OpenAI and SoftBank Group partner with SB Energy
OpenAI
OpenAI and SoftBank just teamed up with SB Energy to build massive AI data center campuses. First one's a 1.2 GW site in Texas, feeding the Stargate project.
Based on reporting by OpenAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The compute arms race just got a new power hookup. OpenAI and SoftBank Group have brought in SB Energy, SoftBank's US energy arm, to build out multi-gigawatt data center campuses purpose-built for AI workloads. The flagship project is a 1.2 gigawatt facility in Texas, and it's slotting directly into Stargate, the sprawling infrastructure initiative OpenAI and SoftBank announced earlier this year.
To put 1.2 GW in perspective, that's roughly the output of a large nuclear reactor, dedicated to a single campus running AI training and inference. It's a reminder that the real bottleneck for frontier AI isn't just chips anymore. It's electrons. GPUs are useless without power delivered at massive, reliable scale, and that requires the kind of energy engineering that companies like SB Energy specialize in rather than the cloud providers themselves.
This partnership fits a pattern OpenAI has been building since Stargate launched: pairing with infrastructure and energy specialists instead of trying to become a utility company overnight. SoftBank brings capital and its existing energy portfolio through SB Energy, while OpenAI brings the demand forecast that justifies building gigawatt-scale campuses in the first place. Texas, with its deregulated grid and existing renewable buildout, is a logical first landing spot.
What's notable is the framing around scale. Nobody's talking about a single data center anymore. They're talking about campuses, plural, multi-gigawatt, as if a few hundred megawatts is now a rounding error. That escalation says a lot about where OpenAI thinks demand for its models is headed over the next several years, and it puts pressure on rivals to secure their own power deals before the available capacity gets locked up.
My take — AI-written commentary, not fact-checked reporting
This is the real Stargate story: it was never really about chips, it was about who controls gigawatts. I'd rather see that capacity funneled into transparent, efficient models than another round of scaling-for-scaling's-sake, but at least building the power layer honestly instead of pretending compute magically appears is a step in the right direction.
Read more about this at: OpenAI