Oracle and OpenAI Abandon Data Center Expansion—Is This the 'First Crack' in the AI Infrastructure Investment Boom?
In 2025, it was revealed through a Bloomberg report that OpenAI and Oracle, the companies driving the AI industry, have abandoned plans to expand their flagship data center, "Stargate," currently under construction in Texas. The plan had called for an increase in capacity from 1.2 gigawatts to 2 gigawatts, but both companies decided to forgo the additional capacity due to prolonged financing negotiations and shifting computing needs.
Construction of the existing campus is continuing as planned, and the overall partnership between Oracle and OpenAI is said to remain intact. However, this move highlights the complexity and risks of executing AI infrastructure projects even as investment accelerates. This article examines the background of this event and its implications for investors from multiple perspectives.
1. What Happened—The Context and Structural Background of the Abandoned Expansion
1-1. A Decision Made After Months of Negotiations
According to Bloomberg, negotiations between Oracle and OpenAI regarding the expansion of the Stargate data center continued for several months, but the companies ultimately decided not to acquire the additional capacity. An OpenAI executive explained in a social media post that they "chose to redirect resources to other locations."
It has also been reported that Meta is involved in the discussions, leaving the future of the unexpanded capacity somewhat fluid.
1-2. The Complexity of Building AI Data Centers
This case illustrates how complex large-scale AI infrastructure construction projects are. Securing power, acquiring and preparing land, coordinating with design firms, and structuring financial schemes must all be done in parallel, and a shift in any one element can affect the entire plan.
Bloomberg Intelligence analyst Matthew Schettenheim stated, "With $650 billion said to be poured into these massive AI projects, I had never heard of such a plan change occurring with big names like Oracle and OpenAI," suggesting that this case is highly unusual.
2. Spillover to the Private Credit Market—Hidden Risks
2-1. The 'Shadow Funding Source' Supporting AI Infrastructure
It is not just bank loans that have supported the construction of large-scale AI data centers like Stargate. The private credit market, a non-public lending market, has provided a significant portion of that funding. Major private credit funds, including Blue Owl, have been aggressively expanding their lending for AI infrastructure construction.
2-2. What the Plan Change Questions
When a plan change like this occurs, how does the valuation of the project—which served as collateral for the loan—change? This is a question that private credit investors cannot ignore. Even if the project itself continues, if it is "used in a way different from what was intended" or if the "initially anticipated customer changes," it could affect the calculation of collateral value.
Grenadilla Advisory analyst Anna Rathbun pointed out, "This is the kind of story that makes private credit investors a little nervous. If something doesn't go as intended, what happens to the collateral, and will the project continue?" She also noted that this is evidence that we are still in the very early stages of AI development, which is why spending is so high, and that it makes sense for companies to change their minds, suggesting that these moves should not be viewed with excessive pessimism.
3. The Relationship Between AI Companies and the Government—New Tensions Revealed by Anthropic's Lawsuit
3-1. An Unusual Designation as a 'Supply Chain Risk'
In the same broadcast, another piece of important news was reported: that Anthropic has filed a lawsuit against the U.S. Department of Defense.
The Department of Defense designated Anthropic as a "supply chain risk." This designation is typically used for foreign adversaries, and there is almost no precedent for it being applied to a private AI company within the United States. Anthropic has indicated its intention to fight this decision in court, calling it "unprecedented and illegal."
3-2. The Background of the Negotiations
According to Matt Schettenheim of Bloomberg Intelligence, the direct sequence of events leading to this designation was the Department of Defense's attempt to retroactively change the terms of a contract that Anthropic and the Department of Defense had agreed upon in July. When Anthropic resisted this, it was allegedly suddenly designated as a 'supply chain risk'.
On the other hand, it is reported that OpenAI is continuing dialogue toward concluding an agreement with the Department of Defense by negotiating similar concerns in a more moderate manner than Anthropic.
3-3. Financial Impact and the 'Blast Radius' Problem
From a purely financial perspective, the contract between Anthropic and the Department of Defense that became the issue this time is said to be worth up to $200 million, but the actual amount executed is believed to have remained at only a few million dollars.
More important is the 'blast radius' of this designation. Even if transactions are not legally prohibited, other client companies that have dealings with the government might decide to 'pass on contracts with Anthropic to avoid pressure from the administration.' The magnitude of this indirect impact is difficult to quantify at this time and remains an uncertainty.
Summary: A Perspective for Re-evaluating AI Infrastructure Investment Risks
The cancellation of data center expansions by Oracle and OpenAI and the Anthropic lawsuit against the Department of Defense are both events that bring instability to the narrative of the 'smooth sailing of the AI boom'.
On the other hand, as local analysts point out, the view that 'we are still in a very early stage' is on target. In the early stages of a major technological revolution, trial and error in capital allocation and changes in direction are inevitable, and that in itself does not immediately mean the 'end of the boom'.
What is important for investors is not to overlook such plan changes and legal/political risks as 'transient noise,' but to maintain a perspective that continuously monitors whether any changes are occurring in the risk profiles of specific companies or projects.
