AI's trillion-dollar data center bet
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Sarah Grillo/Axios
With their latest deals to fund the data center boom, AI firms are making history — in terms of dollar size, convention-busting structure, and astronomical risk.
Why it matters: The U.S. is betting its economic fortunes on the belief that OpenAI's Sam Altman, Nvidia's Jensen Huang and other AI leaders are wizardly innovators dreaming up novel financing vehicles to drive a golden future — rather than salesmen juggling billions and praying the music never stops.
Friction point: The trillion-dollar question neither Silicon Valley nor Wall Street can answer is whether the AI building spree will end up looking more like Google's epochal long-term value creation or Enron's catastrophically faulty financial engineering.
Driving the news: Nvidia announced Monday it would invest up to $100 billion in OpenAI in stages, with OpenAI using the money to "build and deploy at least 10 gigawatts of AI data centers with Nvidia systems."
- OpenAI also announced Tuesday additional commitments for its Stargate alliance with Oracle and SoftBank to build five new U.S. data centers. This brings the sum it has lined up for the effort $400 billion of the way toward a $500 billion goal.
- Google, Microsoft, Meta and Amazon have all projected a total of hundreds of billions more in capital expenditures on data centers for AI.
Yes, but: Some market observers are asking questions about the closed-loop appearance of the OpenAI-Nvidia deal.
- "Nvidia to pay openai so they can get paid by softbank so they can pay oracle to pay nvidia," CNBC's Steve Kovach joked on Bluesky.
- The deal's structure "will clearly fuel 'circular' concerns," one Wall Street analyst wrote.
Flashback: At the raging peak of the dotcom-era bubble in the late '90s and early 2000s, online ad giants like AOL Time Warner and high-flying telecoms like Qwest were accused of inflating revenue figures by essentially flooding customers with cash that the customers would then spend on ads or services.
- Such parallels haunt today's AI boom and reinforce the skeptical view that this investment bonanza might be a shell game and can't last forever.
The other side: The big bucks flowing into data centers are coming not from paper gains — the skyrocketing IPOs fueled by day-trading speculators that fueled the '90s boom and then evaporated — but from tech giants' profit-fed cash hoards.
- "These companies aren't mortgaging the future; they're spending current winnings," Semafor's Liz Hoffman notes.
- And in the case of Nvidia's OpenAI deal, the chipmaker is getting a stake in the ChatGPT maker — an investment opportunity many other companies are also taking right now (or wish they could).
Big Tech has a long history of cyclical overinvestment, most famously with the turn-of-the-millennium fiber buildout.
- As this wheel turns, fortunes are made and lost in a flash, depending on investors' timing.
But even if today's AI market goes south, data centers are durable goods — hardware and buildings.
- The companies that own them can still use them to bootstrap whatever technology wave comes next.
Our thought bubble: The real unknown in this gigantic equation is AI consumption.
- If OpenAI and Nvidia are right in forecasting a hockey-stick growth curve for AI demand in the short and medium term, then their outsize investment plans probably make sense. (That's putting aside AI's other grim worries — environmental costs, social harms and existential fears.)
- But at this point the industry's whole project is built on best-case assumptions.
- Tremors of doubt — slowdowns in chatbot adoption, evidence that AI isn't boosting productivity, a public rebellion against AI "slop" — could bring the game to a messy end.
