OpenAI reaches 10GW capacity milestone early as Stargate strategy shifts from ownership to leasing

OpenAI Accelerates Past Computing Milestone

OpenAI has secured 10 gigawatts AI computing capacity in the United States, reaching a target originally set for 2029 several years ahead of schedule 1

1

. The ChatGPT creator announced the milestone in a blog post on Wednesday, revealing that 3 of the 10 gigawatts under contract had been secured in the past 90 days alone 1

1

. This includes 2 gigawatts from Amazon Web Services and a further 1 gigawatt from an undisclosed provider. A single gigawatt is enough electricity to power approximately 750,000 US homes at any one time 1

1

.

The rapid expansion of AI computing capacity is essential for training advanced AI models, scaling deployment, and meeting growing demand from businesses, developers, and governments 3

3

. Sam Altman's company stated it is assessing additional data center sites across the country and "will significantly expand our compute capacity in the years ahead" 1

1

.

The $500 Billion Stargate Initiative Transforms

What began as the $500 billion Stargate initiative has undergone a dramatic shift in strategy. Originally unveiled in January 2025 by Donald Trump as a joint venture among OpenAI, Oracle, SoftBank Group Corp., and Abu Dhabi fund MGX, Stargate was designed to pool capital to finance and build dedicated data centres for OpenAI's exclusive use 2

2

. The concept aimed to tackle the cost and complexity of AI infrastructure by having OpenAI own its facilities.

That vision has rapidly given way to a more flexible approach, with OpenAI increasingly relying on leasing capacity from third-party providers rather than building and owning its own facilities 2

2

. "I don't know what 'Stargate' means at this point," said a person who was involved in the data center buildout at the outset. "I think it is a completely antiquated line right now" 2

2

. The project has morphed into a catchall brand for all of OpenAI's data center expansion plans 1

1

.

Partnerships With Cloud Providers Replace Ownership Model

OpenAI has abandoned the joint venture in favor of a series of large bilateral deals with multiple partners 2

2

. Oracle has played the most prominent role, agreeing to supply OpenAI with 4.5 gigawatts of computing capacity under a five-year, $300 billion agreement 2

2

. The company has also struck partnerships with AMD, Broadcom, NVIDIA, CoreWeave, and Cerebras in deals that at one point totalled more than $1 trillion in commitments, some of which have since been scaled back 2

2

.

Microsoft and OpenAI recently agreed to drop the software giant's exclusive right to sell the startup's AI models, opening the door for securing computing power from cloud-computing rivals 1

1

. In exchange for ending that exclusivity, Microsoft will no longer pay a revenue share on OpenAI products it resells on its cloud 1

1

. Microsoft will receive a revenue share from OpenAI through 2030, regardless of progress toward artificial general intelligence 1

1

. Microsoft remains OpenAI's "primary cloud provider," and new products will be made available first on Azure 1

1

.

Abandoned Projects and Rising Concerns

OpenAI has recently pulled back from several projects related to the highly publicized initiative. The company declined to lease an expansion to Stargate's flagship Abilene data center site in Texas and paused another effort in the UK, citing the country's high energy costs and regulation 1

1

. OpenAI also halted planned data centres in Norway and saw several senior figures tied to Stargate leave for rival Meta 2

2

.

The willingness to renegotiate or walk away from projects has unsettled partners and raised questions about OpenAI's reliability as a counterparty 2

2

. Questions remain over whether the company can afford its huge infrastructure spending, particularly after a recent Wall Street Journal report suggested OpenAI missed internal targets for revenue and user growth 2

2

3

3

. OpenAI responded that its business is "firing on all cylinders" 2

2

.

The shape-shifting scheme embodies OpenAI's broader approach, which owes much to Sam Altman's "venture mindset" of placing overlapping bets and abandoning all but the most expedient, according to a person close to the company 2

2

. OpenAI defended its strategy, stating: "Stargate is the umbrella for our compute strategy and we've done exactly what we said we would do: secure the compute OpenAI needs at unprecedented scale" 2

2

. The company's relationship with Microsoft is set to be in the spotlight as the companies square off against Elon Musk in court, who is seeking as much as $134 billion in damages 1

1

.