OpenAI's Stargate blew past its 10 GW goal already!

Hey everyone, it's Shii-chan! Today's story is a big-picture one: it's about the very foundation that makes AI run. I got excited reading about just how much compute OpenAI is stacking up!

OpenAI News openai.com

What was announced?

OpenAI's News shared an update on Stargate, the compute infrastructure that powers the Intelligence Age. Stargate is OpenAI's long-term effort to build the compute foundation needed to deliver the benefits of AGI to the world.

When OpenAI announced Stargate in January 2025, they committed to securing 10 GW of AI infrastructure in the United States by 2029. Just over a year later, they've already passed that milestone, with more than 3 GW added in the last 90 days alone. The momentum is wild!

Why it matters

For me, the key point is the claim that compute is the critical input that makes advanced AI possible.

OpenAI calls this the "AI flywheel": more compute enables better models, better models drive more usage, more usage improves products and revenue, and that lets them reinvest in more infrastructure. So the simple reality underneath it all is this: if you want to bring AI to more people, you need more compute.

What changes

The big shift is that, looking beyond the initial 10 GW goal, OpenAI and its partners are now evaluating data center sites across the country, planning to significantly expand compute capacity over the coming years.

And they're not building alone. They emphasize a partner-centric approach that brings in local communities, utilities, chipmakers, cloud providers, construction firms, investors, and skilled trades. This week they donated to a local education foundation (the Port Washington-Saukville Education Foundation) in Wisconsin alongside Vantage Data Centers and Oracle, and they're working with North America's Building Trades Unions (NABTU) to open new job pathways, not just for engineers and researchers but for skilled workers and apprentices too.

Dive Deep

The most concrete and fun part is the Stargate site in Abilene, Texas, where the water story is thoughtfully designed.

Instead of traditional evaporative cooling towers, the site uses closed-loop cooling. Once filled, the water keeps circulating through sealed pipes and is recirculated rather than consumed.

The numbers are charming: the one-time initial fill for each building equals about two Olympic-sized swimming pools. After that, annual water use at full buildout is expected to be comparable to a medium-sized office building, or about four average households. It kind of flips the "data centers guzzle water" image, doesn't it?

And finally, from infrastructure to intelligence: OpenAI's latest and smartest model, GPT-5.5, was trained at this flagship Abilene site. The site runs on Oracle Cloud Infrastructure with NVIDIA GB200 systems.

GPT-5.5 aims to close the capability overhang — the productivity gap between AI power users and everyone else — so the economic opportunities of AI can be broadly shared.

Wrap-up

• Stargate builds the compute foundation for AGI; the 10 GW-by-2029 goal was hit ahead of schedule, just over a year after launch
• More than 3 GW added in the last 90 days, with data center sites being evaluated across the country
• Partner-centric and community-focused: the Wisconsin donation and NABTU collaboration give back locally
• The flagship Abilene site uses closed-loop cooling to go easy on water, and GPT-5.5 was trained there on Oracle Cloud and NVIDIA GB200

This isn't a "try it yourself this weekend" feature, but if you're curious about the physical body — the compute, energy, and water — that AI actually runs on, it's a great read. Engineers curious about the infrastructure behind the scenes should check it out!