Opinions come pouring in on $500 billion Stargate Project, as DeepSeek throws a curve

Last week, just one day into the new administration, the president announced an ambitious plan dubbed Project Stargate that involves to some degree the government, OpenAI, Oracle, SoftBank and MGX investing $500 billion over the next four years to build AI data centers in the U.S.

Meanwhile, Chinese company DeepSeek sent shock waves through the industry when it released a new, cheaper approach to building AI models, one that could call into question whether investments this large are even needed. 

Regardless, challenges aplenty await a project of this magnitude, especially when one company, OpenAI, draws the vast majority of the investment attention. The company wasted no time getting to the heart of the matter in a blog post trumpeting the details. 

"The Stargate Project is a new company which intends to invest $500 billion over the next four years building new AI infrastructure for OpenAI in the United States. We will begin deploying $100 billion immediately," it wrote.

While it sounds impressive on paper, John Dinsdale, chief analyst at Synergy Research, a firm that follows all things cloud, including data center construction, says it will come down to the details. "What I saw was all a bit fluffy and lacking in detail. It will get more interesting if and when more concrete plans are put in place, and there is progress in terms of sites, approvals, power supply, construction, funding and service announcements."

There were also questions raised about the viability of the funding sources from multiple quarters. For starters, Axios reported, presidential advisor Elon Musk, who has his own AI company, xAI, waded in questioning where the money will come from, drawing ire from some in the administration. But investor and former TechCrunch reporter M.G. Siegler, writing in his Spyglass newsletter, also wondered the same thing as Musk:

“I would imagine that SoftBank and MGX are the two largest and real capital commits here. But neither seemingly has enough capital on hand to equal a $100B commit, let alone $500B. SoftBank has about $30B in cash. MGX started last year as a $100B fund, but they've already made several commitments, notably to OpenAI itself. But they also apparently can mobilize more capital as needed, which seems key here. Especially since Oracle "only" has about $11B in cash and OpenAI is a cash-burning machine at the moment,” Siegler wrote.

Patrick Moorhead, founder and principal analyst at Moor Insights & Strategies agreed that details were light, especially around the financing, but he also wondered why OpenAI is the chief beneficiary of this investment, one that appears to have the backing of the administration.

“It strikes me as odd that the number is so huge and that OpenAI is bleeding cash. It’s not like OpenAI has 10x the capability of everyone else. They have the lead, but Llama (from Meta) and Gemini (from Google) are quickly catching up if not surpassing OpenAI [in terms of model performance],” Moorhead said.

As Tom Krazit pointed out in his RunTime Newsletter, it's not just about money or politics either. Even if the administration relaxes environmental rules, these kinds of data centers require huge tracts of land along with a ready supply of energy, water and other resources. Local officials also have to buy in on projects of this magnitude, and there is a growing resistance against massive data centers as communities begin questioning their value.

While this is a big swing, Danny Crichton, writing in his Riskgaming newsletter, says that if this is about taking control of the market, OpenAI might have its work cut out for it, especially now that DeepSeek is showing that this could potentially be done much cheaper.

And as Jon Turow from Madrona wrote in a blog post on the DeepSeek model, it offers the possibility of more powerful and focused models without requiring the raw compute power of models like OpenAI. "DeepSeek's approach suggests how deep domain expertise might matter more than raw compute in building the next generation of AI models and intelligent applications," Turow wrote.

This shows that it might not be about just how much money and compute power you can throw at the problem, and that could be a shock for companies like OpenAI, and possibly chipmaker Nvidia too, if that turns out to be the case. If DeepSeek turns out to be the real deal, and other disruptive companies start entering the market, the costs associated with generative AI could be dropping for everyone, and that could drop one more barrier to entry, and make the capital problem become much less of a factor.

Photo by Robert V. Ruggiero on Unsplash

Note: This is an expanded version of a News Byte article that appeared in the 1/24/2025 FastForward Newsletter.