Quo Vadis is hosting a live, simultaneous online game where every attendee votes on media channels matchups in real time. This virtual event experiment promises to be the most unique content delivery experiment ever created and the most interactive content experience in the advertising space. There is nothing else like it. The gap between your voting instincts and the crowd’s is where the real learning happens. You walk away smarter, with a full breakdown of how the room voted, and get a results transcript in your inbox after the online event.

Join The Crowd on June 11 at 10AM ET

The end of headcount as we know it

There was a time when growth in technology followed a simple equation: More revenue required more people. The more an adtech company (or agency) grew, the more people it hired. As sales teams scaled and engineering teams expanded, the organizational chart (along with lower-cost outsourced staff) sprawled in lockstep with top-line growth. That basic equation has held for a long time.

Between fiscal year-end results in February and the latest Q1 earnings from both Google and Meta, it’s becoming clearer that the old equation no longer holds. Revenue for both companies is still growing at a double-digit pace, but headcount is flattening. We think the gap is one of the most important economic shifts in modern business (and a leading indicator for the sector), particularly for two companies that produce 45%+ return on invested capital and deliver very healthy operating profit margins (Google at 33% and Meta at 45%).

The first signal is evidenced by both companies reaching what we describe as headcount maturity.

Google is approaching peak employment with ~190,000 employees and is now growing its workforce at roughly 2% annually. Meta has ~78,000 employees and will have ~70,000 after the May layoffs are complete, aiming to reduce staff by ~10%.

As our Quo Vadis readers know, we sometimes turn to historical S-curves because they generally give a good indication of what’s true. In our view, reaching peak headcount maturity at Google and Meta should not be viewed as cyclical belt-tightening nor a response to macro conditions. Instead, we think it characterizes the natural end state of a two-decade-old S-curve that followed the typical path of new innovation (e.g., search, social), build (easy-to-buy), scale (eyeballs), optimize to zero marginal cost, and reach maturity as massive cash-flow machines.

Susan Li (Meta’s CFO) reinforced the new circumstances on the 1Q earnings call:

“We ended Q1 with over 77,900 employees, down 1% from Q4 as the impact of headcount optimization efforts in certain functions was partially offset by hiring in priority areas of monetization and infrastructure.”

And more poinently, “Compute is becoming increasingly important as it determines the quality of the services we can provide, including powering more capable models and delivering innovative new products. It is also becoming more critical to how we work at Meta, as we are entering a world where employees are managing agents to help them generate new ideas, run experiments, execute tasks, and build products.”

Revenue Productivity

In 2025, Google generated just over $2 million in revenue per employee, and around the same amount if we examine only its advertising business (search, YouTube, display, etc.), more on that subject below. Notably, Google’s advertising revenue as a percentage of total revenue has fallen from 83% to 73% since 2019, while cloud has grown from 11% to 15%.

Meta’s business remains 98% ad-based, generating just over $3 million in revenue per employee. On its current growth trajectory, assuming reasonable revenue growth (e.g., 15% CAGR) and near-flat headcount over the next few years, revenue productivity will approach ~$4 million soon and ~$5 million+ by the end of the decade.

The mechanism behind this shift should not be mysterious to our readers. For Google and Meta, AI is a labor multiplier on top of strong operating assets, with employees (including lower-cost outsourced staff) increasingly operating less as executors of discrete tasks and more as system managers of agents that generate ideas, run experiments, write code, optimize campaigns, and build new products. Work that once required coordinated teams now collapses into a totally different orchestration of workflows.

Critically, the productivity gains are happening internally just as much as externally. The same models that power consumer-facing products are completely reshaping how these two companies operate internally and make decisions.

Historically, labor was the dominant input and largely fixed with a high correlation to revenue growth for both companies. Today, that relationship is flipping. Labor is becoming less fixed, while computing is becoming the scaling variable. The more that any given output is required, the more computing resources, rather than people, will be deployed.

But unlike labor, computing ability benefits from cost deflation and simultaneous rapid performance improvements over time. The result is an asymmetrical outcome in which variable costs will likely rise in absolute terms, but unit economic costs will decline on an increasingly lower fixed-cost base (lower staff costs).

So, if you’re wondering how Google and Meta will likely capture future benefits from massive capex on AI infrastructure, we recommend considering how the new variable-to-fixed cost paradigm will allow them to materially expand operating profit margins (cash in flows), thereby justifying all the capex with incremental cash flows. Given that Google and Meta generate ~45% and ~50% return on invested capital, respectively, and deliver massive economic profit spreads, we could make a strong argument that both companies are actually underinvesting in AI capex.

How about the broader market?

Zooming out to the broader market presents a stark contrast. We took a random sample of 46 S&P 500 companies and examined both revenue and headcount from 2010 onward (see list below in the footnotes1).

The average company has grown revenue at roughly 3.2% annually since 2010, with headcount growing at about 1.4%, and revenue per employee sitting below $500,000.

Compared to leading-edge, highly entrepreneurial companies like Google and Meta, other large “laggard” enterprises still operate with human-centric workflows designed for coordination, oversight, and incremental efficiency gains.

We can imagine these companies as slow-moving cargo ships. They set a course and sail in a straight line. They are hard to turn or change direction, because cruising in a straight line is the default risk-averse strategy.

If you've worked in a large organization, you've no doubt felt the pain of human-based process flow, process design, and redesign over the last few decades. As far as we can tell, the current process flows that most companies run today are incongruent with how AI works.

In other words, we imagine a phase currently underway in which large enterprises are getting (or are going to get) completely reimagined from their old processes into a new way of thinking that fits with how AI functions. Over the next few years, we'd expect to see a meaningful increase in revenue productivity across the board, as well as a meaningful expansion of operating profit margins.

• Revenue growth grows in line with expectations (more or less)↑

• Fixed headcount costs decrease↓

• Revenue per employee↑

• Variable costs increase with AI compute costs↑ (and eventually get reported as a new GAAP line item)

• In summary, fixed costs decrease while variable costs increase, thus disproportionately benefiting the bottom line.

For Google and Meta, compute has become the central axis of capital allocation because the key question is no longer how many people to hire but how much infrastructure to deploy. We see companies like Google and Meta as leading indicators of what’s to come for all companies.

A number that will surprise many

When you compare advertising revenue productivity across the sector, The Trade Desk emerges as the frontrunner compared to Google and Meta.2 With an estimated $13 billion in gross billings through its platform and 3,843 employees, TTD’s revenue per employee exceeds that of both Google and Meta.

This is not because TTD has suddenly discovered AI leverage ahead of every other adtech company, but because it has been designed for efficiency from the outset. With so much attention on the mega players in the ad business (Google, Meta, Amazon), we suspect TTD’s revenue productivity goes mostly unnoticed.

What about agencies?

We’ve already established that the average revenue per employee across the S&P 500 is roughly $400K, yet across the very agencies those companies rely on for the advertising job done (Publicis Groupe, WPP, OMC/IPG, Stagwell, Havas, Monks), revenue productivity is closer to $160K.

Large advertisers generally operate with economies of scale and pricing power. They own the assets. They own products, audiences, and/or distribution to capture economic rents.

Agencies, by contrast, are a labor-based service layer. Agency headcounts tend to scale with sales (excluding Publicis as the best “rent collector” model across the agency scene), while their pricing is constrained by procurement, benchmarking, and CFO ROI scrutiny. In other words, agencies sit within the client’s cost base, not at the point of value capture, as they used to during the Mad Men days. This is precisely why any agency that can turn its value proposition into a driver of measurable and provable value capture (e.g., through outcome-based compensation) is in a winning position.

Up until recently, agencies have not been designed to maximize revenue per employee. They were designed to minimize client cost per loosely defined outcomes (or none at all). Their role was, and largely still is, to absorb complexity, aggregate expertise, get volume discounts, and convert what would be fixed costs for the client into semi-variable costs from the client’s perspective. If we boil it down, agencies basically operate as an outsourced execution layer to which procurement can flex up or down.

If you’re wondering why clients take this outsourced approach, I asked Brian Wieser the same question while sitting at an outdoor cafe in Cologne, Germany, during DMEXCO in 2015. His answer was brilliantly zen:

“Because agencies are the least-worst alternative.”

At face value, that sounds horrible. But in reality, it’s one of the best market positions a firm can possibly be in. Why? Because it implies switching doesn’t solve the problem. Clients rarely choose an agency because it's perfect. They’re choosing it because the trade-offs embedded in the other alternatives aren’t worth it.

Secondly, agencies don’t need to be exceptional every day. They just need to be consistently better than the next-best option (in-house, consultants, platform team, etc.). That’s simply a much easier position to defend over time than trying to be “best in class.”

Moreover, being the least-worst alternative creates a stable equilibrium.
When all players have flaws, and no clear dominant substitute exists, the market settles into a classic Nash equilibrium of acceptable satisfaction/dissatisfaction.
The reason clients don’t defect from agencies, even though they switch agencies all the time, is that no one expects to materially improve their outcome.

What does the future hold?

We believe large enterprises are highly likely to continue relying on agencies into the foreseeable future. However, while agencies are good at optimizing execution, they do not own the underlying economics. We think agencies will follow Publicis’ lead and take advantage of AI not just to increase revenue productivity, but to shift their business model from being farmers to landowners (data owners that collect recurring rents from clients in exchange for proven/believable outcomes).

Shifting from farmers to landowners is mission-critical for agencies to remain relevant in the coming years amid new entrants like AI tools and platforms with built-in AI. As we sit today, the new alternatives may look superior in isolation, but to displace the incumbent agency, they must be less flawed across the full stack (strategy, execution, scale, and integration).
That’s a much higher bar than most practitioners realize.

As new AI workflows across planning, trafficking, measurement, reporting, and optimization (in media and creative flows) become partially or fully automated, agency revenue will likely continue to grow at current rates while headcount declines naturally. Agencies have historically experienced relatively high annual attrition (~15–25%), so we expect them to leverage natural churn to gradually reduce labor intensity, thereby increasing revenue per employee toward a new, higher state.

The go-forward model is no longer “bigger is better.” It seems more likely that agencies will become thinner and serve clients as software-augmented coordination layers that also happen to play a market-maker role with trading inventory as principals.

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

One of the biggest topics across the advertising industry is the potential scale of OpenAI’s advertising business. Late last week, Axios reported the scoop that the company projects $2.5 billion in ad revenue this year and $100 billion by 2030. Reaching $100 billion over the next five years would mean OpenAI is the fastest-growing company in the history of advertising. Whether reaching such an amazing achievement is plausible, possible, or probable is what we cover in this post.

In January, Quo Vadis used Google search usage history and diffusion curve modeling to estimate $80 billion in 2030 ad revenue for OpenAI. A few weeks ago, Madison and Wall used headcount comparables and estimated that:

“OpenAI has the underlying scale to support as much as $3 billion in global ad revenue this year and significantly more in 2027.”

What do we know?

Let’s start with what we know or believe to be true.

Valuation: We know that investors valued OpenAI at $852 billion in its last financing round (post-money).

Comparables: We believe the two closest historical comparables investors have in mind for understanding OpenAI's future are Google Search and Meta, the world's two biggest advertising platforms. Together, the two companies attracted 40% of global ad spending in 2025 (estimated at $1.1 trillion by Quo Vadis).

OpenAI Revenue Segments: According to the Wall Street Journal’s recent article, An Inside Look at OpenAI and Anthropic’s Finances Ahead of Their IPOs, OpenAI estimates $140 billion in 2030 “Consumer” revenue (Chart 1). Given the recent news suggesting $100 billion in 2030 “Advertising” revenue, we can infer that the company is forecasting non-advertising revenue (e.g., consumer subscription revenue) will represent 30% of consumer revenue. We believe 30% is a bit high, as only ~5% of ChatGPT’s estimated 900 million active users are paid subscribers today.

Speed of Growth: OpenAI believes ad revenue will grow from zero to $100 billion over the next five years. It took Google 20 years to reach that milestone, and it took Meta 16.5 years (Charts 2 and 3). Whereas OpenAI has the benefit of growing on top of a mature global installed base, Google and Meta grew up with the Internet (and vice versa).

CAGR Comparison: Google’s search ad revenue CAGR over its first twenty years of advertising life was 54%. Meta’s ad revenue CAGR over the first 16.5 years to $100 million was 63%. OpenAI’s CAGR will be 151% if it manages to grow from $2.5 billion in 2026 to $100 billion in 2030, which suggests its adoption curve will reach maturity in less than half the time it will take Google/Meta to reach terminal growth levels.

Terminal Growth: If we assume terminal growth rates in the advertising world are ~5% (the average expected growth rate of global ad spending), Google search will hit that level around 2032, and likely maintain the same growth rate into perpetuity as a GDP-like compounding cash flow stream well-anchored to the continued expansion of global advertising demand. Meta is younger with a different future. We expect Meta to grow at a ~14% CAGR over the next five years with $390 billion in ad revenue in 2030.

Economic Moats: We know that Google and Meta produce relatively high and consistent returns on invested capital (ROIC) of 40%+. The higher the ROIC, the faster the printing press prints cash flow. We also know that Google and Meta have generated an average operating profit margin of 30% (EBIT less adjusted cash taxes) over recent years, with little variance. We believe investors expect OpenAI to benefit from a similar economic moat.

Piecing the puzzle together

OpenAI’s investors are smart and experienced. They have access to good information and see a lot of deal flow. For them or any investor, the valuation they assign today is nothing more than the present value of expected future cash flows. If investors in OpenAI consider it as the next big thing after Google and Meta, we can use what we know to back into OpenAI’s advertising revenue. We’ll do that step by step.

Step 1: We know that the value of future cash flows is determined by the following finance formula:

For OpenAI, we know the most recent valuation is $852 billion. From what we can tell, investors are comparing future financial performance to that of Google and Meta, so we think it’s reasonable to use 40% as a plug for future ROIC at maturity.

For Cost of Capital, we think OpenAI is best compared to major ad platforms with betas around 1.2, as well as higher-beta companies in the AI space like Nvidia at 2.5. We lean more toward Nvidia and use a weighted average of 2.1.

As we mentioned above, we think a ~5% terminal rate is a reasonable enough result in the following value equation:

Step 2: Now all we have to do is rearrange it and solve for Operating Profit.

And now, with our reasonable plug values, we estimate OpenAI's Operating Profit at $88 billion at some point in the future, when it reaches maturity.

For context, Google’s operating profit in 2026 will be ~$125 billion (includes advertising, cloud, other bets, etc.) and Meta’s will be around $80 billion. If we assume 30% operating margins at maturity for OpenAI, total revenue at that future time will be $293 billion. Note that WSJ reported $280 billion in total revenue by 2030.

Step 3: Now that we have a good estimate of OpenAI’s Total Revenue at maturity, we can use the WSJ’s revenue segment input guides to arrive at a future ad revenue of ~$209 billion. The question when maturity will be reach is the billion dollar question.

If 20% of OpenAI’s future total revenue comes from Enterprise sales (Google Cloud is 17% of total revenue), and 5% comes from what they call “New Products,” that implies 75% or $220 billion will come from “Consumer Products.”

Finally, if 5% comes from subscriptions, then investors implicitly expect OpenAI to have ~$209 billion in ad revenue at maturity.

When is maturity?

If $209 billion is the maximum number in OpenAI’s future mature state, and the stated goal is to reach $100 billion by 2030, then maturity will happen in about 10 years. The the “best-fit” diffusion curve given our current data points.

If that happens, OpenAI will have grown 3x faster than Google and Meta’s journey to maturity. At first glance, achieving advertising success so quickly might seem plausible but not probable. However, when compared with the first 10 years of growth at Google and Meta, it not only seems plausible but perhaps quite probable even though it is hard to imagine.

Google search experienced a 109% CAGR over its first 10 years, and Meta was similar at 93%. If OpenAI’s forecast is right and it hits $2.5 billion in ad revenue this year, and reaches maturity in ~10 years, then its CAGR will be 62%. Looking at OpenAI’s advertising future through our lens turns what is plausible and possible into likely probabality.

Try Our OpenAI Revenue Growth Estimator Tool

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Let’s say we are examining AdTech Companies A and B. On the surface, Company A appears to be the clear winner, with faster revenue growth and a higher Rule of 40 (revenue growth + EBITDA margin).

Company B, by contrast, has different optics at first glance. Slower revenue growth, but better EBITDA margins and a slightly lower Rule of 40. Beneath the surface, Company B’s economics show a more efficient adtech machine.

The first apples-to-apples mistake observers tend to make is miscalculating EBITDA margins off the wrong “revenue” number.

• Company A reports actual revenue earned ($2.9 billion in 2025) from processing gross ad spend. EBITDA was $697 million, so its EBITDA margin is 24%.

• Company B reports revenue in terms of gross spend, $1.9 billion in 2025. If we incorrectly apply it to EBITDA, we get a 17% margin. Company B’s actual revenue (aka, ex-TAC contribution) is $1.2 billion, so its correct apples-to-apples EBITDA margin is 27%.

The second mistake is caring about EBITDA in the first place. EBITDA is only somewhat relevant if you’re interested in pricing a company (e.g., using comparable company multiples to “value” a target company). EBITDA is widely used because it is easy to calculate, but it is insufficient for properly valuing a company.

Let’s start with Company A

Reporting Guidelines: Company A reports under a GAAP “Agent” model. That means it only records the cut it keeps (its commission or take rate) as revenue, not the full amount of money flowing through its platform.

Revenue: Company A’s revenue in 2025 was $2.9 billion. YoY revenue growth was 18%. Not too shabby. Consensus estimates for 2026 revenue growth range from 13% to 15%. The company faces multiple headwinds, making the low end of the range a real possibility.

Gross Ad Spend: Company A processes or “sees” around $14.3 billion in media flows (~20% disclosed take rate). Before the 2025 reporting year, the company used to provide investors with information about the volume of gross ad spend it processes, but it seems management decided to remove past transparency. In our view, the main reasons likely reflect a combination of mix shift, reporting simplification, and/or underlying volume/take rate dynamics. In any case, we use either accounts receivable (audited number) or accounts payable (audited number) and look at historical days receivable or days payable to get a good enough idea of gross ad spend.

EBITDA and EBITDA Margin (Earnings Before Interest, Taxes, Depreciation, and Amortization): Company A generated $697 million in EBITDA in 2025. Creating EBITDA margins in adtech land north of 20% is not easy. Company A generated 24% EBITDA margins in 2025.

Capitalized R&D Expense: One of the first steps in valuing any tech company is to depart from GAAP by capitalizing all R&D expenses. This approach provides a more accurate view of operating profits. Given the state of rapid technological change, we amortize on a 3-year schedule

Here’s what NYU valuation guru, Aswath Damodaran, recommends, and we follow his advice:

“Expensing R&D leads to lower earnings and book values for firms that invest heavily in it, making them look less profitable than they truly are… by treating R&D as an expense rather than a capital investment, we systematically understate both earnings and invested capital for growth firms.”

Read Damodaran’s persuasive view in: “Research and Development Expenses: Implications for Profitability Measurement and Valuation Aswath Damodaran, Stern School of Business.

So, after we capitalize R&D expenses, the net effective is to add back $108 million in R&D expense to Company A’s EBTDA, then subtract $49 million. The net effect is $59 million in additional EBITDA gains for Company A, bringing total EBITDA to $756 million. Now, in real operating terms, Company A’s true operating EBITDA margin is 26%.

Rule of 40 is a finance rule of thumb and a quick way to gauge whether a company is balancing growth and profitability well. You simply add the revenue growth rate to EBITDA margin. If the total is 40% or higher, the company is generally considered “healthy.” As a company matures, revenue growth is high in the early years and eventually decays toward GDP growth (e.g., ~4%). EBITDA margins behave in the opposite direction, starting low or negative and growing over time. For Company A, its Rule of 40 in 2025 was 44%.

Turning to Company B

Reporting Guidelines: Company B reports GAAP revenue under both the “Principal” and “Agent” models. When a company acts as a principal, that means it serves as a middleman, buying and reselling the product. In this case, GAAP requires recording the full sale amount as revenue and payments to suppliers as costs, leaving the difference as actual revenue (aka, ex-TAC contribution).

When we determine that we act as principal, we recognize revenue and related costs incurred on a gross basis. When we act as an agent, we recognize revenue on a net basis.

Revenue: Company B’s actual operating revenue in 2025 was $1.2 billion. YoY revenue growth was 5%. Keep in mind, Company B is four years older than Company A, and revenue growth rates for every company ever created always decay over time. We’ll come back to this important value driver later on in the article.

Gross Ad Spend: Company A operates under a single business model. Company B runs a diversified business model across two revenue streams. On its principal-based stream, it processed $1.7 billion in gross ad spend in 2025, netting $915 million in operating revenue (principal-based accounting). Its agent-based revenue stream generated $260 million, so total operating revenue in 2025 was $1.2 billion.

EBITDA and EBITDA Margin: GAAP EBITDA for Company B was $313 million in 2025, resulting in a nice 27% EBITDA margin. After we capitalize R&D expense and smooth out amortization as we do for Company A, Company B’s actual EBIDA margin is 28% compared to 26% for Company A.

Rule of 40: Adding revenue growth and EBITDA margin, Company B delivered a healthy 33% in 2025 compared to Company A’s at 44%.

A better measure of true health

“Every time you see the word EBITDA, you should substitute the words “bullshit earnings.”

— Charles Munger and Warren Buffet

Warren Buffett and Charles Munger famously despise EBITDA, viewing it as a misleading metric that ignores real cash expenses like capital expenditures. They argue it allows management to inflate earnings and mask a company’s true, often decaying, financial health.

A better measure of any company's true health results from doing the grunt work to understand the historical spread between return on invested capital (ROIC) and its cost of capital. Using EBITDA to price a company misses the two most important things to properly value a company:

1. Economic Spread: A company can create value only when its ROIC exceeds its cost of capital. Both companies A and B are creating value for shareholders, although one is better than the other.

2. Fundamental Growth: This is the rate at which a company can grow its operating profits, which is solely determined by how much of its operating profits it reinvests (Investment Rate) and how efficiently that capital generates returns (ROIC). In plain terms, growth isn’t just about investing more; it’s about investing well.

Before we compare Companies A and B on a true historical health basis, we need to establish two gold-standard benchmarks across the adtech space that any investor can use to allocate capital. The two high bars everyone looks up to are Google and Meta. Both companies consistently deliver ~45% ROIC (amazing). Since Total Share Returns ultimately map to ROIC performance, you can see the results for yourself in our Quo Vadis BigTech2 portfolio.

In the adtech world, Google and Meta have proven to be far and away the best allocators of capital. That’s why they are the gold standard for investors. That’s also why they can announce ~$150B in new invested capital aimed toward AI and get a favorable reaction from Wall St. They aren’t investing in AI” as a project. They are both completely retooling their entire production function around AI. They are defending and rebuilding their respective moats while simultaneously building AI utility layers. Both companies are asking and answering the right question:

How much capital is required to preserve ROIC in a world where intelligence is computational discipline?

The answer is a lot more than in the pre-AI digital advertising era.

Company A:

• ROIC in 2025 was 29%, and its two-year average was 24%.

• Since 2019, the average ROIC for Company A is 20%

• Key Value Metric: Its economic spread relative to its 2-year average ROIC was 14% at an 11% cost of capital in 2025. Overall, the company’s average economic spread since 2019 is 8%. The economic spreads for Google and Meta (and Publicis too) are north of 30%.

• Company A’s average growth rate of new invested capital since 2019 is 23%.

• Over the same time period, its average revenue growth rate was 28%. Invest to grow, that’s the name of the game.

• In terms of capital efficiency, the company generated $1.40 in revenue per dollar of invested capital in 2025. Not only would investors prefer to see this driver above 2.0, but their expectations are also heading to the 3.0 range with AI in the mix.

• Company A’s current cost of capital for FY25 was 11.6%, averaging 12% since 2019, with minor variation across its capital structure.

• Fundamental Growth: Since 2019, Company A has reinvested 95% of its operating profit into new invested capital, with an average ROIC of 20%, resulting in a healthy fundamental growth rate of 19% (the same rate as revenue growth in 2025).

Company B:

• ROIC in 2025 was 20%, and its two-year average was 29%.

• Since 2019, the average ROIC for Company B is 18%

• Its economic spread relative to its 2-year average ROIC was 10% at an 9% cost of capital in 2025.

• Key Value Metric: Overall, the company’s average economic spread since 2019 is 10%, 2% better than Company A.

• In terms of capital efficiency, the company generated $1.60 in revenue per dollar of invested capital in 2025 compared to $1.40 for Company A.

• Company A’s current cost of capital for FY25 was 8.8%, and averages 7% since 2019.

• Fundamental Growth: Since 2019, Company A has reinvested 25% of its operating profit into new invested capital, with an average ROIC of 17%, resulting in a fundamental growth rate of 4% (the same rate as revenue growth in 2025).

The Future: DCF valuation comparison

Ultimately, what matters most is the forward view because valuation is driven by the future stream of free cash flow a business is expected to generate, not what it has already done.

Company A

• Current market cap (equity value) is $10.1 billion.

• Given multiple headwinds, we forecast 13% revenue growth in 2026 and 9% CAGR through 2030.

• Company A generates $750K in revenue per employee. We expect this key value driver to improve in 2026 with investments in labor productivity, resulting in flat employee growth.

• We expect continued improvement in operating profits, which implies ~30% EBITDA margins in 2026.

• For 2026, we expect the company's ROIC to rise from 29% to 33%, and reach 37% by 2030.

Equity Valuation: Given $1.3 billion excess cash and $437 million in debt + debt equivalents, our fair market valuation for Company A is $11.4 billion. Company A’s stock is trading at $10.2 billion today, so we think investors are slightly undervaluing the company’s future prospects.

Company B

• Current market cap (equity value) is $857 million.

• Consensus estimates imply 2% revenue growth in 2026. We think management will easily deliver 5% growth in 2026 and maintain that rate through our 2030 forecast period.

• Similar to Company A, we expect continued improvement in operating profits driven by AI labor productivity, implying 32% EBITDA margins.

• For 2026, we expect Company B’s ROIC to rise from 20% to ~23%, and reach 31% by 2030

• Rule of 40 by the end of 2026 will tighten: Company A at ~43% vs. Company B at 37%.

Equity Valuation: Given $250 million excess cash and $155 million in debt + debt equivalents, our fair market valuation for Company B is $3.4 billion. Company B’s stock is trading at $860 million today. From our vantage point, we think investors are massively undervaluing the company’s future prospects on what are fairly easy hurdles for management to achieve.

The Big Reveal

Bottom Line: Fundamentals and market narratives don’t always move in lockstep, but they eventually reconcile. Markets often misprice trajectory, but they rarely misprice destination over time.

For The Trade Desk, the gap between market value and fundamental value has historically been wide, reflecting periods of both exuberance and skepticism. But today, that gap is narrowing. As free cash flow scales and reinvestment efficiency improves, investor sentiment appears to be converging toward the business's underlying economics.

For Criteo, the opposite dynamic is emerging. While fundamental value, driven by steady free cash flow, remains stable to improving, market valuation is still lagging (for now). From what we can observe, investor sentiment is diverging from operating performance, suggesting either a lack of conviction about durability (wait-and-see mode) or a meaningful and temporary mispricing of cash-flow quality.

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

We have something special in store at the Advertising Economic Forum this Thursday at The New York Times Center. We partnered with AdWeek to create a live interview studio and talk about a single important subject.

Back in 2024, when we first launched the forum, an important question arose on stage that caused a ton of debate:

Will there be websites in the future?

Flash forward to March 2026, and we can all see that websites are still around, albeit with reduced traffic and pressure to tweak the publisher business model as fast as possible in the AI era.

Throughout the morning this Thursday (March 19), we are handpicking ~25 attendees to ask them a simple question in 5-minute interviews. One question, many points of view.

“What is the future of the open web? And how must it reinvent itself to increase its relevance in the advertiser’s mind?”

A few sub-questions might be:

Why should advertisers care about the open web? What has to change for the open web to stay relevant as an adverting medium?

If you want to chime in at our AdWeek interview studio, get your ticket today.

Get Your Spot Today

History of the Open Web told through the lens of supply and demand

In the Beginning, 1994-2005: Let’s start with the initial equilibrium state of a healthy market for open web banner ads in 2005.

Website display ads started on October 27, 1994, when the first web banner ad appeared on HotWired.com (the online counterpart to Wired magazine). It was created by agency Modem Media for AT&T. The ad featured a 468x60 pixel banner with the text:

“Have you ever clicked your mouse right here? You will.”

By 2007, Advertiser demand grew, and publisher supply kept pace

By 2007, demand and supply looked normal. Total ad spend on web banners was roughly $15B (the grey rectangle). In the first era of digital advertising, most ad-supported websites were focused on the “onlining” of old-school offline newspapers and magazines. There were also new digital-first publications.

CPM prices were around $10 (P) on a weighted average, for example:

• Premium publisher display = $15 – $30 CPM

• Run-of-site banners = $5 – $15 CPM

• Ad network inventory = $1 – $5 CPM

• Remnant/performance networks = $0.50 – $2 CPM

Back then, in 2007, there were around 1.2 billion internet users. Today, there are over 6 billion. That means quantity sold (Q) was around 1.5 trillion impressions or about 3.5 ads per day per user on average. Who knows what the number is today? All we know is that users get tons of open web display ads, of which many they never see.

Programmatic is Born: ~2007 to 2016

When programmatic came along, the supply curve started to flatten and shift because it was easy to produce new inventory. Like alchemy that produces false gold, there was an increasing ability for any publisher, big or small, to create infinite audiences, new traffic, and new supply.

As a result, CPM prices for open web banner ads started to fall while the magic elixir and allure of “audience targeting” grew at a robust pace, increasing the total market size.

Supply is Flat: 2016 to 2023

When something is too good to be true, it probably is. By 2023, open web display ad sales were mostly done programmatically between DSPs and SSPs with data providers in the middle.

Even though supply was essentially flat, demand continued to grow despite warnings from various industry associations, with one “working media” research report after another. Quo Vadis wrote about this in an article called Funky Town and The Way The [AdTech] World Works. The article describes how demand was procurement-led, and suppliers simply produced what they wanted — “cheap reach.”

Present Day 2026

Just as market theory and the lemon market theory would predict, demand from advertisers has run out of patience.

Now they are pulling back and shifting attention to the GenAI world, where eyeballs are increasingly flowing.

Good news taking form in 2026 and 2027

The interesting thing about LLMs is their need for good information. Good sites provide useful information, bad sites don’t. The more users spend time in chatbot environments and use agents to complete tasks, the more LLMS will need to eat the good stuff.

We can also envision a future in which LLMs will have to account for a Cost of Goods Sold (COGS) expense called “publisher data acquisition cost.”

As of 2025, total global ad spending was ~$1.2 trillion. We think it’s safe to assume 6% annual growth over the next five years, so total global ad spend will be ~$1.6 trillion in 2030. Note that Brian Wieser from Madison & Wall is forecasting 8.1% this year.

That means $400 billion of new media money is coming online over the next five years. We don’t think it is hard to imagine GenAI ads on OpenAI and other apps reaching $100 billion by 2030. If we assume a 5% COGS going to publishers, that’s $5 billion extra on top of ad revenue.

The Case for 2028

The more publishers convert old programmatic thinking into better advertising propositions, the quicker the supplier curve will recover and return to a normal state, like it was way back in 2000.

At the same time, advertisers will start to see the open web in a new light and, hopefully, reward great outcomes with increased demand.

Lower revenue with more profit is a great outcome

Here’s a test for publisher management teams:

Would you rather have $100 in ad sales and $10 in operating profit OR $60 in ad sales and $13 in operating profit.

• The supply curve recovers to a normal state in the Recovery Stage.

• Low-quality publishers get boxed out of the market.

• Some variable costs ꜜdecrease (TAC, tech fees).

• Other variable costs ꜛincrease (data movement, AI usage).

• Add in LLM licensing revenue (COGS for LLM, revenue for publishers).

• All fixed costs decrease primarily due to increased AI-assisted labor productivity.

Result: A smaller market universe, but more profits.

If you want to chime in at our AdWeek interview studio, get your ticket today.

Get Your Spot Today

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Get Your Seat To Enlightenment Today

Preview of Advertising Economic Forum Agenda with Brian Wieser

Yesterday, we sat down with Brian Wieser from Madison and Wall to get his take on the agenda/speaker sessions taking pace at the Advertising Economic Forum. Similar to part years, the agenda is like a storybook that we unfold one session at a time. In our video review, we quickly run through what’s in store for the session and offer little tidbits and observations along the way.

Enjoy! See you next week!

Get Your Seat To Enlightenment Today

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Get Your Seat To Enlightenment Today

4Q25 and FY25 AdTech, MarTech, Agency Update

With 4Q25 and full-year 2025 earnings now reported (except RAMP, fiscal year ending March 31), we can assess how our equal-weighted portfolios have performed since our Q3 update in November.

As a reminder, each portfolio starts with a $100 investment divided equally across all constituent stocks. If a company exits public life (e.g., IAS) or a relevant company goes public (e.g., MNTN), we remove or add it and rebalance on an equal-dollar basis.

Here’s how it works using our AdTech11 portfolio as an example. You start with $100, divide it equally across 11 companies, and buy shares in each based on the stock price on January 1, 2022. If things go well, you find alpha by beating the NASDAQ or S&P500.

Key Benchmarks: The NASDAQ and S&P 500 are both up ↑43% since January 2022. We also include our BigTech2 (Google and Meta) to benchmark what top “durable moat” performance looks like for companies with incredibly deep and defensible economic spreads.

What is a moat? Quo Vadis defines a “moat” as the spread between a company’s return on operating invested capital (ROIC) and its cost of capital (aka, economic spread). In other words, the criteria for success is the economic spread and the measure of success is ROIC > Cost of Capital.

For Google and Meta, that all-important spread is 30% or more, depending on the quarter. With top quartile spreads, both companies are happy to make giant investments in new invested capital to sustain their moat in the AI age. In our Portfolio Chart chart above, investors are basically saying, “Earnings talk is cheap. Get a sustainable moat or we’ll penalize you.”

AdTech11

Our AdTech11 portfolio is ↓54% since inception. Note that IAS was taken private by Novacap for $1.9 billion and removed from the portfolio in January 2026. MNTN was added in June 2025 following its IPO, currently trading at a $720 million market capitalization and $510 million enterprise value. Following its IPO in May 2025, the “CTV” company had a market capitalization of over $2 billion.

Overall, AdTech11 was trending upward toward the end of last year but has since reversed course and is now trending downward. In other words, the same $100 invested in a NASDAQ ETF would have returned 43% with less risk.

Gainers: ZETA (+96%) and DSP (+1%).

Downers: MGNI (-24%), RAMP (-46%), CRTO (-54%), TBLA (-59%), MNTN (-63%), DV (-68%), TTD (-74%), PUBM (-76%), and TEAD (-94%).

AdTech11+ (with AppLovin)

When we include AppLovin (APP) in the mix, our AdTech11+ portfolio is only ↓33%. APP alone is up ↑361%, single-handedly carrying the portfolio.

MarTech8

Our MarTech8 portfolio remains deeply in the red, ↓57%. The portfolio includes SHOP, HUBS, SPT, SEMR, TWLO, BRZE, YEXT, and AMPL. It turns out that SaaS revenue models are not superior to media percentage models for generating cash flow.

Shopify (-11%) used to dominate our MarTech portfolio and contribute to nearly all the gains. From our perspective, this shows how disruptive AI has become in investors' minds, particularly when they consider SaaS-based businesses.

BigTech2

We track two Big Tech ad players, Google (GOOG) and Meta (META). Had you invested $100 on a 50/50 basis in January 2022, you’d be ↑103% GOOG (+115%) and META (+91%), respectively.

Between the two companies, they controlled 40% of global ad spending in 2025. Given reasonable growth expectations, we estimate their combined share will increase to 46% by 2030.

SocialTech4

Our SocialTech4 portfolio is ↑18%. The portfolio includes META, SNAP, PINS, and RDDT. Reddit (RDDT) was added in April 2024 following its IPO. Reddit is the biggest winner, up 217% since entering the portfolio.

Meanwhile, Pinterest made news last quarter with its acquisition of TVScientific. The company reported a bullish outlook for the first quarter, citing the impact of expanding its connected-TV capabilities within its Performance+ AI optimization portfolio.

Agency5

Our agency portfolio consists of WPP, Publicis, Omnicom, Stagwell, and S4 Capital (SFOR.L). Note that IPG was removed following its acquisition by Omnicom, making the combined agency the largest in the world and displacing Publicis (for now).

The Agency5 portfolio is down ↓43% since inception. Publicis and Omnicom are the only agencies in the black, up 33% and 15%, respectively.

They say a picture is worth a thousand words. We don’t need a thousand words. We only need seven: creating and sustaining an economic moat matters. From our vantage point, Publicis has done the best job of converting itself from a farmer to landowner (data owner). Whomever owns the best soil (data) in advertising land gets to build the biggest and durable moat. That seems to be Publicis’ acquisition strategy in a nutshell.

A word on Betas

Looking at betas across our AdTech and MarTech portfolios, they are relatively high, ranging from ~1.3 to ~1.8. High betas materially affect the cost of equity and the overall cost of capital. In a tightening cycle with AI opportunities and threats everywhere, players with higher discount rates disproportionately affect efforts to create long-duration cash flow streams. The result is valuation pressure and multiple compression, even if operating performance remains intact.

If you want to hear what the experts have to say about topics like this, then come to the Advertising Economic Forum NYC on March 18 and 19.

Get Your Seat To Enlightenment Today

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Get Your Seat To The Fun Today!

Advertising Economic Forum Returns to New York City: March 18–19, 2026.
Day 1 at Horizon Media features eight AI founders pitching to our esteemed investor panel across agentic commerce, creative, and emerging agency models. Day 2 at The New York Times Center convenes 400 senior leaders for closed-door sessions examining market structure, incentives, and the economic forces reshaping advertising today. Early-bird registration is still open for 3 more days; space is limited.

Get Your Spot Today

Rebuilding the media value chain for an agentic AI world with Andrew Mole.

Quo Vadis had an insightful chat last week with Andrew Mole about how the media value chain is getting rebuilt for the agentic AI world. To set the scene, we began with a topic dominating the adtech/financial news cycle — Google’s giant AI capex bet.

Google told investors on its Q4/FY25 earnings call that it plans to invest $180 billion in capex (aka, operating fixed capital) over the near-term investment cycle. Brian Wieser from Madison & Wall covered this story last week (I recommend checking out his perspective).

Let’s put $180 billion of new investment in fixed assets into perspective. At the end of 2024, Google had $210 billion in total operating invested capital. $133 billion was fixed capital, and $77 billion was working capital.

With $180 billion in new investment, we interpret this to mean Google is aiming to replace its entire capital base (hardware, compute, etc.).

Case and Point: As far as we can, Google’s giant investment in new fixed capital is signaling a bet that AI tech and adoption are not only happening fast (obviously), but it’s going to happen even faster than the majority of the market thinks. In other words, it’s better to overinvest and be right than underinvest and be wrong.

4 Questions We Covered With Andrew

Check out the video recording to hear his answers and related commentary.

Q1 with Andrew: Why is AdCP (Predid Sales Agent) the most consequential shift across programmatic?

Q2 with Andrew: Prebid.org has taken over the code for the Ad Context Protocol (AdCP) to develop the Prebid Sales Agent, an open-source framework enabling publisher-side AI to autonomously manage, negotiate, and execute ad campaigns. This initiative aims to prevent vendor lock-in and standardize AI-driven advertising, providing an alternative to proprietary AI solutions while ensuring interoperability across the digital ecosystem. Good, Bad, or Indifferent?

Q3 with Andrew: Everyone is talking about “agentic AI,” but the industry is split on what actually changes: workflows, incentives, control points, and who captures value. What is structurally different in this cycle?

Q4 with Andrew: What does agentic unlock in today’s publisher stack to make curation scalable rather than manual and work in the publishers’ favor to avoid the sins of the programmatic era?

The Ideal State

We also covered the trampled topic of “transparency.” In adtech land, Quo Vadis treats the word “transparent” as synonymous with “trust.” Then we take it one step further and define trust as:

Confidence in a counterparty’s intentions.

This concept is important to prevent the agentic era from repeating the lemon-market mistakes of the RTB era. As our Quo Vadis community knows, the RTB era presented a classic “lemon market” asymmetry, where the sell-side had all the information about the quality of the impressions being offered. When marketing and procurement teams thought it was a good idea to allocate 20% or more to $1.00 “cheap reach” inventory, it incentivized the supply side to flood the market with subprime inventory. On the other side of the trade, buyers had difficulty distinguishing good from bad quality, so they priced programmatic inventory downward. As with all lemon markets, buyers eventually find equilibrium and exit the market altogether, a trend that became more noticeable in 2025 and continued into 2026.

Translation: The buy side of RTB had lost confidence in the seller’s intentions as the promise of programmatic failed to materialize (efficiency, transparency, quality, etc.).

The agentic world is a fresh start. Time will tell whether old habits and vested interests are hard to break or whether the tsunami of creative destruction ultimately delivers on the original promises. Quo Vadis is hopeful for the latter but paying attention to the former.

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Thanks to our amazing Quo Vadis community, there have been nearly 400 games played on Last Media Dollar and 30,000 individual 1-to-1 media channel matches since launching beta v1.0 as an experiment in September 2025. Thanks to user feedback, we are going from a “generic” simple experimental game to beta v2.0.

Last Media Dollar is a simple behavioral economics game that pits 22 competing media channels against one another in head-to-head matchups to reveal “hidden” preferences when compared to where actual media money flows.

Last Media Dollar “Live” is the most unique event in Ad Land

It’s based on the online game you’re reading about right now. The event is coming to NYC on February 25 at the Roxy Cinema in Tribeca. This is not a panel. It’s a live, on-stage decision-making with Chad Stoller, VP of Media at PMG, and Joe Zawadzki, partner at Aperiam Ventures. Two more speakers will be announced soon! Secure your ticket now to be in the room where the thinking happens.

Get Your Seat To The Fun Today!

Actual Media Money Flows

In September 2025, we published our bottom-up research and estimates across 22 media channels, where nearly $1 trillion in media spend flows globally. Roughly 76% goes to eighteen digital channels, the remainder goes to four traditional channels.

Beta v.20 Game Updates

Simplified landing page: We condensed the landing page to get you into gameplay faster. Users get a quick sense of what the game is, why they should try it, and how to play.

Game Use Cases: The most important finding in our MVP is that playing Last Media Dollar requires context and a defined role. For example, are you playing a game as a brand advertiser or a performance advertiser? Are you spending media dollars as CPG, auto, finance, travel, or other verticals? Just like a real-world media plan, it’s obvious that both matter.

So, on the home page, players can now choose to play one of our 18 “House Cases” across eight different verticals. For example, you might step into the role of a CPG brand like Kraft Mac & Cheese, a luxury automaker like Genesis, or an insurance leader like Progressive. Each case is available as either a Brand or Performance case.

Building Your Own Case and Becoming a “Game Master”: Let’s say you don’t find a suitable House Case. No problem. Now you can use our GenAI helper tool to build your own case. All you have to do is answer a few prompts, and voila, you’ll get your own use case presented in the same clean “McKinsey” style as the house cases. Now you’ve become the “Game Master” of your very own use case(s).

Private Game Cases: As a game master, you can distribute your URL game link to a private group of players. For instance, let’s say you work at an agency and you just landed a new account. You have the client brief and will use it to create your LMD game case. Maybe you end up making it better than it was when you first read it. You send your game URL link to as many colleagues as you want. They play the game, and you get incredible insights that you cannot get today. It’s like asking for a new color to be invented and getting it.

Private Game Cases: Another new feature for game masters is the ability to turn a private use case into a public use case by adding it to the Case Library. For instance, perhaps you are an ad tech founder or investor. You want to know where the last marginal dollar is moving to better understand where durable demand exists. Or maybe you are a publisher revenue leader, and you want to know how buyers truly value your inventory relative to substitutes in order to reposition your offering, pricing, or go-to-market strategy. Now you have an easy way to move, knowing what you know to knowing what you don’t know, but need to, and as fast as possible.

Get Advice on Channel Matches: Stuck on a tough decision? During any match, tap the "Get AI Advice" button to receive instant, context-aware guidance from our media decision coach. The advice takes into account your specific case details, brand objectives, and the unique strengths of each channel in the current matchup. Whether you're weighing the reach of broadcast TV against the precision of digital targeting, or debating between influencer marketing and traditional print, the AI provides balanced insights to help inform your decision—without making the choice for you. It's like having a media planning expert in your pocket.

Take AI Notes for Future Reference: Capture your thinking as you play! Use the notes feature to record your rationale for each channel choice. Type your thoughts or tap the microphone to speak them. Your notes are automatically saved with each match and compiled into a complete transcript on your results page. This creates a valuable record of your decision-making process that you can reference later, share with colleagues, or use to explain your media preferences to stakeholders. It's perfect for turning a quick game session into documented strategic insights.

Get Your Seat To The Fun Today!

Last Media Dollar “Live” is based on the online game you’re reading about right now. The event is coming to NYC on February 25 at the Roxy Cinema in Tribeca. This is not a panel. It’s a live, on-stage decision-making with Chad Stoller, VP of Media at PMG, and Joe Zawadzki, partner at Aperiam Ventures. Two more speakers will be announced soon! Secure your ticket now to be in the room where the thinking happens.

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Last Media Dollar “Live” is coming to NYC: February 25, 2026, at the Roxy Cinema in Tribeca. This is a live, on-stage decision-making experience where senior practitioners and the audience debate assumptions and reveal how media choices are actually made under real constraints, in real time. Secure your ticket now if you want to be in the room where the thinking happens.

Get Your Seat To The Fun Today!

Advertising Economic Forum Returns to New York City: March 18–19, 2026.
Day 1 at Horizon Media features eight AI founders pitching to our esteemed investor panel across agentic commerce, creative, and emerging agency models. Day 2 at The New York Times Center convenes 400 senior leaders for closed-door sessions examining market structure, incentives, and the economic forces reshaping advertising today. Early-bird registration is still open; space is limited.

Get Your Spot Today

Ads are coming to ChatGPT

“OpenAI has announced plans to introduce advertising in ChatGPT in the United States. Ads will appear on the free version and the low-cost “Go” tier, but not for Pro, Business, or Enterprise subscribers. The company says ads will be clearly separated from chatbot responses and will not influence outputs.”

OpenAI listed five principles in their January 16 announcement:

1. Mission alignment: Ensure AGI benefits all of humanity. The company’s economic pursuit of advertising is intended to support that mission and make AI more accessible to more people.

Note that OpenAI has committed to approximately $1.4 trillion in AI infrastructure, including data centers, over the next eight years to support future development. The keyword is “committed,” which we translated to mean capitalized lease commitments, not actual cash outlays.

2. Answer independence: Ads will not influence the answers ChatGPT gives you. Answers will be optimized based on what’s most helpful to users. Ads will be clearly separated and clearly labeled.

3. Conversation privacy: Conversations with ChatGPT remain private from advertisers, and we never sell your data to advertisers. This is absolutely key if OpenAI wants to create a giant economic moat. In other words, their positioning (water-tight-walled garden) is the polar opposite of the “open” Internet.

4. Choice and control: Users control how their data is used. Users can turn off personalization and can clear the data used for ads at any time.

5. Long-term value: OpenAI does not aim to optimize for time spent in ChatGPT. They prefer to prioritize user trust and user experience over revenue. Again, this is key because their atomic unit of production is trust. We translate this to align with Google’s philosophy that utility trumps time spent. As long as consumers use Google’s products like toothpaste (twice per day), growth and free cash flow are guaranteed.

A quick historical look at the expressions of advertising

Despite constant innovation in media and technology over the past 5000 years, there are still only three fundamental expressions of advertising: print, audio, and motion (video). Therefore, every past and future adtech investment ultimately maps to one (or a combination) of these three.

As far as we know, print was the first expression of advertising and still dominates $1 trillion in media flows today. It appeared around 3000 BC and was found in the ruins of Thebes, Egypt. The ad was a papyrus print ad promoting a weaving workshop. Audio was the next expression of advertising in 700 BC China, where candy shop owners paid flute players to attract kids to the store.

Print is the original and still dominant expression (~54% of total ad spend). You can trace its roots from ancient written ads (e.g., early papyrus notices) through the Gutenberg press, early classifieds (Boston News-Letter, 1704), and the evolution from black-and-white to color posters via lithography.

Jumping forward to 1995, digital formats such as display, search, social (including influencer marketing), and retail media ads are mostly just continuations or branches of print. For example, search ads can be framed as a structured “directory” lineage (e.g., Yellow Pages), and retail media really just an evolution of catalog-style merchandising to meet online consumers where they are today.

Audio became the second major expression in the early 20th century, with the first radio commercial in 1922 by AT&T's WEAF station in New York City. It was a 10-minute commercial for the Queensboro Corporation, promoting apartments in Jackson Heights. The Queensboro Corporation paid $50 for the commercial.

Motion (video with audio) was the third expression to come onto the advertising scene. It started with silent-era theater ads (slides and pre-feature promotions) as early motion-era commercialization. The first U.S. TV ad in 1941 (for Bulova watches) was the beginning of modern televised advertising.

TV ad spend experienced hypergrowth in the 1970s and expanded again with cable ads in the 1980s. In 2026, we are well into the decade of CTV ad growth. It’s not a new expression, but rather a dollar-for-dollar migration within motion. Of the nearly $1 trillion in global ad spend in 2025, the majority is a print expression.

With our historical backdrop in mind, we can more clearly see how ChatGPT ads are unlikely create a new form of advertising. Far from it. The ad units will simply be an expression of print (similar to a sponsored listing), audio (when interfacing with ChatGPT via voice prompt and listening), or motion (a video ad experience inside chat results).

What’s Different: What will likely be different is how ads inside chatbot experiences represent a new division of labor (in the spirit of Adam Smith) by breaking the media and creative job-to-be-done into more atomic tasks to create new outcomes for advertisers. This atomic shift will change how advertising (media and creative) is produced, transacted, targeted, served, optimized, and measured with higher-calibrated instrumentation, but the underlying advertising expressions will remain print, audio, and motion.

ChatGPT Ads TAM Background

Back in 2000, it seemed like Google search came out of nowhere. Today, there are an estimated 6 billion internet users, and ~4 billion of them use Google search. Business models don’t get much more ubiquitous than that.

And according to various estimates, there are ~1.3 billion ChatGPT users, give or take. As of late 2025, the official ChatGPT mobile app has achieved massive adoption, with approximately 1.36 billion cumulative global downloads since its launch in May 2023.

Google is the top candidate for the most amazing business ever created. Microsoft is probably a close second. Back in the early 80s, IBM did not fully appreciate the strategy of Bill Gates, Paul Allen, and Steve Ballmer. They simply failed to see how Microsoft would create a massive and enduring economic moat that it still enjoys today.

Moral of the Story: Just because Google is so dominant today, like IBM was in 1980, doesn't mean it will be dominant forever. Why? Because there is always a 100% chance of disruption, and it is highly probable that a substitute or new entrant will eventually come along. From what we can observe, the most likely new player in town is OpenAI.

Sizing The Opportunity

If you work in advertising and use ChatGPT, you’re probably envisioning and discussing what these future ad experiences will look like. As far as we can tell, it is reasonable to think of ChatGPT ad experiences as a substitute good for search.

Assuming such a comparison makes sense, we looked at Google search growth since 1999 and built a best-fit diffusion curve model through 2025.

If you’re interested, here’s our best-fit Gompertz model:

We were unable to locate historical data on Google Search users; however, available reports suggest the average user conducted roughly four searches per day in 2025. We use 4.0 searches/day as a reasonable 2025 proxy and built a best-fit model such that 2.0 searches/day occured around 2011 (blue curve below). This is when Larry Page first publicly discussed the “toothbrush” strategy during Google’s second-quarter earnings conference call. During the call, Page outlined his goal to create services that people use “twice a day, just like a toothbrush.”

With a decent time-series growth in Google searches and another curve for searches/day, we derived the number of Google users over time and estimate there ~4 billion users today.

Mapping ChatGPT to Google’s Past Experience

Now that we have a decent baseline for imagining ChatGPT’s ad market size, our next step is to compare the first four years of Google’s search user growth (1999 to 2002) with ChatGPT's first four years of user growth (2022 to present). From there, we can construct an addressable market for ChatGPT.

Our model estimates that Google had 88 million unique users in its fourth year of life in 2002. Keep in mind, the Internet was still quite young in 2002 with a meaningful portion of people (young, old, 2nd and 3rd world countries, etc.) still not online. Back then, only 10% of the globe was online compared to 70% today.

In 2025, ChatGPT’s fourth year of life, it had over 1 billion unique users. If we update our model parameters to account for faster growth across a much bigger Internet install base compared to when Google started (the blue line), we can imagine ChatGPT scaling new users much faster than anything that has come before it. We don’t think such an acceclerated curve should be surprising to see because new technology cycles tend to compress adoption timelines, leading to faster adoption toward a maturity limit defined by the number of global internet users.

Forecasting ChatGPT DAU, Prompts per Day, and ChatGPT Revenue

With an estimate of unique users on the table, we convert it into an estimate of daily active users (DAU) using a factor of 15% (e.g., 15% of annual uniques converts to DAU).

We then asked ChatGPT: How many chatGPT queries or prompts to does the average user do per day?

Answer: On the order of ~3–5 prompts per user per day.

We use 4 promts per user per day as the mid-point starting value in 2026 and step up to 10 per day as a terminal value when adoption reaches maturital around 10 to 15 years from now (maybe sooner).

On average, advertisers pay Google search a $2.69 CPC and expect to achieve an average 3.17% CTR. That equates to an $85 eCPM. According to OpenAI, they are aiming for $60 CPM.

If we take these plugs as face value, our model suggests ChatGPT’s revenue could be $17 billion in 2026 and grow to $80 billion in 2030 based on five prompts per user per day. At maturity, our model suggests $160 billion at a $60 CPM and $185 billion at an $85 CPM.

To our forecsat into perspective, if Google search revenue grows at a 7% CAGR over the next 5 years (conservative), it will generate a whopping $310 billion. $80 billion, ChatGPT would sixth largest media channel.

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Last Media Dollar “Live” is coming to NYC: February 25, 2026. Last Media Dollar “Live” is coming to New York City on February 25 at the Roxy Cinema in Tribeca. This is not a panel, it’s a live, on-stage decision-making experience where senior practitioners and the audience debate assumptions and reveal how media choices are actually made under real constraints, in real time. Seats are very limited and moving quickly. Secure your ticket now if you want to be in the room where the thinking happens.

Get Your Seat To The Fun Today!

Advertising Economic Forum Returns to New York City: March 18–19, 2026.
This year features an expanded two-day format focused on where media capital, ad innovation, and financial strategy are heading. Day 1 at Horizon Media features eight AI founder pitches to our investor panel with live debate across agentic commerce, creative, and emerging agency models. Day 2 at The New York Times Center convenes 400 senior leaders for closed-door sessions examining market structure, incentives, and the economic forces reshaping advertising. Early bird registration is still open for now; space is limited.

Get Your Spot Today

Liftoff IPO Valuation

As reported by Reuters last Tuesday:

Liftoff, a mobile app ​marketing provider backed by Blackstone, filed for an ‌IPO on the NASDAW under the symbol “LFTO”. Reuters reported last year that Blackstone was exploring a sale of Liftoff that could value the ‍business at $4 billion or more, including an IPO or a private sale. The company intends to use IPO proceeds for general corporate purposes and to repay $1.8 billion in outstanding debt.

Rather than anchor on the rumored $4B value, let’s instead triangulate it from fundamentals with a simplified discounted cash flow approach starting with revenue and margin structure, and then turn to investment capital, optimistic forecast assumptions, and what it all implies to arrive at a fair equity value. If you’re into TLDR, after we assess past performance and plug in optimistic forecast drivers, we get to $3.1B in equity value. That said, we have a feeling the IPO will likely trade higher than our estimate.

Revenue History: Liftoff’s S1 filing provided data for FY23 and FY24, along with data for the 9 months ending in September 2024 and 2025. The filing also provided eight quarters of revenue data from 4Q23 to 3Q25 (see table below).

We decomposed the 2024 quarterly data into a seasonal index and took a whack at estimating what Liftoff will report for Q425 and FY25 in its next filing. If the seasonal index is consistent from 2024 to 2025, we get $187 million in 4Q25 revenue and $678 million for the FY25. If our revenue estimate is accurate, then YoY revenue growth for Liftoff is quite healthy at 31%.

As pointed out by Brian Wieser at Madison & Wall:

“Liftoff is a smaller peer when compared with AppLovin. But it is pursuing the same strategy: building a single platform that controls both ad buying and app monetization and using its data to improve performance over time.”

On a trailing 12-month basis from 3Q25 reporting, AppLovin’s revenue is $6.3 billion, or over 9x larger than Liftoff. According to ChatGPT and app tracking service 42matters, Apple’s App Store and Google Play publish ~1,700 new apps every day. We assume a decent portion seek the kinds of services provided by companies like AppLovin and Liftoff.

Operating Profit: Turning to operating profit margins (EBIT), Liftoff generated a material jump from 4.9% in FY23 to 18% in FY24, and has sustained that level through 2025. As we’ve pointed out in previous posts, achieving EBIT margins greater than 20% across all NASDAQ-listed companies is a low-probability event.

Notably, AppLovin is in a different class with 40% EBIT margins in 2024 and trending higher (~55%) on a TTM basis as of 3Q25 reporting.

Flywheel Business Model

Those are insane operating margins. So, before moving on, let’s take a step back to examine the basics of AppLovin/Liftoff’s flywheel business model:

1. App developers want revenue, so they plug tools into their apps (ad mediation + in-app bidding) to maximize ad yield across many ad buyers.

2. App advertisers want efficient user acquisition (UA), so they buy ads through AppLovin/Liftoff’s demand-side products to acquire users at the lowest cost possible.

3. More spend + more inventory = more data. As more ads run across more apps, AppLovin/Liftoff gains more signal about what works, e.g., which creatives, audiences, placements, and times drive installs and downstream value.

4. Better targeting/optimization improves results. Their optimization engines use that data to improve performance, e.g., lower CPI, better ROAS for app advertisers.

5. Better performance attracts more advertisers and spending. If app advertisers hit their ROI goals more consistently, they write a blank check.

6. More advertiser demand increases publisher revenue as more demand increases competition for impressions, which generally increases eCPMs and monetization for developers.

7. Higher publisher revenue attracts more apps. More developers integrate, adding more inventory, and the loop repeats.

Net Operating Profit After Cash Taxes (NOPAT)

Looking at Liftoff’s income statement, we make a few adjustments to get a pure view of operating vs. non-operating performance. The first adjustment we make is the difference between reported taxes and cash taxes on an all-equity basis.

Although minor, we first add operating lease interest expense based on the reported operating lease liability sitting on the balance sheet. We apply a BBB interest rate (AppLovin is rated BBB according to S&P Global). We also add back impairment expense (a non-operating expense) to get the purest sense of operating profit possible (Adjusted EBIT).

As mentioned above, Liftoff has $1.8 billion of long-term debt, resulting in $98 million in FY24 interest expense and $91 million through 3Q25. Since these are flows to debt holders, we add back the tax shield effects of interest expense based on a 27% marginal cash tax rate (federal, state, and local tax rate assumption).

The last adjustment to make before arriving at our cash tax estimate is to account for the YoY change in deferred tax liabilities.

We estimate actual cash taxes of $36M in FY24 and $21M for the first 9 months of 2025. Unsurprisingly, the difference in cash taxes (what you cannot avoid paying at some point) and reported taxes on the income statement (GAAP accounting, timing differences) is material to equity holders.

After accounting for cash taxes, Liftoff generated $70M in FY24 in pure operating profit (13.4% margins) and $72M through the first 9 months of 2025.

Invested Capital

Now we turn to a few balance sheet adjustments to identify operating assets and liabilities vs non-operating items to get a decent view into invested capital. The first adjustment is in “Other Assets,” where we reclassify interest rate swaps as non-operating and keep the remaining portion as operating assets.

The second adjustment is on the liabilities side, where we reclassify “Contingent Consideration” (earn-out payments to former stockholders of Liftoff Mobile) as non-operating and keep the remaining portion as operating liabilities.

The last material adjustment we make is to remove goodwill from invested capital as a non-operating asset. FWIW, we strip goodwill out of invested capital because it’s a purchase-accounting plug that reflects what the acquirer paid, not incremental operating assets deployed.

Retained Earnings/Losses

The first thing our eyes go to on Liftoff’s balance is –$1 billion in retained losses. Interestingly, AppLovin went public in April 2021 and also had –$1 billion in retained losses in FY21. However, post-IPO AppLovin’s fortunes changed dramatically with retained earnings standing at +$1B as of 3Q25 reporting. That’s a $2 billion swing.

Invested Capital

Liftoff produced respectable Return on Invested Capital (ROIC) in FY24 at 19.2%. We estimate FY25 will be around 24% and rise over the coming years. FWIW, AppLovin’s ROIC in FY24 was a whopping 73% and might go higher in FY25.

One thing to keep an eye on with Liftoff is the downward trend in invested capital. Between FY23 and FY24, invested capital decreased by 20% and by another 3% as of September 2025. This non-investment trend seems to be sector-specific because AppLovin is also not investing back into its business. We find that a bit odd because with such a high ROIC and a large spread of its cost of capital (large economic moat), you’d think a company would be investing as much as possible to grow and protect it.

When companies don’t invest in new invested capital, investors should be wary of future growth prospects. For instance, if you look at Google and Meta, you’ll find that both companies invest 20%+ per year in new invested capital during normal years. With the AI arms race in play over the past year, investment in new invested capital for Google and Meta is 30%+.

In any case, we think rational investors should take note and expect the same bold thinking from any advertising platform that is serious about winning, building an economic moat, AND sustaining it well into the future.

Valuation

When it comes to valuation in general, we prefer taking an optimistic approach (heroic assumptions) with forecast assumptions to arrive at what the highest valuation could be and get a directional idea of where Liftoff’s IPO valuation might land.

• Let’s start with revenue growth at 31% in FY25, and gradually move downward to a 20% target value in 2030. If that happens, Liftoff’s average annual revenue growth would be 25%. We think that would be phenominal growth. That said, with the potential for so many new-AI apps coming online, our growth 25% rate could also be on the light side. Time will tell.

• AppLovin’s NOPAT margins were 33% in FY24, Liftoff’s were 13%. We estimate that AppLovin’s margins will surpass 40% when they report FY25. That’s an insanely productive business model and an incredibly rare operating margin. With that in mind, we set Liftoff’s NOPAT margins to gradually grow to 40% by 2030.

• Liftoff’s current ROIC sits somewhere around 24%. Applovin’s ROIC is much higher at 73% (note that Google and Meta’s ROIC are ~45% and ~40%, respectively). Let’s keep our forecast picture positive and assume Liftoff’s ROIC will reach 50% by 2030.

With just these few value drivers in mind, we derive a 5-year forecast of invested capital, net investment, and end up with a sense for operating free cash flow creation.

Capital Structure and Cost of Capital

AppLovin’s capital structure is 95% equity and 5% debt. What’s more interesting is how it trades at a very high 2.5 beta, resulting in a high cost of equity of around 14%.

We noted above that Blackstone reported valued Liftoffat $4B. If we assume their valuation is enterprise value with $1.8 billion in debt, then the equity portion in 2025 would be 55% and the debt at 45%.

As noted above, post-IPO funds will be used to pay down the debt, so we assume 2026 equity/debt portions will be more like 90/10 and tend toward a normalized equity portion over time. We also assume the beta for the sector will move down to 1.3 over time, which is highly favorable to our equity valuation for Liftoff.

Valuation Stack: Given everything presented so far, we end with the following valuation stack for Liftoff.

• $5.3 billion in operating value

• Plus $14 million in excess cash today and $93 million in net operating loss (NOL) carryforwards gets us to an enterprise value of $5.4 billion.

• Less current debt and debt equivalents, resulting in $3.1 billion in equity value or $29/share based on 106 million common shares.

As we mentioned above, “We have a feeling the IPO will likely trade higher than our estimate”. Here’s why…

If investors simply apply AppLovin’s EBITDA or revenue multiple to Liftoff’s FY25 revenue, you could see an IPO equity value around $5.4 billion.

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Last Media Dollar “Live” is coming to NYC: February 25, 2026. Following a wildly successful London debut, Last Media Dollar “Live” comes to New York City on February 25 at the Roxy Cinema in Tribeca. This is not a panel, it’s a live, on-stage decision-making experience where senior practitioners and the audience debate assumptions and reveal how media choices are actually made under real constraints, in real time. Seats are very limited and moving quickly. Secure your ticket now if you want to be in the room where the thinking happens.

Get Your Seat To The Fun Today!

Advertising Economic Forum Returns to New York City: March 18–19, 2026.
This year features an expanded two-day format focused on where media capital, ad innovation, and financial strategy are heading. Day 1 at Horizon Media features eight AI founder pitches to our investor panel with live debate across agentic commerce, creative, and emerging agency models. Day 2 at The New York Times Center convenes 400 senior leaders for closed-door sessions examining market structure, incentives, and the economic forces reshaping advertising. Early bird registration is still open for now; space is limited.

Get Your Spot Today

WEBINAR RECORDING: Inside C Wire’s AI Stack, How In-House Tools Are Creating Superhuman Productivity

On one hand, adtech companies can layer AI on top of legacy workflows and get incremental gains. On the other hand, “AI-native” companies redesign processes end-to-end on their own, resulting in leaner and faster organization.

Last week, Quo Vadis had a detailed discussion on how Rui de Freitas (CEO/co-founder of C Wire) has replaced fragmented SaaS tooling (external email, CRM, marketing apps, etc.) with purpose-built, lightweight, and internally-built systems that the company fully owns and tweaks regularly.

Our conversation tied performance culture and outcomes. For example, as AI absorbs “busy work,” employees and teams are increasingly evaluated on demonstrable impact rather than activity. The best example of this is how Andy Jassy is handling it at Amazon.

“To kick start 2026, Andy has asked his corporate workforce a very simple question: What did you do last year?”

In essence, Amazon is asking its workers to list three to five accomplishments that show the best things they achieved at the company in 2025, where accomplishments are defined as specific projects, goals, initiatives, or process improvements that show the impact of their work. If you’re just listing “busy work” that AI can do, that’s not enough. Newer companies like C Wire are thinking the same way and acting on it very early in their existence.

We discussed two concrete examples anchor our discussion.

1. We covered C Wire’s internal events/meeting setup tool used to operationalize conference presence (tables, availability, duties, booth workflows), including a persona-style interactive experience built quickly due to falling development costs.

2. We also dug into C Wire’s internal email/CRM layer designed around its specific customer and supplier world, which categorizes inbound messages, proposes tasks, standardizes follow-ups, and creates a foundation for richer automation with LLM-driven agents.

It’s All About Net Revenue per Employee

When we stack up our AdTech10 by age and filter out net revenue for each company on an ex-TAC basis (TTD, CRTO, DSP, MGNI, PUBM, DV, IAS, TBLA, TEAD, and RAMP), we can visualize how net revenue per employee grows over time.

By the 12th year of life, an average pre-AI adtech company should be $363K/employee. MarTech is a bit lower at $241K. In the future (e.g., now), investors expect a meaningful AI-based expansion in revenue productivity, a much higher hurdle rate on revenue productivity, and much better financial performance relative to other investment alternatives.

How Investors Think About AdTech

Although investors might not verbalize it, this is how they think about the intersection of adtech and investing criteria.

Consider a 5-year old adtech company that processes gross ad spend, collects a take rate, and earns $10 million in net revenue.

• In the old paradigm, the cost of revenue expectation would be around 12%. In the new AI paradigm, let’s say it’s 5% because AI computes more efficiently than before.

• Let’s also say investor expectations on revenue per employee have doubled from $150K to $300K, cutting required headcount in half.

• Let’s increase all-in salary/benefits by 50%. The humans working in the new paradigm are top-notch, creme de la creme talent. On average, they earn more than before.

• In the old adtech financial structure, 80% of fixed costs are people costs. In the new AI workplace, it drops to 65%. On a nominal basis, non-people fixed increases because the cost of AI is not, and will not be, free. But the benefits far outweigh the costs.

• The difference in expected operating profit (EBIT) is material, increasing from 6% (C-) to 18% to (A-).

• In addition, investor expectations on capital efficiency (net revenue ÷ invested capital) will likely increase with AI in play everywhere. That means new adtech companies will generate more net revenue for the same $1.00 in invested capital. Put another way, the new paradigm adtech company needs less invested capital to generate more revenue and faster.

Bottom Line: Whether investors verbalize it or not, or perhaps prefer to simplify the world by talking about EBITDA or adjusted EBITDA, what always matters most is return on invested capital. The grand majority of older adtech companies have had difficulty breaking above 10% ROIC. Notably, the average cost of capital in adtech land is greater than 11%, which is value-destructive if ROIC is less than that hurdle rate.

For new AI adtech companies (or re-invented companies from the last adtech vintage), investor expectations on ROIC are more in line with what they get from Google, Meta, Applovin, etc. If you can’t deliver it, then why should they invest in you instead of them?

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Advertising Economic Forum Returns to New York City: March 18–19, 2026.
This year features an expanded two-day format focused on where media capital, ad innovation, and financial strategy are heading. Day 1 at Horizon Media features eight AI founder pitches to our investor panel with live debate across agentic commerce, creative, and emerging agency models. Day 2 at The New York Times Center convenes 400 senior leaders for closed-door sessions examining market structure, incentives, and the economic forces reshaping advertising. Early bird registration is still open for now; space is limited.

Get Your Spot Today

Last Media Dollar “Live” is coming to NYC: February 25, 2026. Following a wildly successful London debut, Last Media Dollar “Live” comes to New York City on February 25 at the Roxy Cinema in Tribeca. This is not a panel, it’s a live, on-stage decision-making experience where senior practitioners and the audience debate assumptions and reveal how media choices are actually made under real constraints, in real time. Seats are very limited and moving quickly. Secure your ticket now if you want to be in the room where the thinking happens.

Get Your Seat To The Fun Today!

London PubCast Interview with Revving

In a candid pub-side conversation, Revving co-founders (Chris and David) unpack a structural problem at the heart of adtech:

Long payment terms force every intermediary in the supply chain to act like a bank. Brands pay late, platforms get paid early, and agencies, DSPs, SSPs, and publishers are left financing growth with their own balance sheets.

Drawing on their backgrounds in corporate finance and private equity, the two founders explain how Revving emerged from an insight first observed in mobile gaming and later across the broader digital economy.

Real-time transaction data exists everywhere, yet funding decisions remain anchored to slow, invoice-based models.

Revving’s solution is what they call transaction-based funding. A modern evolution of invoice factoring designed specifically for digital marketplaces. Instead of waiting for invoices and reconciliations, Revving integrates directly into platforms like DSPs and SSPs, ingesting live transaction data via APIs. This approach allows Revving to underwrite credit based on actual delivery and payer quality (e.g., Publicis, Google, The Trade Desk, Magnite, etc.), not the balance sheet of a fast-growing intermediary.

Funding can be advanced daily or weekly, giving finance teams precise control over working capital without giving up equity or posting blanket security over the business. Customers can dial funding up or down as needed, use Revving alongside existing bank facilities, and keep Revving invisible to their commercial partners.

Our discussion emphasized why this model is fundamentally different from traditional receivables factoring or bank debt:

• No setup / legal fees

• No minimum utilization

• No intrusive debtor notifications

• No asset-wide liens.

Beyond capital, the founders highlighted RevHub, Revving’s unified reporting and reconciliation layer, which consolidates receivables, transactions, and payment performance across dozens of counterparties, which dramatically reduces the finance team’s overhead.

Over time, Revving’s growing dataset also improves risk models, enabling more accurate pricing and lower costs for customers.

The conversation closed on scale and strategy: Revving has raised £107m (£7m equity, £100m funding capacity) backed by global asset managers, and is expanding aggressively in the U.S.

Revving is also embedding AI and agentic workflows to keep headcount low while scaling volume. The ambition is not just to fund adtech more efficiently, but to become the agentic credit infrastructure layer the industry has been missing.

If you’re interested in learning more, reach out to David and Chris

• david@revving.io

• chris@revving.io

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Advertising Economic Forum Returns to New York City: March 18–19, 2026. AdEF is back with an expanded two-day format focused on where media capital, ad innovation, and financial strategy are heading next. Day 1 at Horizon Media features eight AI founder pitches to an esteemed investor panel built around Bezos-style memos, live debate, and finance-first evaluation across agentic commerce, creative, and emerging agency models. Day 2 at The New York Times Center brings 400 senior leaders together for closed-door sessions to examine market structure, incentives, and the economic forces reshaping advertising. Early discounted registration is now open. Space is limited.

Get Your Spot Today

Last Media Dollar “Live” is coming to NYC: February 25, 2026. On the back of a wildly successful live experiment in London last week, we’re bringing Last Media Dollar “Live” to New York City on February 25 at the Roxy Cinema in Tribeca. This is not a panel. It’s a live, on-stage decision-making experience where senior practitioners and the audience debate, challenge assumptions, and reveal how media choices actually get made under real constraints, in real time. Seats are very limited and already moving quickly.
If you want to be in the room where the thinking happens, secure your ticket now.

Get Your Seat To The Fun Today!

Reinventing TV Buying and Agentic CTV

In this Quo Vadis live webinar, we sat down with Jules Minvielle, CEO and co-founder of Olyzon, to explore how TV buying is being reinvented through agentic CTV planning.

We began by grounding the conversation in a shared reality:

“TV remains the most powerful medium for mass reach and attention, but the mechanics of buying it have fundamentally changed.”

The fragmentation between Linear TV and CTV has introduced complexity, inconsistency, and an overwhelming number of planning decisions. What was once a relatively simple exercise in reach is now a multi-dimensional optimization problem spanning formats, platforms, contexts, and creative variations.

Jules outlined why Olyzon addresses this problem from first principles. Rather than layering more tools onto a broken workflow, Olyzon positions itself as The Agentic CTV Advertising Platform, using AI-powered agents to automate and continuously optimize media planning. The goal is not to replace human judgment, but to augment it, which in turn allows planners to operate at the speed and scale required by modern TV.

Our discussion focused on three core pillars of this agentic model:

1. Media plan automation, where AI agents dynamically adjust plans in real time as performance signals evolve.

2. Program-level contextual targeting, shifting the unit of planning from broad audience proxies to the actual CTV programs that best align with brand message and viewer intent.

3. Multi-format creative activation, enabling smarter sequencing and immersive formats that improve attention and downstream outcomes.

Lively Audience Q&A

1. Like Netflix or Hulu are not passing this data today, what incentives would motivate them to do so?

2. Will placement-level reporting be available at the individual content or program level, or will it be limited to broader content categories?

3. Do you support third-party independent measurement partners for ad serving and creative verification?

4. Recent social media platform moves suggest an increased focus on CTV ad dollars—for example, Pinterest acquiring tvScientific and Meta launching Instagram for TV. What are your thoughts on these developments?

5. If Netflix were to acquire Warner Bros. Discovery, what ripple effects might this consolidation have across the CTV ad tech ecosystem?

6. Are you able to manage reach and frequency deduplication versus platforms like YouTube?

7. How is brand safety managed in CTV environments, and does it rely on content ratings (e.g., PG, PG-13) or more granular contextual signals?

8. Are you currently participating in any multi-model attribution frameworks, and how is TV perceived within those models?

For anyone trying to understand where TV buying is heading, and why agentic systems may be the missing layer, this conversation offers a clear, practical lens.

If you would like to meet Jules, you can contact him here: jules@olyzon.tv

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

The future of TV buying is becoming more automated, more contextual, and very soon it will be far more agentic. Join Quo Vadis on December 18 at 10:30 AM ET, where I’ll be hosting Jules Minvielle (CEO & Co-Founder of Olyzon) for a live webinar on how Agentic CTV Planning is reshaping reach, attention, and automation across the fragmented TV ecosystem.

Register Today

Case and Point

Across 14K match-ups in our Last Media Media Online Game, we can observe that Linear TV has a low preference rank (12) but a very high actual spending rank (3). We see the opposite with CTV.

• LinearTV is 3.5x overfunded relative to media allocation preferences, and

• CTV is 3.6x underfunded relative to preference. What is the friction in the middle? Let’s discuss!

Jules Minvielle and I will dig into

1. How agentic aligns planning signals with real attention and outcomes

2. Using AI agents to manage CTV complexity dynamically

3. Creating a unified, explainable workflow across data, media, and optimization

Don't Miss It. Register Today

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Get Your Seat at AdEF NYC 2026

DEAL Roundtable on The French Adtech Invasion: On Tuesday, December 16, 11:00 AM to 12:00 PM ET, Landmark Ventures is hosting an invite-only Zoom discussion on the surge of French adtech companies expanding into the U.S. via NYC. Jules Minvielle (Olyzon), Arthur Larrey (Audion), Arthur Querou (Vibe.co), Mathieu Roche (ID5), and Fabien Magalon (xpln.ai) will unpack what’s fueling this expansion, how these founders are approaching the U.S. market, and where the strongest partnership and M&A opportunities are emerging.

Register for DEAL

WEBINAR RECORDING: The Future of Persona-Driven Advertising

How persona creation, contextual AI, and transparent supply signals pave the way for autonomous planning and buying agents

In our live conversation last week from London, Quo Vadis sat down with Rui de Freitas, CEO/co-founder of C Wire, to unpack the strategy and implications behind MatchPersona™, C Wire’s new AI-powered persona agent. We know it’s a meaningful step toward autonomous, transparent media-buying systems because CWire’s clients love it.

Our discussion explored how MatchPersona™ converts traditional static personas, which are typically trapped in decks and brand books, into shared, machine-readable objects that drive live media activation across trusted publishers. Agencies and brands can now collaboratively build personas, see exactly how each attribute maps to contextual and supply signals, and activate curated deals or outcome-based bidding without cookies, IDs, or opaque black boxes.

The webinar session also previews how this architecture lays the foundation for autonomous planning and buying agents that can simulate strategies, optimize dynamically, and operate with common-sense transparency and compliance.

Part 2 with Rui January on 14, 2026

That was Part 1 with Rui on external-facing AI tools. He’ll return to Quo Vadis Live on January 14th for a deeper dive into how C Wire has built in-house AI tools across multiple functions to create superhuman productivity. Wait until you see how their internal email toolkit boosts sales and account management speed and quality of communication, and we’ll also show you C Wire’s conference demo tool, originally developed for DMEXCO last September.

Every company should take note. CWire’s pace and depth of AI integration are showing adtech the way forward and how all players will need to compete.

Register for Part 2 with Rui de Fruites

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Don't Miss Out. Get Your Seat Today!

Dec 10 at 10:00 AM: Join Quo Vadis and CWire’s CEO, Rui de Freitas, for a live conversation on how collaborative persona creation, contextual AI, and transparent supply signals are paving the way for autonomous planning and buying agents. We’ll unpack how to turn static personas into shared, machine-readable objects that drive real activation today and lay the foundation for fully autonomous media-buying systems. Registration here.

Measuring Marginal Media Value

For the past few months, we have been running the Last Media Dollar (LMD) game experiment with real marketers and media practitioners. The game is simple:

• There are 22 places where most media money flows (See Updated Media Money Flows, October 2025).

• You’re down to your last media dollar and must choose between two channels in pure head-to-head matchups.

No budgets. No politics. No gaslighting. No attribution dashboards that create bias in your head. Instead, you get loss aversion in motion and gut instinct as if Tversky/Kahneman were framing the trade-offs for you to make.

Methodology: Across nearly 200 game sessions and over 14,000 1-to-1 matches, we collected real decisions and converted them into win rates. The win rates are then simulated into the 231 matches it would take for all 22 media channels to play each other one time. That would take too long to play, so we designed LMD to be played in ~10 minutes, where between 60 and 110 matches are played. That gives us reliable win rate statistics. For example, if YouTube has a 72% win rate from our sample data set and Linear TV has a 49% win rate, then YouTube beats Linear TV in the backend simulation.

Finally, we fed the simulated results into a Bradley–Terry model to extract a clean “preference strength” for each channel. In parallel, we line up the actual market shares (where the money actually goes today, see Updated Media Money Flows, October 2025) and compare them to the preference rankings (where people would want the money to go if freed from the system’s frictions).

This creates a rare kind of map between the psychological market versus the media market. It’s the difference between a frictionless market and one structured on all kinds of incentives that can diminish advertising outcomes in favor of financial outcomes.

Market Frictions Between Preferences and Actuals

The Last Media Dollar is a designed experiment to reveal true preference, but the media economy isn’t built to follow preference. Between structural constraints, organizational incentives, and legacy systems, an entire stack of market frictions sits between what marketers want to do and what the industry actually does. In classic economic terms, LMD attempts to reveal an efficient market signal of true marginal value, while today’s media budgets reflect an inefficient market distorted by frictions, incentives, and legacy structures.

This isn’t unusual. Most real-world markets drift toward inefficiency whenever the following ten habits accumulate faster than the ability to reallocate capital decisions like media spending:

1. Budget silos: Even if a marketer prefers a channel, the silo walls prevent money from moving across categories.

2. Procurement & contracting inertia: Long-term upfront commitments (especially in Linear TV) lock in spend regardless of marginal preference.

3. Measurement systems anchored in the past: Channels with weaker or newer measurement frameworks can’t “win” on reports even if they win in preference.

4. Organizational politics & incentives: Teams optimize for internal scorecards, not economic truth, while agencies favor channels tied to rebates and favorable margin structures. Everyone deserves to make a buck in the media trade.

5. Ease-of-execution (friction costs): Channels like Search, Meta, and YouTube are very easy to buy and optimize at scale.

6. Legacy planning frameworks: Annual planning cycles allocate to TV first because “that’s how planning is done,” while newer channels only get incremental dollars.

7. Risk aversion & CYA behavior: No CMO gets fired for putting money into Google, Meta, or Amazon, even though all three of the big players get CMOs stuck in a game called escalation of commitment.

8. Reporting ecosystems built around the big platforms: What gets measured gets spent. Dashboards are built around Google Ads, Meta Ads Manager, and TV GRPs.

9. HoldCo trade economy: Agencies are financially incentivized to maintain spend in specific channels, regardless of marketer preference. Preferred channels at the marketer level often lose inside agency trading systems.

10. Complexity penalties: Simplicity wins budgets for now, but dealing with complexity with agentic AI will very likely win the future and perhaps align preferences to more efficient media spending allocations.

What We Found: Preference ≠ Spend

Let’s start with the broad strokes. The preference-vs-actual ranking reveals a far more nuanced Big Tech story than the broad “platforms dominate regardless of value” narrative.

Meta and YouTube are remarkably close to equilibrium, with their actual market shares closely matching what game players choose in forced trade-off scenarios, making them the most efficiently funded channels in the ecosystem.

Amazon Retail Media shows only moderate underfunding, suggesting structural constraints rather than a lack of belief. It seems clear that marketers want to allocate more to Amazon. That’s why we have Amazon's ad revenue growing at a 15% CAGR over the next five years.

The lone outlier is Search, which is significantly overfunded relative to preference. However, advertising in GenAI environments is in the top half of preference ranking as the “new search advertising space.”

Retail, Social, and CTV channels show latent upward pressure. Amazon Retail Media, Walmart Connect, Influencer Marketing, and CTV Direct Premium all scored significantly higher in preference than in actual budget ranking. This tells us something important:

These channels are “where marketers want to go,” but the machine hasn’t let them get there yet.

The TV story is a tectonic imbalance

TV deserves its own section because the magnitude is stunning. Across all matchups, preferences, and Bradley–Terry scoring, we can see that Linear TV has a low preference rank (12) but very high actual spending rank (3). In other words, Linear TV is 3.5x overfunded relative to preference, but CTV is 3.6x underfunded relative to preference.

It’s a mirror image illustrating the tectonic imbalance between what marketers believe in a pure decision moment and what their systems allow them to do today. Linear TV persists because the current machinery of media planning was built around it. However, CTV is gradually accelerating as eyeballs shift because the marginal dollar righty believes in a TV future with context and audience targeting at its core.

Brian O’Kelley Thought Experiment: Imagine a world where buyer and seller agents simply match themselves based on mutual preference, letting the market discover its own equilibrium. In that world, the sytem’s belief system gets aligned to outcomes. (See A Matching Algorithm for an AdCP Stable Market). That’s the future of TV. It’s about a structural overhang yielding to the behavioral momentum of where the market wants to move before the dollars actually move.

What It Means, Especially on December 9

On December 9, there are two great events in London. One is “The Future of TV,” an all-day conference with nine themes and ~20 speakers offering industry reflection.

The other great event is an experiment called Last Media Dollar Live at Everyman Theatre in Borough Market. LMD Live is a boutique, behavioral-economics-driven experience. We aren’t debating the future of TV because our preference data and layman observations already show us the way. You’ll get to see how Cadi Jones (SVP EMEA, Index Exchange), Dan Plant (Chief Strategy Officer, Starcom), and Richard Kramer (Managing Director/Founder at Arete Research) make and rationalize their media choices live on stage.

As far as the future of TV goes, it’s already answered. CTV wins by preference while Linear TV wins by structural inertia. The only question is:

For how long will Linear TV inertia hold?

🎟️  Get one of our remaining tickets today and experience it yourself. 
🎟️ Use discount code LMD25 for a £75 ticket. 
🎟️ That's a lot of bang for the buck!

Don't Miss Out. Register Today!

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Day 2 is our “Davos of Advertising” experience, a closed-door, off-the-record forum where leaders reassess the probabilities of who’s winning and losing in the rapidly changing advertising market. If you’d like to attend, register your interest today!

Apply to Attend Today!

HOUSEKEEPING 2: Last Media Dollar “Live” in London is on December 9 at Everyman Theatre in Borough Market. If you’re based in London/UK or happen to be in town on December 9, Quo Vadis is hosting our first LMD Live event. The courageous Cadi Jones will be playing the game live on stage as Tom Triscari plays color commentator. It’s an unscripted dialogue of media decision trade-offs in a room filled with inquisitive minds. The theatre only holds 100 people and is filling up fast. Quo Vadis subscribers get 25% off tickets with code LMD25.

Get Your Seat Today

3Q25 AdTech, MarTech, Agency Update

With 3Q25 earnings now in the rear-view, we can assess how public adtech names are holding up in a market increasingly shaped by channel shifts and sharper investor scrutiny. As a reminder, last quarter we rebuilt our equal-weighted portfolio beginning to start January 1, 2022, as a clean starting point since every company in our coverage universe was already public by then.

Here’s how it works using our AdTech12 portfolio as an example. You start with $100, divide it equally across 12 companies (see list below), buy shares in each based on the stock price on January 1, 2022, and hope to find alpha by beating the NASDAQ or S&P500.

AdTech12

• Our portfolio was up ↑6% in 2Q25 and is now up ↑13% as of 3Q25 reporting.

• Peak returns happened in February 2025, performing in parity with the NASDAQ, but have since fallen off and dipped downward heading into year-end.

• The AdTech12 includes: TTD, CRTO, APP, DSP, ZETA, MGNI, PUBM, DV, IAS, TBLA, TEAD, and RAMP. If we exclude APP, the AdTech11 would be down ↓34%.

• When TTD was hot, it pretty much held up our entire portfolio, but it’s been rough sledding with dramatic shifts across the open web inventory scene. And with news like Omnicom Allegedly Pivoted A Chunk Of Its Q3 Spend From The Trade Desk To Amazon, the future is not going to be any easier. The Trade Desk is trading below a $20 billion market cap and heading toward our modeled fair market value closer to $10 billion or $20/share. That said, the company still has a massive opportunity, but it’s not going to be easy without a plan to get there. Management’s plan is unclear to us at the moment.

• Besides AppLovin, Zeta is the only other positive performer across our portfolio.

• The same $100 invested in a NASDAQ ETF would have returned 47% with less risk.

MarTech8

• Our MarTech8 portfolio is dismal. SHOP, HUBS, SPT, SEMR, TWLO, BRZE, YEXT, and AMPL.

• It was bad in previous quarters and getting worse, down ↓47%. Software used to eat the world. Now it seems that AI is eating SaaS.

• Shopify is still the only bright spot, up ↑14%.

Agency6

• Our agency portfolio consists of: WPP, Publicis, Omnicom, IPG (soon to be OMC-IPG, expected to close by year-end), Stagwell, and S4 Capital.

• The agency world has been rough going, except for Publicis, up ↑48%.

• WPP is still struggling, down ↓75%. WPP shares leaped amid takeover bid speculation from Havas, but Havas has denied the rumors. In other rumors, we’re hearing about internal infighting amongst the ranks at WPP. Similar to TTD, WPP still has a massive opportunity (for now), it’s going to be difficult anyway we cut it, and management’s 4-point plan is totally underwhelming. A better plan would be to simply say that:

“WPP is going to buy and create more of the best land (data) available. We are going to constantly enrich our land and rent it to clients (farmers) over and over again.”

Big Tech

• We track two “Big Tech” players (Google and Meta ) in the ad space, with a third one on the rise (Amazon, see below).

• Had you invested $100 in GOOG and META on a 50/50 basis in January 2022, you’d be up 90%.

• Both Google and Meta had impressive quarters (again). We expect the good news to continue for both companies in the emerging outcome-driven ad world (more on that below).

Leaderboard Highlights

• Teads (f/k/a Outbrain) looks to be in serious trouble after Outbrain acquired Teads and rebranded as Teads. Its market cap is down to just $75 million with $650 million in debt from the private equity-like deal.

• We like DoubleVerify at $10/share. It does not take much in a discounted cash flow model to get to $20/share as DV’s M&A strategy is well-positioned for the outcome-driven world and agentic future.

• IAS was taken private by Novacap. We covered our valuation thinking and deal rationale here.

Outcome-Driven Ad World and Advertiser Types

Advertisers love to talk about transparency. They talk about it on stage in Cannes, at Possible, in trade press interviews, and on procurement checklists. But if you ignore the speeches and follow the money, a different truth emerges.

Transparency is a “nice to have,” but not a “must have” for the vast majority of marketers. The real market is in Boxes 3 and 4 below, where advertisers are free to express a preference for transparency but don’t behave as if they value it. It’s preference falsification in its purest form, and that’s okay. On hand, they say one thing because it sounds principled, and on the other, they spend on outcomes (real or perceived) because that’s the best place to look good. Welcome to the advertiser world 😂.

This is why black-box systems, such as Meta Advantage+ and Google PMax, continue to compound, and why principal-based agency trading continues to grow. The revealed preference of the market isn’t transparency, fairness, or control. It’s about performance, real or manufactured, and the political safety that comes with it.

Founder Tip: If you’re a founder of an adtech/mediatech startup, it is better to recognize this fact of life sooner rather than later, when you have to pivot into it to achieve going-concern status and keep your investors happy.

Interestingly, the only exception is Box 1. Linear TV is the third largest area of global ad spending (see Quo Vadis #130: Updated Media Money Flows) and incredibly resilient. However, with every passing day, it is shrinking as the growth engine of CTV. On traditional TV, marketers know where their ads run. On CTV, particularly across the growing long tail of thousands of streaming channels, they increasingly don’t know where all those CTV dollars are going. They are going to Boxes 3 and 4.

It is equally interesting to overlay the advertiser's long tail, ranging from large advertisers to SMBs and tiny advertisers. The deeper you go into the tail, the less transparency is demanded and the more outcome-centric the buyer mindset becomes.

Here’s a test to prove our thinking is realistic. Would you rather:

Spend $100K on a campaign that offers the amazingly angelic, crystal clear 100% transparency, and get $150K in outcomes?

OR

Spend $100K on a campaign and get $200K in outcomes?

If you’re a marketer speaking on stage at an industry conference, you’d obviously go with the former. But when you get off stage and sign the insertion order, you’d go with the latter, and you know it.

Our little test helps explain why Google, Meta, and Applovin are crushing it quarter after quarter. When you have the best mousetrap, the world will beat a path to your door.

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

Get Your Seat Today!

Welcome back to Quo Vadis Green Shoot AdTech Interviews. In this series, we’re stepping into the shoes of “AdTech Equalizers” — the greenshoot companies changing how the advertising job gets done. We’ll delve into their worldview and unveil new strategies, tools, and perspectives that make advertisers more effective.

Today, we have the great pleasure of speaking with David Mandeno, Co-founder and COO of Revving, the category leader of transactional-based funding.

If you want to learn more about transactional-based funding, hit up David here: david@revving.io.

Q1: Why is adtech’s cash-flow problem the right problem to solve now?

A: David

It’s very simple. 120-day payment terms and chronic late payments are starving growth at precisely the moment the industry needs to invest. That’s not a “finance quirk”; it’s a competitive handicap. We see great teams forced to choose between payroll, vendor bills, and new inventory because revenue recognized today won’t land for months. That’s why our view is simple: if you’ve earned it, you should be able to use it — this week, not next quarter.

Q2: What’s fundamentally different about Revving versus old-school factoring?

A: David

Traditional factoring was built for paper invoices and manual checks. Adtech is data-rich and real-time. Revving integrates directly with all the primary digital platforms and marketplaces, automates verification of earned revenue, and advances funds by acquiring the underlying transaction data, all within the funding framework of traditional invoice factoring.

We call this new model Transactional-Based Funding (TBF) because we’re using technology to evolve from an invoice-based funding model to a transaction-based one, purpose-built for the digital economy. TBF reduces friction for publishers, networks, agencies, and advertisers, enabling decisions based on live performance signals, not tangled PDF archaeology.

Q3: You raised a significant war chest. How does that translate into value for customers?

A: David

The £107m investment we secured (led by DWS) lets us fund growth at scale, not just a handful of pilots. Practically, it means we can turn long payment cycles into same-week liquidity for more partners, more often, without compromising on risk discipline.

Our analysis suggests this funding capacity could free up as much as £1.8bn for digital businesses over three years, with a potential wider economic multiplier effect. That’s how we inject oxygen back into the system.

Q4: There are plenty of funders in the market. What’s the real edge?

A: David

As pioneers of Transactional-Based Funding, we’re offering a completely new and innovative liquidity model for the digital economy. By leveraging technology with data, we’re dragging old-school invoice factoring, with all its inherent benefits around flexible payor-focused credit risk, into the 21st century. Think high funding capacity combined with flexible, automated drawdowns without the commitments, restrictions, and hidden costs typical of standard lending products.

Q5: What should a prospective partner expect in week one?

A: David

Clarity, then acceleration. We start with data connections and eligibility so you can see exactly what revenue can be unlocked, and on what cadence. From there, it’s rinse-and-repeat: advance against earnings, stabilize working capital, then redeploy into growth, inventory, talent, or tech without waiting for the slowest payer in your chain.

My background is in engineering and operations so I’m allergic to “financial theatre.” If it doesn’t measurably reduce cash-conversion time and admin drag, then it doesn’t ship.

Bonus Q: I’ve seen the recent partnership announcement between Luma and SLR Digital Finance. Is SLR a competitor in this space?

A: David

We know SLR Digital quite well. They’re very much a standard asset-backed lender that has some exposure to digital media. So while they’re technically competitor-adjacent, in that we both provide capital to similar sectors, their approach is far more traditional. SLR (and other lenders) focus on the customer’s credit risk to provide senior secured loans over a fixed term with all the associated security, restrictions, and setup/utilization fees.

In contrast, Revving’s Transactional-Based Funding model leverages technology to capture and fund on the underlying transaction data (e.g., impression sales). Revving’s credit risk assessment is also based on the payment risk of our customers’ payors. This means we can often provide more funding, more flexibly, with fewer security requirements and restrictions, all while charging a single daily fee, which drives down overall funding costs.

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.

A few weeks ago, I found myself having tea 39 floors above Manhattan’s skyline with my old friend and ex-Yahoo, Cadi Jones, SVP of EMEA at Index Exchange. We were catching up about life, adtech, and the topic du jour, AI.

As our conversation shifted from topic to topic, I told her about a game created for Quo Vadis called Last Media Dollar (LMD). As our readers know, LMD is a behavioral economics game that helps media professionals uncover their hidden budget biases. It’s not a game about winning; LMD is about creating the conditions for interesting trade-offs to surface meaningful debate and discussion. It’s about how decisions reveal relative preferences and values, and also provide leading indicator clues about which way the media wind is blowing.

The week before meeting up with Cadi, I tested LMD during an Advertising Week NYC presentation with Ana Milicevic at Horizon Media. The 200 attendees played it beforehand, and then we showed them how their media choices played out against actual market allocations. Trade-offs create cognitive dissonance. What followed is exactly the kind of discussion that leads to enlightenment.

Last Media Dollar is very simple to play:

• Players face matchups between media channels such as CTV vs Social, Search vs Retail Media, Out-of-Home vs Display, and so on across 22 different media channels.

• Each match reveals preferences, and new matches are automatically weighted with a running count of past wins.

• When you’re done, you get various charts and graphs to see your relative media allocations.

Cadi was leaning into the concept, so I suggested:

Shall we play a game? You make the game-play decisions and I’ll play a sherpa role offering up extra information just for fun.

We had a blast! Playing Last Media Dollar live in person with back-and-forth banter turned out to be a conduit to a conversation rarely had in the media business.

That’s when it hit me:

Why not play Last Media Dollar live on stage?

Last Media Dollar “Live” — London Premiere

Join Cadi Jones and me for the first-ever Last Media Dollar “Live” in London at Everyman Theatre in Borough Market on December 9, 2025.

This is the first live experiment at the intersection of media allocation and behavioral economics. The experience will be a bit of data science, a bit of psychology, and a lot of fun, all wrapped into a pre-holiday morning with London’s media community.

📍 Everyman Theatre, Borough Market
🗓️ Tuesday, December 9
🎟️ Tickets: £100. Only 100 seats available, so act fast to get your seat at the best little event of the year!

Get Your Ticket Today!

Agenda

9:00 AM | Check-in, coffee, breakfast treats
10:00 AM | Game 1 Live: Cadi Jones vs The Algorithm
11:00 AM | Game 2 Live: One brave audience contestant selected randomly
12:00 PM to 1:00 PM | Networking, snacks, and drinks in the bar

Live Game Play

Lifelines: Each contestant will get a few lifelines where you, the audience, get to help them through the toughest matchups.

Live Media Allocation Comparisons: A few days before the event, attendees will receive a private game link to play a custom-coded game. We’ll compare the aggregated audience results to Cadi’s and to our random contestant’s outcomes.

LMD “Live” is an event experience that London’s media and adtech community won’t want to miss!

Get Your Seat to LMD "Live" Today!

See you in London on December 9!

Special thanks to Revving

The official sponsor of Last Media Dollar Live.

Revving is redefining how capital flows through the media supply chain with transaction-based funding that gives companies the ability to unlock growth without waiting on payment cycles. Their support helps make experiments like Last Media Dollar possible, connecting creativity, data, and decision-making in the most literal sense. Revving keeps the media economic engine running.

Day 1 features our high-energy Founder + VC pitch competition hosted at Horizon Media. If you’re working on Agentic Agency, Agentic Creative, Buyer/Seller Agents, or Cognitive Commerce, and you’re bold enough to pitch your worldview thesis to a live audience on March 18, then get on our application list.

Apply to Pitch Today!

Day 2 is our “Davos of Advertising” experience, a closed-door, off-the-record forum where leaders reassess the probabilities of who’s winning and losing in the rapidly changing advertising market. If you’d like to attend, register your interest today!

Apply to Attend AdEF Today!

From Matching to Clearing: Designing the Protocol Stack for Agentic AdTech

Last week in our post, A Matching Algorithm for an AdCP Stable Market, we laid out how the classic theory of Gale-Shapley stable matching offers a blueprint for a coordination layer in AdTech Context Protocol (AdCP). Similar to how dating apps match potential mates, we explored how seller-side agents (publishers) can propose, buy-side agents (advertisers) can accept or defer, and how such a proposal model can generate stable matching under conditions of well-specified preferences and capacity constraints.

When we examine the buyer-seller universe, we estimate there are ~13 trillion unique buyer-seller pairs possible. That’s precisely why some kind of matching system is necessary.

Assuming our estimates are close enough, if every advertiser operates through a buyer-side agent and every meaningful media seller (excluding the 4 million long-tail sites that monetize passively through pipes) operates through both buy-side and sell-side agents, the addressable matching and trading surface is massive. Imagine a human being’s brain trying to find optimality and stability with so many options. Not gonna happen.

If we include the millions of long-tail sites and imagine how they get tranched into curated deal IDs like ETFs that get sliced and diced into tons of potential baskets, then the number of unique pairs gets way bigger.

When each buyer agent can, in principle, evaluate and match against every seller agent, the marketplace does not resemble today’s auction environment at all. Instead, it becomes a dense, multi-bidirectional matching graph with trillions of potential buyer–seller pairs.

If we restrict the model to the true apex of the market, roughly ~1,500 to 3,000 global mega-advertisers who spend tens to hundreds of millions per year, and the ~50–150 mega-scale media sellers who control the majority of consumer attention, the agentic marketplace becomes much more compact.

At that level, the total trading surface consists of approximately ~200,000 unique buyer–seller match pairs. This is the layer where strategy, negotiation, and long-term relationships actually matter (think Cannes Lions). The challenge here is not so much about supply discovery. It’s more about clearing and coordinating persistent counterparties efficiently and in a manner where middle players can make their take.

The Gale-Shapley matching framework proposed in our last post seems ideal when the “deal” looks like the old world of direct guaranteed media buying for a specific impression, at a specific CPM, within a specific timeframe. In that sense, agentic systems are simply automating what human media salespeople and media buyers do today. That alone is a compelling innovation, particularly for high-value CTV inventory/audiences, but it’s not the full picture of what advertisers actually value.

Two layers, one market fabric

Whether we look at trillions or a few hundred thousand unique matches and how these relationships actually operate, we see that the “match” itself is only the starting point. What advertisers observedly prefer is real-time audience targeting. That preference in the brave new agentic world introduces a second layer of complexity:

Once buyer and seller agents are matched, how does the trade clear in real time? And at what price, under what rules, and with what learning feedback?

That’s where market-design (stable matching) ends, and mechanism design begins. Mechanism design is the study of how protocols, auctions, and incentives govern value exchange. In other words, matching establishes who can trade. Mechanism design defines how they transact. Together, they form an architecture of where we think AdCP could go to make programmatic advertising what it always wanted to be (e.g., not a lemon market).

Let’s consider a system as two broad layers:

Layer 1 is Matching (Market Design). This layer is about who trades with whom. It defines a network of counter-parties. Think of it in terms of Metcalfe’s Law, which states that the value of a network grows with the square of its number of connected participants (aka, “nodes,” where value = n²). In other words, each new participant adds value not linearly but exponentially, because they can potentially connect with all others.

Let’s imagine AdCP as the “network operating system” for agentic advertising markets. On one hand, Metcalfe’s Law describes the network externality where the more buyer and seller agents join the protocol, the more valuable it becomes to every participant. On the other hand, market design ensures stability and incentive alignment within that growing network such that new connections don’t just create noise or spam (the current “lemon market”) but instead produce stable matches and efficient clearing.

Without sound market design, the growth predicted by Metcalfe’s Law very likely ends up collapsing under its own weight because more participants simply means more friction, more instability, and more arbitrage (sound familiar?). With a viable market design, the exponential potential of network effects can generate enduring value (e.g., a market of peaches instead of lemons).

Layer 2 is Clearing (Mechanism Design): Given a matched pair (or set of eligible pairs), how do they negotiate or bid to determine price, volume, and execution terms in milliseconds?

If the smart people out there in AdTech Land only design matching and ignore clearing, then we end up with static pairs that don’t transact optimally. If we only run real-time bidding and ignore the matching universe, then we end up with the current auction chaos (which only a small handful of people truly understand), a non-persistent structure, high friction, and wasted media dollars.

Why both layers matter

Before matching and clearing can work together, the market needs to 1) remember who it’s dealing with, 2) compete to discover real value, and 3) execute all of it in milliseconds.

Trust and efficiency require persistence: Matching embeds a memory of past outcomes, preference rankings, and confidence scores. Efficient matching between potential mates turns one-time auctions into relationship-based trading. That’s critically important when you put it into the perspective of trust levels introduced by psychologists Roy Lewicki and Benton Bunker. In their work at Stanford and Berkeley in the 90s, they refer to three levels of trust:

1. Deterrence-based: This is the most basic level, based on the idea that people (agents) will behave trustworthily because the costs of breaking trust outweigh the benefits as a rational calculation of risk and reward. It seems fairly clear that programmatic has had issues with making it to level one.

2. Knowledge-based: This level develops as you get to know a person (agent to agent) over time and see their track record of reliability and competence. This is where trust is based on predictable future behavior and consistent past performance (e.g., outcomes).

3. Identification-based (aka, relationship-based): With each level building on the previous one, this is the deepest level of trust. It forms when individuals (agents) have a strong sense of shared preferences, values, and goals. It involves a deep understanding and identification with another person’s (agent’s) motivations, intent, and aspirations.

Price discovery requires competition

Matching tells us who is eligible to trade, and competition reveals what the trade is worth. Once two agents are paired, they still need a mechanism for price discovery or a process that surfaces the true marginal value of an impression at that instant.

Without real-time competition for an ad opportunity, the price becomes an assumption instead of a signal of information processing. For AdCP, this means that even inside a stable match, bids and counters must still flow. On one side of the market, seller agents offer dynamic floors (e.g., what pubX.ai and Swivel do) based on probability models of quality and context. On the other side, buyer agents (e.g., what AtomicAds.ai and C Wire do) respond with probabilistic bids derived from performance priors.

The exchange between them becomes a continuous micro-auction. It’s a negotiation dance compressed into milliseconds, where both sides reveal information for the truest clearing price to emerge. Every cleared trade refines the next round of preference weights and confidence scores. And that’s precisely how a stable market can remain adaptive without collapsing into stasis.

Latency and scale demand protocol engineering

For this system to work at the speed of digital advertising like RBT does today, the clearing function between ad call and an impression render must operate in milliseconds.

With this constraint in mind, it seems obvious that AdCP must decouple matching and clearing into parallel processes. Matching runs continuously in the background, updating as outcome data flows in. Then, when a bid request triggers, the pre-computed match state allows the clearing engine to act instantly, drawing only from eligible counter-parties rather than the entire market.

Such a parallel architecture lets AdCP transform programmatic advertising from a series of one-off auctions into a real-time, stateful market fabric where matching provides memory and clearing provides speed.

A Two-Stage Protocolic Flow

1. Proposal Stage (Matching):

   • Sell-side agent (S) proposes (think hyper-scaled RFPs) to buy-side agent (B) with inventory/context.

   • Buyer (B) tentatively accepts if it meets preference thresholds or capacity.

   • The result is a temporary match where μ (s) = b, s ∈ u (b), e.g., if a seller is matched to a buyer, then that buyer’s list of matches must also include that same seller.

2. Clearing Stage (Auction/Negotiation):

   • Within that matched pair, the buyer and seller execute a real-time bidding/clearing mechanism, which could be a declared first-price, second-price, negotiated, or hybrid auction. As long as it’s declared, it doesn’t matter what type of auction the seller runs.

   • The matched pair can also execute a direct order (e.g., an automated direct deal, but with more expansive and efficient matches)

   • The outcome determines the actual assignment and price paid, with an updated feedback loop of learning signals.

Bridging theory to practice

• In matching theory with seller-side agents proposing, the equilibrium will be seller-optimal, resulting in better inventory, higher yield, and fewer mismatches.

• In mechanism design, the clearing mechanism must respect preferences revealed in the matching layer. For example, a buyer who ranks a seller highly should be able to realize certain advantages in exchange.

• Preference lists can then become utility functions in mechanism design and influence both ranking in matching and bidding strategy in clearing. For example:

  • Buyer agent utility = Pₖ ×Pᵢ ×Pₒ (buyer’s base price value × probability inventory is suitable × probability outcome happens).

  • Seller utility = Pᵣ × Pₛ × Pₒ (e.g., seller wants to charge the rate card price, then adjusts that price based on how suitable this impression is and how likely it is to produce the desired outcome for the buyer)

What’s Next In This Series? Real-Time Settlement

Once we know who should trade (matching) and at what price (clearing), the next question is:

When and how does the money move?

Matching gives us counterparties. Clearing gives us a price. And settlement gives us closure. Article 3 in our AdCP series will explore how real-time settlement could change the economics of working capital and reshape how money actually flows through digital advertising.

Apply to Attend AdEF NYC March 2029

Disclaimer: This post, and any other post from Quo Vadis, should not be considered investment advice. This content is for informational purposes only. You should not construe this information, or any other material from Quo Vadis, as investment, financial, or any other form of advice.