Oracle Project Jupiter Force Majeure Notice Stalls $18B Data Center Loan, September 2026

Oracle has invoked force majeure on its New Mexico Stargate campus, sending the project's $18 billion construction loan to 89–91 cents on the dollar and exposing grid power as the binding constraint on the AI infrastructure buildout.

Oracle's invocation of force majeure on Project Jupiter — its flagship AI campus in New Mexico and a central node in the Stargate consortium with OpenAI — has thrown a spotlight on what is becoming the defining bottleneck of the AI infrastructure cycle: grid power. The notice, sent to Blue Owl Capital as a key lender in the project's financing structure, cited unforeseeable circumstances affecting power supply. Reports indicate that campus completion has been pushed to 2028, and Oracle shares fell 3.5% following disclosure of the delay concerns.

Oracle's roughly $18 billion in Project Jupiter construction debt was reported trading at 89 to 91 cents on the dollar, with syndication described as stalled — a direct repricing of execution risk in a sector where institutional capital has accumulated on the assumption that infrastructure delivery timelines would prove manageable. Sending a force majeure notice to a project lender signals that Oracle is managing liability exposure stemming from circumstances outside its control. But as one report noted, the effect on broader AI data center financing sentiment is measurable regardless of the legal mechanics.

What gives the situation its particular texture is what Oracle has been doing in parallel. Rather than retreating from its New Mexico energy strategy, Oracle reaffirmed its 2.4 gigawatt fuel cell supply agreement with Bloom Energy — a confirmation that sent Bloom Energy shares up 8%. The rationale is visible in the operating record: Bloom delivered fuel cell power to a separate Oracle data center in 55 days, faster than traditional grid upgrades. Oracle has also issued a 2 gigawatt renewable energy request for proposals in New Mexico and signed 1.7 gigawatt wind supply agreements with RWE and Engie in Texas. Taken together, these moves reflect a deliberate posture that on-site and near-site generation must substitute for grid reliability — a posture the force majeure notice confirms was warranted.

This episode plays out against a financial backdrop that otherwise holds up. Oracle reported Q1 fiscal 2027 revenue of $19.3 billion, with management characterizing the AI buildout as self-funded from operating cash flow. The company's AI cloud backlog stands at a reported $664 billion, and Oracle delivered 300,000 GPUs in the first fiscal quarter. The capital commitment behind these figures is substantial: Oracle's fiscal 2026 capital expenditure reached $55.66 billion, up 162.4% from $21.21 billion in fiscal 2025, representing 83% of the year's $67.36 billion in revenue — the highest capex intensity among tracked hyperscalers in Slicast's SEC XBRL compilation. At that pace of outlay, construction delays translate directly into a lag between capital deployed and revenue recognized, compressing the return timeline the market has been pricing. Oracle has also expanded its AI transition restructuring budget by $700 million. On 14 September 2026, an SEC filing disclosed that Larry Ellison had cancelled his 10b5-1 stock sale plan, with Oracle stating no shares were sold and that he has no current plans to sell.

The broader implication reaches beyond Oracle's balance sheet. Project Jupiter sits at the center of Stargate, the OpenAI-Oracle-SoftBank compute initiative. One analyst estimate places OpenAI's compute expenditure at $278 billion by 2030. When a campus at this scale encounters a power constraint serious enough to trigger a lender force majeure notification, it provides an empirical data point the sector has largely lacked: AI data center construction is harder and slower to execute than the capital formation around it has assumed. The issue is not that the demand story has weakened — Oracle's backlog figures suggest it has not — but that the physical delivery side of the equation has proven more complex than financial modeling anticipated.

Three signals are worth tracking. First, whether Oracle's fuel cell and renewable pipeline resolves the New Mexico power constraint on a timeline that preserves the 2028 completion target's credibility. Second, whether the Project Jupiter construction loan returns to par in secondary trading — recovery to par would indicate lenders view the delay as temporary rather than structural. Third, how Oracle's Q2 fiscal 2027 results characterize backlog conversion pace: at 83% capex intensity, execution speed has become the variable on which the investment thesis pivots.