Stargate Project Major Shift: OpenAI Abandons Building, Pivots to Leasing, the $1.4 Trillion Computing Power Empire Dream Awakens

$14 trillion. This is the total estimated value of OpenAI CEO Sam Altman’s Stargate computing infrastructure plan presented to investors by the end of 2025. 14 months later, this figure was cut to $600 billion.

According to The Information on March 16, OpenAI has undergone a major restructuring of the Stargate computing infrastructure project, abandoning plans to build its own data centers and shifting entirely to leasing computing power from cloud providers like Microsoft Azure, Oracle, and Amazon AWS. Stargate has been split into three functional teams, all under the unified management of former Intel Chief Technology and AI Officer Sachin Katti.

The direct reason for the shift isn’t complicated. Stargate was publicly announced at the White House in January 2025, with plans to jointly build large data centers with SoftBank and Oracle, initially investing $100 billion, with a total investment of $500 billion over four years. But more than a year after the project started, no employees had been hired, and no substantial data center was built. According to CNBC, lenders were unwilling to provide billions of dollars in construction financing for a company still operating at a huge loss. Earlier this month, OpenAI also withdrew from negotiations to expand the Oracle Stargate project in Abilene, Texas.

Over a year, with zero employees and no construction, the “self-build” path for Stargate never truly got off the ground.

According to investor materials, Altman’s $1.4 trillion total commitment is distributed among seven suppliers. Venture capital analyst Tomasz Tunguz analyzed the investor materials and found that Broadcom accounts for $350 billion, Oracle $300 billion, Microsoft $250 billion, NVIDIA $100 billion, AMD $90 billion, and AWS and CoreWeave combined account for $60 billion.

In February 2026, CNBC reported that this number was reset to approximately $600 billion (by 2030), a 57% reduction. The same report provided a slightly different but consistent figure, estimating that OpenAI will spend $665 billion on cloud servers by 2030.

$600 billion remains a figure that needs anchoring to truly grasp. According to internal OpenAI forecasts, the company’s revenue target for 2030 is $280 billion, implying a roughly 2:1 ratio of five-year cumulative expenditure to revenue. Internal financial data cited by ainvest predicts a loss of $14 billion in 2026, with gross margins reportedly only 33% (note: gross margin reflects the profitability of the product itself, while net loss is the final result after deducting all costs like R&D and management; both can coexist).

Placing OpenAI’s expenditure goals within the broader Big Tech arms race for computing power makes the proportional relationships clearer.

Based on financial reports and public guidance from various companies, Amazon plans to spend $200 billion on capital expenditures in 2026, Alphabet $180 billion, Meta $125 billion, and Microsoft about $120 billion. These four companies have generally doubled or tripled their spending over two years, totaling over $650 billion, with roughly three-quarters directed toward AI infrastructure.

OpenAI’s $600 billion is a five-year cumulative target, averaging about $120 billion annually, comparable to Microsoft’s annual capital expenditure. The difference is that Microsoft’s annual revenue exceeds $240 billion, while OpenAI’s annualized revenue just reached $25 billion, and it is not expected to turn cash flow positive before 2030.

The Stargate restructuring is not just a change in budget figures; the organizational adjustments reveal a deeper strategic shift.

Post-restructuring, Stargate is divided into three lines. The Epic Business Partnerships team is led by OpenAI veteran and former Deloitte manager Peter Hoeschele, managing cloud contracts with Microsoft, Oracle, and Amazon, as well as deals with chip manufacturers, including multi-year agreements with AMD (using up to 6 gigawatts of chips at a cost of up to 10% of AMD’s common stock) and a partnership with chip startup Cerebras Systems.

The Technology Engineering and Design team is jointly led by former Meta and Google engineer Chris Malone and former Microsoft engineering director Adrian Caulfield, responsible for redesigning OpenAI’s AI server clusters. The Physical Infrastructure Operations team is led by former Google data center director Nick Saddock, replacing Keith Heyde who left weeks earlier.

The semiconductor team, led by former Google chip executive Richard Ho, is outside Katti’s scope and reports directly to OpenAI President Greg Brockman. This team is collaborating with Broadcom to develop in-house chips, aiming to reduce inference costs for products like ChatGPT.

The name “Stargate” remains, but what it refers to has completely changed. In January 2025, it was a joint venture with SoftBank and Oracle to build data centers. By March 2026, it had become a broad strategy to deploy gigawatt-scale server capacity. From “I want to build my own power plant” to “I want to sign the best lease.” The total planned capacity across all sites remains nearly 7 gigawatts, with a total investment exceeding $400 billion over three years. OpenAI is shifting its compute focus toward NVIDIA’s Vera Rubin platform, aiming to launch its first gigawatt-scale capacity by the second half of 2026.