Summary

  • Project Jupiter’s lease, per the Wall Street Journal’s September 25, 2026 reporting, obligates Oracle to make rent payments under “hell-or-high-water” terms regardless of whether the planned New Mexico data center is operable, and the September 2026 force majeure notice extends that obligation across up to three years of permitting and power delay.
  • The financing structure — approximately $18 billion in bank construction loans, roughly $3 billion in Blue Owl Capital equity, and a built-in three-year delay buffer — concentrates value on the lenders and developer during routine operation, but converts permitting and power disorder into a fixed obligation when those contingencies fail.
  • Credit-market signals — at least one bank selling New Mexico loan paper below 90 cents on the dollar, S&P Global Ratings’ July 2026 downgrade of Oracle’s credit rating to one notch above “junk” territory, and Oracle’s AI-related capital expenditure exceeding operating cash flow for several quarters — indicate that the market is repricing the lease structure’s tail risk before the operational question is resolved.
  • Nassim Taleb’s framework in Antifragile classifies arrangements in which rare-event losses amplify while ordinary operations appear stable as technically fragile; the lease configuration the Wall Street Journal reports fits that profile, with the rent response curve bending downward as permitting and power stress accumulate.

The Wall Street Journal reported on September 25, 2026, that Oracle issued a force majeure notice to Stack Infrastructure, the developer of its planned New Mexico AI data-center campus known as Project Jupiter. The notice cites three contingencies — power supply, permitting, and local resistance in Doña Ana County — and allows Oracle to defer full rent payments for up to three years without reducing its overall financial exposure, according to people familiar with the deal. The lease, per people familiar with the matter, carries “hell-or-high-water” terms that bind Oracle to rent payments regardless of whether the data center is operable. The analytical claim the disclosure supports is that the structure, by decoupling the rent obligation from the project’s underlying viability, converts permitting and power delay from a planning constraint into a fixed corporate liability that intensifies as disorder accumulates.

The lease architecture behind the notice

The force majeure notice reported by the Wall Street Journal is the visible event; the lease that produced it is the underlying structure. Oracle signed the New Mexico lease, people familiar with the matter told the Journal, with so-called “hell-or-high-water” terms — a contractual formulation in which the tenant’s obligation to pay does not depend on whether the premises are fit for use, and in which the lease cannot be terminated. Pulling the force majeure lever, per the same people, triggers an automatic extension of the lease’s duration, so the ultimate amount Oracle pays Stack does not decline; it shifts to a later payment profile. The force majeure notice allows Oracle to defer full rent payments for up to three years, but it does not reduce Oracle’s overall financial exposure to the development.

An Oracle spokesman said in an email: “Project Jupiter remains on our planned schedule. We are fully committed to New Mexico and confident in our path forward.” Oracle also said, according to the Journal, that force majeure notices are commonplace in projects like the one in New Mexico and do not necessarily mean there is a delay or change in expectations. A Blue Owl spokesman said in an email: “Blue Owl, Stack Infrastructure and Oracle remain fully aligned on Project Jupiter. This notice does not change the financial commitments to this multiyear project.”

The corporate posture, taken at face value, is one in which the notice is a contractual mechanism absorbing a routine delay. The substrate does not establish whether Oracle has any mechanism to renegotiate the underlying “hell-or-high-water” obligation itself; it establishes only that the force majeure notice allows payment deferral for up to three years. The substrate reports the clause exists and that force-majeure notices “are commonplace in projects like the one in New Mexico,” per the Oracle statement; it does not specify whether the deferral extends, modifies, or merely postpones the obligation curve.

The payoff structure between Oracle, Stack, and the lenders

The financing of Project Jupiter, as described by the Wall Street Journal, places roughly 20 banks in the role of construction lenders with $18 billion in committed financing and Blue Owl Capital funds in the role of equity provider with approximately $3 billion. Stack Infrastructure, the project’s developer, is itself owned by Blue Owl Capital funds. Stack is currently earning the equivalent of a 9 percent annualized return for Blue Owl from money raised to build the project, the Journal reported, citing people familiar with the matter; the developer’s forecast return jumps to 11 percent once lease payments begin.

The financing package set aside funds for up to three years of potential delays before lease payments begin. That three-year buffer is the structure that allows the force majeure notice to operate without triggering an immediate renegotiation; it is also the structure that, once consumed, brings the next phase of the contract into view. Across the range from routine construction delays to the tail event of a denied air permit or a permanently blocked pipeline, the configuration concentrates value on the lender-developer side in ordinary scenarios, and converts disorder into a fixed obligation that runs in the opposite direction from the project’s fortunes.

Taleb’s framework in Antifragile classifies arrangements in which losses amplify with rare events while ordinary operations look stable as concave in response profile — fragile in the technical sense. By the description in the Wall Street Journal, Project Jupiter’s lease structure appears to fit that profile: across the plausible range of stress levels, the obligation curve bends downward. Only the revenue side of the chain — Oracle’s ability to bill OpenAI under the underlying cloud-services contract — depends on operability; the rent obligation does not.

The three load-bearing components

The setbacks the Wall Street Journal identifies are concentrated in three components: an air permit, a natural-gas pipeline right-of-way, and the partner interface.

On the air permit, Oracle has yet to obtain the air permits it needs from the state of New Mexico. The Journal reports that the company overhauled its original power plans to address local resistance to using gas turbines and diesel generators; earlier in 2026, Oracle and BorderPlex announced a switch to Bloom Energy fuel cells. The substrate does not establish whether the fuel-cell substitution resolves the air-permitting question or merely relocates it.

On the pipeline, a small portion of a proposed 17-mile natural-gas pipeline would pass through state land and has been rejected twice, according to the Journal. Whether the Bloom Energy transition makes the natural-gas pipeline dispensable depends on the fuel cells’ gas-input requirement, which the available reporting does not specify.

On the partner interface, the Journal reports that Oracle has removed BorderPlex Digital Assets, a key land and power partner that started working on the project in late 2024 and had early support from Gov. Michelle Lujan Grisham. BorderPlex is set to receive a payment from Oracle to relinquish the rights it held to the project’s land and power, per people familiar with the matter. The substrate does not detail what rights BorderPlex held, what was lost in the transfer, or what the relinquishment cost the New Mexico development beyond the disclosed payment.

The credit-market signal

The substrate reports two credit-market reactions to Project Jupiter that run in the opposite direction from the corporate posture.

In July 2026, S&P Global Ratings downgraded Oracle’s credit rating to one notch above “junk” territory, saying its “rapidly expanding AI infrastructure business is increasing its overall credit risk.” Oracle’s capital expenditure on AI equipment and property, the Journal reports, has exceeded its cash flows from operations for several quarters. Bond investors have raised questions about Oracle’s financial stability and its investment-grade credit rating.

In the New Mexico loan market specifically, at least one bank that arranged the New Mexico loans sold off a portion of the debt for a price below 90 cents on the dollar, implying a paper loss of at least $1.8 billion for original lenders, according to people familiar with the matter. The Financial Times earlier reported the bank debt sale. Most banks hold construction loans they make until their due dates and do not adjust their valuations based on market fluctuations — the disclosed transaction is one bank’s price, not the system’s price. But the Journal previously reported that earlier in 2026, banks also struggled to sell billions of dollars of loans they made to build data centers leased to Oracle in Texas and Wisconsin.

The pattern is the analytical point. Philip Anderson’s diagnostic in the complexity literature, applied to the substrate’s account, identifies a system-level pattern not visible from any single component and not predictable from any one in isolation as the signature of emergence. The cross-site correlation — bank exposure to Oracle-leased AI infrastructure becoming correlated across Texas, Wisconsin, and New Mexico — meets that description. Diane Vaughan, studying the Challenger accident, identified the practice of treating past success as evidence of safety even as underlying tail risk has shifted as normalization of deviance. The S&P downgrade and the secondary-market loan prices are the substrate’s documentation of normalization-of-deviance’s inverse: tail-risk repricing by participants who are not bound by the hold-to-maturity convention that insulates most construction lenders’ marked values.

Forward trajectories

Two failure narratives are visible from the substrate’s reported state; the substrate does not establish which is more likely. Both trace back to the same root: the lease structure that decouples the obligation from the project’s underlying viability.

Under the first narrative, the permitting and pipeline pathway remains stuck past the three-year delay buffer. The lease, by its reported terms, continues. The financing, by its structure, requires refinancing or restructuring, since the buffer the package was built to absorb has been consumed. The 90-cent signal in the secondary market becomes the reset price for the next lending round rather than an isolated sale; the S&P one-notch-above-junk rating becomes the operating constraint on the next issuance rather than a forward-looking warning.

Under the second narrative, the OpenAI side of the contract encounters its own stress. The Wall Street Journal reports that Project Jupiter is part of Oracle’s $300 billion, five-year cloud-computing agreement with OpenAI, signed roughly a year prior; Oracle signed leases to occupy roughly 4.5 gigawatts of capacity across Texas, Wisconsin, Michigan, and New Mexico to deliver against that contract. If OpenAI’s demand profile weakens or its own capacity plans shift, Oracle’s “hell-or-high-water” obligation — designed as a commitment device for the OpenAI contract — converts from a strategic asset into a stranded-capacity liability, and the rent obligation continues to bind against a revenue stream that no longer justifies it.

Donella Meadows and Peter Senge’s systems-dynamics tradition identifies the “Limits to Growth” archetype as the structure in which visible infrastructure stock grows faster than the supporting stocks that constrain it. The substrate’s reported configuration — Oracle growing roughly 4.5 gigawatts of computing capacity across four states while the supporting stocks of power generation, transmission upgrades, permitting throughput, and the social license to operate in affected counties lag — fits that archetype. A reinforcing loop runs from AI demand to capacity build to permitting-and-grid stress to construction delays to reputational and credit costs to higher cost of capital on the next phase. A balancing loop — the three-year financing reserve — provides runway, but the runway is consumed by use and is not replenished. James Reason’s Swiss-cheese model, applied to the substrate’s account, treats each defensive layer — the delay buffer, the equity cushion, the hold-to-maturity convention — as a slice with holes, and asks whether the holes align.

How the disclosure is being framed

The disagreement between the corporate statements and the market signals is unresolved on the substrate as it stands. Oracle and Blue Owl spokesmen frame the notice as routine and the financial commitments as unchanged; the S&P action and the secondary-market debt prices frame the same event as evidence of credit deterioration. The substrate does not resolve which reading is correct. It does establish that two audiences — the issuer side, with its contractual language of commonplaceness, and the credit side, with its 90-cent price and one-notch-above-junk rating — are looking at the same structure and assigning it materially different values.

Applying Taleb’s via-negativa discipline — remove fragility rather than add strength — the highest-leverage target is the lease structure itself. The “hell-or-high-water” clause is the single component whose renegotiation would most shift the response profile from concave toward linear. The substrate reports the clause exists; it does not establish whether Oracle has any mechanism to renegotiate the underlying obligation, only that the force-majeure mechanism allows a payment deferral of up to three years.

Open indicators

Several indicators would distinguish whether the three-year deferral resolves within the buffer the financing was built to absorb. The disclosed load-bearing components — air permit issuance, natural-gas pipeline right-of-way resolution, completion of the first phase originally scheduled for the third quarter of 2028 — are not yet positive. Whether the Bloom Energy fuel-cell configuration requires a different permitting pathway than the prior gas-turbine and diesel-generator plan, and whether that pathway is shorter, is a question the available reporting does not resolve. Whether the 90-cent bank debt sale is a one-off price or a reset price for the next lending round is not established by the disclosed transaction. Oracle’s stock fell as much as 8 percent in morning trading on the disclosure date before closing down 4 percent; Blue Owl stock declined about 3.6 percent.

Additional considerations

The substrate’s account contains one boundary decision worth flagging separately. Power generation and permitting have been treated, in the conventional framing of AI build-outs, as exogenous constraints on a planning problem. Bank credit availability has been treated as exogenous to the data-center story. The substrate’s report — bank debt selling below 90 cents on the New Mexico loans, earlier reports of Texas and Wisconsin loan-sale difficulty, and S&P’s downgrade citing “rapidly expanding AI infrastructure business” that is “increasing its overall credit risk” — suggests the boundary may need to be redrawn, with capital-market stress as a feedback into the build-out pace rather than a backdrop to it. The substrate does not state this conclusion; it supplies the components from which it follows.

Analytical techniques used in this piece

This analysis applies the methods below. Each links to a short, plain-English explainer you can read and reuse.

Fragility / Antifragility Audit
Asks whether a system gains or loses from volatility, shocks, and disorder (Taleb).
Pre-Mortem (Fragility)
Imagines a system has already broken and traces the structural fragilities that let it.
Systems Dynamics (Structural)
Maps a system’s structure — stocks, flows, and the architecture that shapes its behavior.