Here comes the AI data center bubble

Welcome to Runtime! Today: OpenAI reveals its plan to build $500 billion worth of data centers in the U.S., why Oracle's decision to keep TikTok running could backfire down the road, and the latest funding rounds in enterprise tech.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Will you come if we build it?

It was hardly a surprise that one of the first things President Trump did upon returning to the White House was overturn the Biden Administration's executive order on AI, which was one of the top priorities of tech supporters like Elon Musk and Marc Andreessen. The decision paved the way for companies building AI computing, models, and algorithms to go nuts over the next four years — even beyond the frothy standards of the last two years — and they didn't waste any time.

OpenAI announced plans Tuesday to launch Project Stargate, "a new company which intends to invest $500 billion over the next four years building new AI infrastructure for OpenAI in the United States," it said in a statement. OpenAI CEO Sam Altman joined Softbank CEO Masayoshi Son and Oracle CTO Larry Ellison at the White House to announce the project alongside Trump, and OpenAI said it "will not only support the re-industrialization of the United States but also provide a strategic capability to protect the national security of America and its allies."

• Money for the project will be provided by SoftBank, OpenAI, Oracle, and Abu Dhabi's MGX, which invested in a similar project involving Blackrock and Microsoft last September.
• Arm, Microsoft, Nvidia, Oracle, and OpenAI will build the technology used in the project, according to OpenAI, and they're getting started right away in Texas with an initial $100 billion investment.
• Microsoft released a separate announcement reaffirming most of the details surrounding its partnership with OpenAI through 2030, after several reports over the last year suggested the two companies had hit a rocky patch.
• However, Microsoft is no longer the exclusive infrastructure provider for OpenAI; the companies are "moving to a model where Microsoft has a right of first refusal" for such projects after OpenAI "made a new, large Azure commitment that will continue to support all OpenAI products as well as training."

The Stargate Project arrives at a time when the progress of AI models has slowed as access to data starts to run dry, despite the enormous amounts of investment in computing capacity over the last two years. While the companies involved are primarily interested in building out AI computing capacity to dominate the future of both consumer and enterprise computing, invoking national security concerns also gives them cover to frame a literal land grab as patriotic.

• Scale AI CEO Alexandr Wang took out a full-page ad in the Washington Post Tuesday addressed to President Trump, urging an increase in AI investment because "America must win the AI war."
• “What’s undeniable, if you think about what the future is going to look like, is the degree to which the amount of computational capability you have will be directly related to how strong your AI capabilities are,” he told Semafor, a statement that applies to both companies and countries.
• However, in an editorial published the same day by MIT Technology News, Alvin Wang Graylin and Paul Triolo wrote that "now it appears that access to large quantities of advanced compute resources is no longer the defining or sustainable advantage many had thought it would be," arguing that starting a new AI compute "arms race" with China will backfire.

In a press conference Tuesday Trump suggested he would invoke "emergency" powers to make it easier to get Project Stargate up and running, which likely means the suspension of environmental reviews and energy-related regulations. But even if the federal government greases the skids, it's still going to be very hard to build $500 billion worth of data centers in the U.S. during Trump's second term.

• That level of investment dwarfs what the Big Three cloud providers have been spending on data-center construction around the world, and they are running into problems sourcing enough power and placating local opposition to see their projects through.
• Oracle has had its own struggles building data centers, and while the infusion of Softbank's capital certainly solves one of its problems it still doesn't have nearly as much experience as the Big Three in getting capacity online.
• In the late 1990s telecom companies, excited by the promise of the internet, spent and lost billions on the construction of fiber-optic cables linking the world, and it took decades for that infrastructure to pay dividends.
• Project Stargate's initial outlay might make sense, but the rest of that $400 billion sounds like deja vu all over again.

Nice cloud service you've got there

Meanwhile, last weekend's ban brinkmanship was just the latest episode in an exhausting multiyear back-and-forth over how the U.S. government should handle TikTok. Oracle, which stores TikTok user data and provides infrastructure services, had engineers standing by Saturday night awaiting an order to shut down TikTok as of midnight, when the law banning the app was supposed to go into effect, according to The Information.

TikTok was briefly offline Saturday, but came back after the incoming Trump administration indicated it wouldn't enforce the law and while TikTok can no longer be found in Apple or Google's app stores it's still serving up videos. On his first official day back in office Monday, President Trump signed an executive order delaying any enforcement of the law for 75 days and assuring TikTok's enterprise service providers that they'd be left alone, but legal experts aren't sure it's that simple.

After all, the Supreme Court upheld the validity of the law, which means service providers like Oracle and Akamai could eventually face billions of dollars in fines for keeping TikTok up and running. "[The executive order] does not stop, let's say, Oracle, from violating the law — which, as far as I can tell, it is doing right now," as Alan Rozenshtein of the University of Minnesota Law School told NPR, and Bloomberg Intelligence's Matt Schettenhelm noted that the companies are relying "on a promise from a president who flipped 180 [degrees] on this very issue, & under statute of limitations that outlasts his presidency," according to Variety.

Enterprise funding

Aligned Data Centers raised $12 billion in debt and equity funding to expand its data-center operation "across the Americas."

Synthesia scored $180 million in Series D funding as it builds out its AI-generated video platform, which is used by corporate customers for sales and marketing videos and for internal communication.

Instabase landed $100 million in Series D funding with plans to expand "the automation, analysis, and search capabilities of its unstructured data platform."

Render raised $80 million in Series C funding to improve its cloud application development platform with more bells and whistles for enterprise-grade customers.

Prophecy scored $47 million in an extension to a previous Series B round for its data-management tool, which provides a visual user interface to make building data pipelines easier.

Quantum Brilliance raised $20 million in Series A funding to support its goal of building a quantum computer that uses "quantum diamonds," which could allow its quantum computers to run at room temperature.

The Runtime roundup

HPE is looking into claims that a hacker made off with sensitive data, including source code, but hasn't found any evidence as of yet to support those claims, according to Bleeping Computer.

AWS halted plans to expand its data-center network in Israel, according to Globes.

A database issue caused a worldwide outage for Atlassian Bitbucket customers over several hours until service was restored Tuesday afternoon.

Thanks for reading — see you Thursday!