Project Stargate — the headline-grabbing $500 billion plan to build massive AI data centers across the U.S. — may have launched with far more ambition than preparation. A new report says the initiative has been slowed by partner disputes, financing hurdles, and shifting strategies as OpenAI scrambles to secure the computing power needed for the AI arms race.
A 1-gigawatt facility is now under construction in Texas, but insiders say the original announcement came before a concrete roadmap existed, leaving partners to figure things out on the fly.
From moonshot to workaround
When plans to build data centers directly ran into resistance from lenders wary of the enormous risk, the structure reportedly shifted. Oracle $ORCL ( ▲ 0.89% ) now borrows the money and leases capacity back to OpenAI, allowing the AI company to access compute without carrying the full debt load — while still influencing the facility’s design.
SoftBank $SFTBY ( ▼ 1.82% ) , another major partner, has also been part of the evolving arrangement as the project seeks stable footing.
Falling behind in the AI infrastructure race
The delays have real consequences. OpenAI reportedly missed its own target of securing 10 gigawatts of computing capacity from partners by the end of 2025 — a massive shortfall in an industry where access to power and chips increasingly determines who leads.
In short, Stargate illustrates the messy reality behind AI’s glossy headlines. Building the physical backbone of artificial intelligence isn’t just a software problem — it’s a capital-intensive, utility-scale undertaking that looks more like constructing power plants than launching an app.