Stargate's Troubled Launch
The highly anticipated Stargate data center project, envisioned as a groundbreaking AI infrastructure initiative costing a staggering $500 billion, has
encountered considerable difficulties. Originally announced with significant fanfare, the venture is now grappling with delays and complications in securing necessary funding. A core issue has been internal disagreements concerning control and ownership, which have led to partners forging individual agreements outside the main joint venture. This has created financial pressures for key collaborators like Oracle. The initial ambition was to establish 10 gigawatts of AI data center capacity by the end of 2025, with an initial investment of $100 billion. However, the project has only managed to secure approximately 7.5 GW. Furthermore, the projected compute expenses through 2030 have escalated dramatically, rising from an initial $450 billion to a much higher $665 billion. The Stargate joint venture itself, intended to be the central coordinating body for these efforts, has remained largely inactive, as OpenAI has pursued direct agreements with various entities.
OpenAI's Solo Endeavor Fails
Facing a stalemate within the partnership, OpenAI explored the possibility of independently developing and operating its own data centers. The company scouted potential locations across the United States, engaged a general contractor, and began laying the groundwork for large-scale campus development. The objective was to reduce reliance on external cloud providers, which incur higher long-term costs, and to gain direct command over its computational resources. However, financial institutions were hesitant to back such ambitious construction projects. Lenders expressed reservations about financing multi-billion dollar developments tied to a company that was consuming billions in cash annually—an estimated $8 billion in 2025 alone—without a clear, proven path to consistent profitability. Consequently, OpenAI had to abandon its self-building strategy, and the project lost its general contractor due to the protracted delays. This situation compelled OpenAI to return to the negotiation table, but instead of working through the established Stargate joint venture, the company opted to bypass it entirely. In mid-2025, a direct agreement was made with Oracle for 4.5 GW of capacity across several US sites, which later evolved into a substantial $300 billion, five-year cloud computing contract. Additionally, separate compute agreements were established with Amazon Web Services and Google Cloud to address immediate capacity needs, and a $22.4 billion arrangement was secured with CoreWeave. In total, OpenAI has committed approximately $600 billion across various infrastructure agreements, all negotiated outside the framework of the joint venture originally intended for centralized coordination.
Ownership Disputes Emerge
The Texas campus, initially conceived as OpenAI's flagship self-constructed data center, became a focal point of contention. Following the pause in construction due to Oracle's negotiations and the departure of OpenAI's contractor, SoftBank was brought in as a partner for the site. However, immediate disagreements arose over the ownership structure. OpenAI sought to maintain control over the facility, while SoftBank aimed to develop and own it. Reports suggest that these impasses necessitated multiple trips to Tokyo between September and October, involving extended negotiation sessions at SoftBank's headquarters. A compromise was eventually reached: SoftBank would assume ownership and development of the Texas site, while OpenAI would oversee its design and secure a long-term lease. Despite this agreement, SoftBank encountered regulatory obstacles that forced it to halt a planned $50 billion acquisition of data center operator Switch. During the announcement period, Elon Musk publicly questioned the project's financial backing, suggesting SoftBank had secured far less than the reported amount. Altman responded by inviting Musk to visit the initial site and subtly referencing the differing priorities between national interests and individual companies.
Oracle's Financial Strain
Oracle, a key financial backer of the Stargate project, is experiencing significant repercussions. The company has accumulated over $100 billion in debt, including $38 billion allocated to data centers in Texas and Wisconsin and $20 billion for a New Mexico campus. Its stock value has plummeted by approximately half since its peak in September 2025, resulting in a market capitalization loss of around $463 billion. S&P Global has placed Oracle's BBB credit rating under negative watch, and some of its bond offerings, such as a 10-year note from September, are trading at levels typically associated with high-risk debt. The cost of credit default swaps, which serve as insurance against default, has surged and remained elevated. Investor Michael Burry is reportedly betting against the company's stock. TD Cowen, an investment bank, estimates that the OpenAI contract alone will necessitate roughly $156 billion in capital expenditure and approximately 3 million GPUs. To manage these costs, Oracle may reduce its workforce by 20,000 to 30,000 employees, potentially generating $8–10 billion in cash flow. The company is also reportedly considering divesting Cerner, its healthcare software division acquired for $28.3 billion in 2022, and has begun requesting upfront payments of up to 40% of contract value from new clients. Oracle's attempts to mitigate the negative impact have often backfired; a statement in early February asserting that the stalled Nvidia-OpenAI deal had no effect on its financial relationship with OpenAI caused its shares to drop nearly 3%. Venture capitalist Alex Kolicich described the situation using language akin to a bank run.
Revised Projections and Funding
A clear indication of a recalibrated approach is OpenAI's significant reduction in its spending forecasts. After previously projecting $1.4 trillion in infrastructure commitments, the company is now informing investors that its total compute spend target through 2030 is approximately $600 billion, a figure less than half of the initial estimate. While OpenAI achieved revenues of $13.1 billion last year, exceeding its $10 billion target, it still incurred $8 billion in expenses. The company now anticipates generating $280 billion in revenue by 2030, with contributions evenly split between consumer and enterprise segments. Furthermore, OpenAI is in the process of finalizing a substantial funding round expected to surpass $100 billion, valuing the company at $730 billion before the investment. Nvidia is reportedly in discussions to invest up to $30 billion, replacing a previously announced non-binding $100 billion multi-year commitment. SoftBank is also negotiating an additional $30 billion investment, and Amazon may contribute up to $50 billion as part of a broader strategic partnership. The intricate financing structure continues to raise concerns among analysts, as the interdependence between investors and OpenAI creates a fragile ecosystem. The current arrangement, where Nvidia invests in OpenAI and buys its own chips, and Oracle builds data centers for OpenAI and charges for their use, means that the financial stability of each entity relies heavily on the others fulfilling their obligations. As one market observer noted, the system is sustainable only until one party falters, triggering a widespread impact.
