Everyone agrees AI is changing advertising. The numbers are a lot more complicated.
Key Takeaways
- The data is in: AI revenue growth rates are impressive; the actual revenue, not so much
- Cannes decreased its AI panel discussions by 80% compared to last year
- IPO history: three companies, three trillion-dollar valuations — what's ahead for AI advertising revenue
In this episode of Media Monitor, Kelly Sweeney and Sean Wright examine a series of major developments shaping the future of artificial intelligence and advertising. Three of the most valuable AI companies in the world are heading toward public markets simultaneously. OpenAI is running ads for the first time. The Pope issued a formal encyclical on artificial intelligence. And Cannes Lions — the advertising industry's most-watched annual gathering — just dedicated a small fraction of its agenda to AI.
Cannes Dialed AI Back — and That's Meaningful Data
Last year, roughly a third of all Cannes content was AI-related across 150 hours of panels. This year, the total is 10 hours. That's a big change.
Two of the seven panels discussing AI this year aren't even necessarily pro-AI. They're framed around how human creativity survives in an AI age — how agencies and creatives differentiate themselves as automation matures. The conversation has shifted from AI as a requirement to AI as a supplement.
Part of that mindset is the audience. Cannes is full of people who built careers on the premise that creative output is a human skill that makes it unique and interesting — and AI is a very real threat to that premise.
Part of it is also the technology itself. Last year's panels were anchored on video generation, focusing on AI-produced creative with no production crew required. This year, that narrative has pulled back. ChatGPT stepped away from its video generation component. The whirlwind innovation that fueled last year's hype cycle has slowed to something more incremental. The gap between the press releases and reality is narrowing, and Cannes is reflecting that.
Three Companies, Three Trillion-Dollar Valuations
Anthropic has filed its S-1 targeting an October IPO at a projected $1 trillion valuation. xAI — Elon Musk's AI company and home of the Grok model — is targeting a listing at a valuation north of $1.75 trillion. OpenAI is expected to follow in roughly the same window.
That's three companies, each valued at over a trillion dollars, all converging on public markets while their revenue models are still being worked out in real time. All three face the same fundamental question: how do you close the gap between extraordinary infrastructure costs and revenue that is still being figured out?
What public markets will demand from these companies is the same thing they demand from every high-growth tech listing: a credible path to profitability. And that's where things get complicated. The infrastructure costs behind these platforms are extraordinary, including the R&D, construction, and powering of data centers. Investors will be looking hard at the gap between what these companies spend versus what they earn.
That pressure is already reshaping decisions. Anthropic recently raised token costs significantly. One major company reportedly burned through its entire annual AI budget in just 3.5 months, driven by higher per-token pricing and faster-than-expected internal usage growth. This is the frog-in-a-pot approach: incremental price increases that downstream clients absorb quietly, designed to close the profitability gap before the IPO filing date to look financially sound.
OpenAI's Advertising Revenue Is Growing Fast. The Actual Dollars Are Not.
When OpenAI launched its advertising product, the growth rates were staggering. The first month-over-month was 17,000%+, then 954%+, then 644%+. For context, when other platforms launched comparable ad products, their average first-month growth rate was around 340%. OpenAI's launch was roughly five times that pace — but what matters more is the actual dollar volume.
Five months in, revenue remains in the low-to-mid single-digit millions. OpenAI's stated target was to build a billion-dollar ad business by year-end. The current trajectory does not support that. While the growth rates are impressive, the actual revenue is not so much.
This is why advertising revenue matters so much to the broader IPO story. Token fees and enterprise subscriptions are the primary revenue levers right now. Advertising was supposed to add a third leg to the stool, but so far it's functioning more as a proof of concept.
Anthropic Went to the Vatican. OpenAI Did Not.
That's not a trivial detail. It's a branding decision playing out in real time.
When the Pope issued his recent encyclical on AI — calling for the technology to be "disarmed" and urging the world to pause before blindly accepting its acceleration — Anthropic was in the room. OpenAI was not. Anthropic ran a Super Bowl ad earlier this year explicitly distancing itself from advertising as a revenue model. The biggest eye-catcher: Anthropic turned down a government contract it deemed a "values conflict." OpenAI signed it the following day. For two companies competing for the same enterprise budgets and the same public market attention, that asymmetry matters.
Whether you find this positioning credible or strategic, it is working as a marketing approach. The "responsible AI" frame gives enterprise buyers — particularly in regulated industries, large agencies, and brand-sensitive categories — a rationale for vendor preference that goes beyond just feature comparison.
The reality is this: the Pope's message and Anthropic's presence at that conversation are pointing at the same thing the Cannes agenda is pointing at — a meaningful segment of the market wants to slow down versus expediting even more AI integration.
96% of Executives Say AI Is Transformative. 77% of Workers Say It Made Their Job Harder.
A recent survey asked C-suite executives whether AI would fundamentally improve their companies — 96% said yes. The same survey asked workers, and the results were starkly different.
77% of workers said AI increased their workload — not because the tools failed, but because they created a new category of work. Fact-checking AI output. Editing for tone. Cleaning up artifacts that take longer to correct than they would have taken to produce manually in the first place. The term circulating now is "AI slop" — generated output that isn't quite right and requires human intervention before it can be used, causing additional overhead work for employees.
That gap is appearing consistently across multiple major surveys now, almost identically each time. It's the same tension showing up at Cannes and in the Pope's encyclical, expressed in workforce data. A large and growing portion of the people closest to AI adoption are saying: the details matter.
Where AI Advertising Actually Goes From Here
AI advertising is real, growing, and nowhere near its final form. The channel that exists today — probabilistic, search-adjacent, limited in ad unit variety — is not the channel that will exist in five years. That's not optimism, it's just how new advertising surfaces develop. Search ads looked nothing like what they became. Social ads started as banner units on profile pages.
What is clear right now:
- Advertising alone will not get these companies to profitability on IPO timelines
- Token repricing is shifting infrastructure costs downstream to enterprise clients and agencies
- Ad dollars are growing fast in percentage terms and slowly in absolute ones
- The industry's most-watched creative event just voted with its agenda — AI is a factor, not the story
The brands and agencies that navigate this well will be the ones that treat AI advertising as a channel still finding its format — investing to understand it now, rather than waiting for it to arrive fully formed.
Written by Cassie Bryson-Evans.
Media Monitor is a weekly podcast breaking down what's happening in media and advertising — and what it actually means. New episodes every Wednesday. Subscribe wherever you get your podcasts.
