Observe ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- The information facilities will collectively present 5.5 gigawatts of energy.
- They’re the newest part of the so-called Undertaking Stargate.
- Knowledge facilities require large portions of water for cooling.
OpenAI is pushing ahead with its mission to energy more and more giant AI fashions with the development of 5 new information facilities throughout the US, the corporate announced Tuesday.
Additionally: How Nvidia and OpenAI’s staggering $100 billion deal could fuel a new age of AI
The announcement of the brand new information facilities marks the newest growth in Project Stargate, the AI infrastructure plan OpenAI launched(?) in January alongside the Trump administration, Oracle, SoftBank, and Emirati funding agency MGX. The undertaking secured $500 billion in funding to construct information facilities over the subsequent 4 years that can cumulatively generate 10 gigawatts of energy — sufficient to gas roughly 7,500,000 properties for one 12 months.
The 5 new information facilities — which shall be constructed, respectively, in Shackelford County, Texas; Dona Ana County, New Mexico; Lordstown, Ohio; Milam County, Texas; and an undisclosed location within the Midwest — will present 7 gigawatts of energy to be put towards OpenAI’s proprietary AI fashions. In line with OpenAI, that quantity retains Stargate “forward of schedule” to get to 10 gigawatts by the tip of 2025.
“Collectively, these websites are anticipated to create over 25,000 onsite jobs, and tens of hundreds of extra jobs throughout the US,” OpenAI wrote in its announcement, including that it is nonetheless evaluating extra websites throughout the nation. US chipmaker Nvidia introduced on Monday that it could invest up to $100 billion in OpenAI for the development of extra information facilities.
Additionally: Will AI damage human creativity? Most Americans say yes
(Disclosure: Ziff Davis, ZDNET’s guardian firm, filed an April 2025 lawsuit in opposition to OpenAI, alleging it infringed Ziff Davis copyrights in coaching and working its AI programs.)
A race to construct
As tech firms’ ambitions amid the continued AI growth have escalated, so too have their vitality payments.
Additionally: Worried about AI’s soaring energy needs? Avoiding chatbots won’t help – but 3 things could
It takes a lot of energy to train and run AI models. Builders have been locked in a race with China to remain forward of AI growth, which suggests constructing large native information facilities full of graphics processing models (GPUs) to power AI tools like ChatGPT and Gemini and develop extra superior fashions. That race has reworked US chipmaker Nvidia into the wealthiest firm on the planet.
It additionally has main environmental implications. Knowledge facilities are vulnerable to overheating, which suggests they require big portions of water for cooling. On high of that, they’re typically constructed in arid Western states, which have ample land however are additionally vulnerable to seasonal droughts. Many information facilities are powered by fossil fuels, exacerbating international greenhouse impact and local weather change.
Microsoft introduced plans final September to reopen Three Mile Island, the nuclear energy plant and the location of the notorious partial reactor meltdown in 1979, to energy its personal AI ambitions.
Additionally: These consumer-facing industries are the fastest adopters of AI agents
OpenAI CEO Sam Altman and different AI business leaders are inclined to justify the financial and ecological prices of constructing information facilities by claiming that the long-term advantages of the expertise shall be price doing every part inside our energy proper now to construct infrastructure as rapidly and as copiously as attainable.
As Altman wrote in a personal blog post on Tuesday: “Perhaps with 10 gigawatts of compute, AI can work out methods to remedy most cancers. Or with 10 gigawatts of compute, AI can work out methods to present custom-made tutoring to each scholar on earth. If we’re restricted by compute, we’ll have to decide on which one to prioritize; nobody desires to make that alternative, so let’s go construct.”
“Perhaps” could be very a lot the operative phrase there.
