OpenAI is scrambling to secure computing power after its Stargate data center venture stalled, turning to cloud partners and alternative hardware to fill the gap, according to The Information.
Gridlocked from the start
Stargate — a US$500 billion joint project with Oracle and SoftBank, announced at the White House in early 2025 — was designed to deliver 10GW of AI computing capacity. More than a year on, the venture remains largely dormant. No dedicated team exists, no facilities are under development, and disagreements over leadership, structure, and financial responsibility have left the partnership gridlocked.
The debt dead end
OpenAI initially looked to build and own its own data centers. Executives scouted US sites and explored raising billions in debt to fund large-scale campuses. Lenders balked, however, wary of backing a cash-burning company without a proven long-term business model. OpenAI shelved those plans.
Pivoting to partnerships
In July 2025, OpenAI struck a landmark deal with Oracle to jointly develop 4.5GW of data center capacity across multiple US sites. The two companies share construction risks and cost overruns, giving OpenAI a say in facility design without the full capital burden.
OpenAI has since deepened its cloud reliance. Reuters reports the company signed additional compute deals with Amazon Web Services and Google Cloud to cover near-term shortages. Bloomberg, meanwhile, says OpenAI is also moving beyond Nvidia chips, partnering with AMD and AI accelerator startup Cerebras for alternative hardware supply.
Falling short — and spending more
Despite these moves, OpenAI missed its target of locking in 10GW of capacity by the end of 2025, securing only about 7.5GW. The shortfall hit its finances hard. The company revised its projected compute spending through 2030 from US$450 billion to US$665 billion.
At the World Economic Forum in Davos, OpenAI CFO Sarah Friar confirmed the shift, saying the company is prioritizing partnerships to keep its balance sheet lean. Building its own facilities remains a long-term goal — not an immediate one.
Competition closes in
The reset comes as rivals move fast. According to The Wall Street Journal, Google DeepMind and Anthropic have rapidly expanded their computing footprints, raising fears that OpenAI's delays could erode its technological edge.
To regain ground, OpenAI hired former Intel executive Sachin Katti to lead its infrastructure organization. The Information says the appointment is aimed at tightening control over its compute roadmap and data center design.
For now, OpenAI's strategy is control without ownership: locking in priority access, custom designs, and long-term capacity — while letting partners carry the financial weight.
Article edited by Jerry Chen
