OpenAI and Oracle unveil plans to develop 4.5GW of additional US data center capacity in an expanded partnership, hitting 5GW in total and running 2M AI chips
OpenAI and Oracle Corp. announced they will develop 4.5 gigawatts of additional US data center capacity in an expanded partnership …
It also follows Oracle’s reported plan to deploy about 400,000 GB200 chips at a 1.2GW Texas site for OpenAI, making this a move from an initial site toward a much larger powered-capacity program.
First-order effects
OpenAI gains a stated path to 5GW of U.S. capacity and roughly 2 million AI chips, while Oracle takes on a substantially larger role in supplying its compute infrastructure.
The partners must turn a capacity target into data-center development, power procurement, and chip deployment, extending the operational scope of their Stargate-linked work.
Second-order effects
The scale concentrates more of OpenAI’s near-term infrastructure demand with Oracle, raising the strategic value of Oracle’s ability to secure powered sites and bring them online.
Large commitments such as the reported 1.2GW Texas deployment put greater pressure on the constrained inputs behind AI buildouts—especially powered data-center capacity and accelerator delivery—rather than cloud availability alone.
Third-order effects
If comparable multigigawatt commitments continue, AI providers and cloud operators will increasingly compete over long-duration access to electricity and ready-to-build sites, treating compute capacity more like utility infrastructure.
The Stargate model points toward deeper partnerships that combine model demand, cloud operations, financing, and physical buildout; its durability will depend on whether planned capacity can be energized and utilized on schedule.
The trend:AI infrastructure is shifting from incremental cloud expansion to multiyear, power-constrained capacity commitments anchored by a small number of model builders and infrastructure partners.
It's official: we're developing 4.5 gigawatts of additional Stargate data center capacity with Oracle in the U.S (for a total of 5+ GWs!). And our Stargate I site in Abilene, TX is starting to come online to power our next-generation AI research. https://openai.com/...
I am, at this point, calling into question: — 1. Whether Stargate ever existed, including the supposed $18bn-$19bn commitments promised by SoftBank/openAI — 2. Whether SoftBank ever had any involvement with Abilene Texas or *ANY* OpenAI data center — 3. Whether SoftBank act…
OpenAI will, starting in 2028, have to start paying $30bn a year to Oracle, $28bn to Microsoft for Azure, and also $2.9bn to CoreWeave for compute. That's $60.9 billion *in compute costs alone*. This company is an absolute joke and should be treated as such. — www.theinformat…
Are the disagreements on deal terms because: — 1. OpenAI has no moat and is burning money without a realistic path to profitability? — 2. SoftBank can't really afford such a risky investment when gen-AI is yet to find a product/market fit? — 3. Something else? [embedded pos…
NEWS: A $500 billion effort unveiled at the White House to supercharge the U.S.'s artificial-intelligence ambitions has struggled to get off the ground and has sharply scaled back its near-term plans. After 6 months, they have yet to complete a single deal for a data center. [ima…
Masayoshi Son is in the mob. His father ran illegal alcohol distribution on illegally obtained land from the Japanese government, then Masa appeared in California and mysteriously graduated high school in just one month by taking “special” exams. https://www.wsj.com/...