10-figure bills for data centers ultimately leave advertisers on the hook

Look at the sums involved in keeping the AI revolution rolling and you begin to wince.

More than $350 billion will be spent by Google, Amazon, Microsoft and Meta on data centers alone this year, according to an analysis by the Financial Times. Next year, that figure will top $400 billion. By 2029, it’s projected to cross $500 billion. Those are yearly, not cumulative, figures.

That investment in infrastructure is meant to support the anticipated surge in demand for generative AI tools developed by the companies in question — alongside OpenAI, which is now looking beyond its long-term deal with Microsoft for so-called “compute,” the computational power AI is built on.

The scale of that spending has some industry watchers worried. Johnny Ryan, senior fellow at the Open Markets Institute who studies the ad market, calls the present moment “very precarious.” Despite AI’s ubiquity, most companies are still experimenting and seeing only modest profits. Bain Capital estimates that AI will need to generate $2 trillion in revenue by 2030 to justify the $500 billion Big Tech is expected to spend on infrastructure by then. The core question, according to Ryan, is what will be left when the hype clears — and who will foot the bill? “The pieces are going to start falling out of the sky at some point,” Ryan says.

And there’s growing speculation about where Big Tech will recoup those costs. Advertising is the most obvious candidate. If so, advertisers could end up subsidizing AI’s massive hardware costs, often without knowing exactly how much of their budgets are going to working media versus being siphoned off to cover infrastructure.

A lack of transparency has long plagued advertisers working within walled gardens, and those concerns are likely to intensify as the pressure to fund AI ramps up. After all, both Google and Meta draw most of their revenue from ads, while Amazon’s ad business is only accelerating. OpenAI, for its part, recently said it is not “categorically” ruling out introducing ads at some point to ChatGPT — but would only do so in a “very thoughtful and tasteful” way.

That possibility has some experts uneasy. “Advertisers are going to be the ones that pay the price for it, literally, because there’s no transparency,” says Iesha White, director of intelligence at the Check My Ads Institute. White worries that prices will rise without advertisers’ knowledge. “I think we’re going to see cost per [mille and other metrics] continue to increase year over year,” she says. But that price hike will be hidden in some way by the opaqueness of the walled gardens. “In Google Ads and Meta, I have no idea what my platform fee is,” she says. “Everything is baked together.”

That lack of visibility raises a key question that advertisers ought to be asking: “How do I know that my platform fee that I can’t see has increased from, say, 5% to 10%, and how much of my money is actually going towards working media versus offsetting Meta or Google’s investment into AI?”

What this means for brands and agencies

Add on the way AI is changing how users obtain information — and therefore how advertisers can reach consumers — and it becomes a challenging situation for advertisers. Alongside potentially paying more, advertisers are now being asked to spend on untested ad products, such as placements in generative AI search results.

Recent reports suggest the few advertisers that have made it into AI-powered answer engine Perplexity’s ads program have been disappointed or underwhelmed with the results they’ve seen. “The more talk of AI and the more talk of data there is, I suspect the less value actually gets to advertisers,” Ryan says.

White compares the broader AI-driven economy to the famous rule that governs TV improv show Whose Line Is It Anyway?: “The numbers are made up and the points don’t matter. I think we’re going to see a lot that,” she says.

In response to AI creeping into ad automation at the behest of tech companies, ad agencies are starting to develop tools that bring more transparency into what’s going on under the hood. To further counter this, White recommends advertisers collectively use their power to compel changes and bring about transparency — ensuring that they know exactly what they’re paying for and why.

“I think advertisers should demand transparency,” she says. “They should demand the log-level data where possible. They should be insisting upon it.”

It’s a sentiment echoed by Ryan, who believes the AI moment is crystallizing thought around the “perennial problems” that have plagued walled garden advertising: lack of transparency, incorrect metrics and confusing results. “There are some big storm clouds ahead,” Ryan says. “The ad industry should do exactly what it should have done a terribly long time ago and remind itself that its business is advertising.” Using the moment to redress that imbalance and trying to understand what advertisers are paying for and why will be crucial, White further argues. “[Advertisers] should know how many conversions did such and such drive. I don't understand why this is controversial.”