Oracle has sent a force majeure notice to the developer of Project Jupiter, a Stargate data-centre campus in New Mexico, Bloomberg reported on 24 September. Bloomberg's sources, as relayed by TechCrunch, say Oracle is not trying to exit as the campus's main tenant. The notice would instead let it delay payments if the facility misses its 2028 target to come online.
We have tracked the data-centre build-out for four weeks running, from Crusoe's $3.9bn raise to Google's Project Tembo campus. This is the first big AI campus we have seen hit a contractual wall in public. The notice reserves Oracle's rights, and nobody has declared a delay. The causes behind it will still be familiar to any project manager, quantity surveyor or contractor: a late utility connection, refused permits and a regulatory decision still pending.
The AI build-out is usually described as a race for chips. On this site the critical path runs through a gas pipeline and an air-quality permit, and that is where data-centre bidders should look.
What has slipped
Project Jupiter is designed for 2.45GW and is meant to run on gas-powered fuel cells from Bloom Energy, so a reliable gas supply is central to its timeline. The Energy Transfer pipeline that would deliver the gas has been delayed nearly six months, to 1 February 2027, after regulators repeatedly denied permits for the line.
Source New Mexico reported in August that the state's Commissioner of Public Lands, Stephanie Garcia Richard, had turned down the developer's second request to build on state trust land. Bloomberg reported that the route was then changed.
The fuel cells also need an air-quality permit, and the New Mexico Environment Department faces a 23 November deadline to decide. The state Supreme Court paused the permit hearing in August while it considered challenges from environmental groups, then rejected them on 17 September.
At that point, Source New Mexico reported, the department had yet to replace the hearing officer who had recused himself or set a new hearing date. Lawyers for Yucca Growth Infrastructure, one of the developers, had argued in a court filing that keeping the hearing paused could make the deadline "impossible to meet".
What Oracle and Blue Owl say
Both companies say the schedule and the money are intact. Oracle maintains that it does not expect a delay, and told CNBC: "Project Jupiter remains on our planned schedule. We are fully committed to New Mexico and confident in our path forward."
We see this as a tenant protecting itself against a risk it does not expect, but cannot control.
This notice does not change the financial commitments to this multi-year project.
For firms further down the supply chain, that is the sentence to test. If a tenant's payments to a developer can be deferred, the next question is whether anything in the chain of contracts lets that deferral travel downwards.
Questions to take to your commercial team
TechCrunch describes force majeure clauses as ones that excuse a party from contractual obligations when events outside its control get in the way. Whether a late pipeline or refused permit counts depends on each contract's wording and governing law. For commercial and legal teams, these are the questions we would raise before pricing or signing data-centre work. This is not legal advice.
Does your appointment or subcontract define force majeure, and does that definition reach delays to utilities or permits that the client is responsible for securing?
Is late power or gas supply to the site listed as a relief event, a compensation event or a client-side risk, or is the contract silent on it?
Does any wording link your payment to the client's own income, so that a deferral further up the chain could reach you?
Does your programme show utility connections and permit decisions as client-side milestones with dates, so that a slip is visible and can be notified?
Are notice periods for extension-of-time claims diarised, given that many contracts penalise late notice?
The money behind the concrete
The notice landed a day after Brookings published a summary of a paper by Stijn Van Nieuwerburgh, Earl Kazis and Benjamin Schore Professor of Real Estate at Columbia University. It projects AI investment in data-centre buildings, power systems, networking infrastructure, and specialised chips and other equipment totalling $10.3tn from 2025 to 2032, an average of 3.63% of US GDP a year. It warns that risk is moving from corporate balance sheets to off-balance-sheet structures such as joint ventures, private credit and lease commitments.
Those structures, Van Nieuwerburgh writes, depend on cash flows exposed to risks that include "timely access to power and hardware". Project Jupiter shows the power item in practice.
It would be premature to conclude that AI infrastructure already poses systemic risk comparable to earlier credit booms.
We take the same view. One notice on one site does not tell us much about the wider build-out, but it does show where some of the pressure points are.
Takeaway
Oracle and Blue Owl both say Project Jupiter is on schedule and the commitments stand. The dates to watch are the environment department's 23 November air-permit deadline and the pipeline's 1 February 2027 in-service date.
For UK firms pricing or delivering data-centre work, the practical point is about their own contracts. Energy supply and permits belong on the programme as named dependencies with owners and dates, and the contract should say plainly who carries the time and cost if they slip. That is far easier to settle at tender than after a notice arrives.
We will report what the 23 November permit decision means for Project Jupiter's programme. If you price, programme or build data-centre work, follow the story in the Project Flux newsletter.
Links and Stuff
All content reflects our personal views and is not intended as professional advice or to represent any organisation.
