Oracle invokes force majeure on a 2.45GW AI campus
Oracle sent a force majeure notice over Project Jupiter, the $165B Stargate site in New Mexico. Your 2028 AI compute depends on a gas pipeline permit.
Oracle sent a force majeure notice to the developer of Project Jupiter, the Stargate data center campus in Doña Ana County, New Mexico. Bloomberg broke it on September 24; TechCrunch confirmed the details. The cause is not a chip shortage or a funding gap. It is a natural gas pipeline that regulators keep refusing to permit. If you are building on AI infrastructure that comes online in 2028, this is the failure mode nobody put in the roadmap deck.
What actually happened
Project Jupiter is roughly 1,400 acres near Santa Teresa, with a build-out cost that could reach $165 billion. It is one of the flagship sites in the Stargate program that OpenAI, Oracle and SoftBank announced together. The campus is designed to run on as much as 2.45 gigawatts of Bloom Energy fuel cells, which burn natural gas.
The gas has to arrive. An Energy Transfer pipeline meant to feed the site slipped nearly six months, to a February 1, 2027 target, after regulators repeatedly denied permits for the line. A separate air-quality permit for the fuel cell system is still pending with the state, which faces a November 23 deadline. Local residents and environmental groups have pushed back on water use and air quality since the project was announced.
The notice is a contractual hedge, not an exit. Oracle stays on as the main tenant; the filing lets it delay payments to the Blue Owl Capital unit developing the site if the facility misses its 2028 target to come online. Oracle's own line is that the project "remains on our planned schedule." Oracle shares fell on the news anyway, which tells you how the market read a company protecting itself from its own timeline.
Why a delayed AI data center matters for your business
You will never sign a contract with Project Jupiter. You will feel it through the token price and the rate limit.
The industry has been selling 2028 capacity as though it were already racked. Oracle just put in writing that a marquee site's schedule depends on a pipeline permit and a state air-quality decision — regulatory events with no engineering workaround and no vendor you can escalate to. When a campus this size slips, the demand does not evaporate. It queues somewhere else, and small buyers sit at the back of that queue.
The lesson is not "avoid Oracle." It is that AI capacity is now a construction and permitting business, and construction slips. Treat any vendor roadmap dated 2027 or later as a forecast, not a commitment. If your plan to triple inference volume next year rests on on-demand pricing holding steady, go get that in writing from your provider before you build against it.
We keep the model call behind one interface with a fallback chain, so switching providers is a config change and a deploy. That is not architectural purity. It is what lets a business keep shipping when somebody else's gas pipeline gets denied a permit for the third time.
Key takeaways
- Oracle filed a force majeure notice over Project Jupiter, the ~1,400-acre Stargate campus near Santa Teresa, New Mexico
- The trigger is infrastructure, not silicon: an Energy Transfer pipeline slipped ~6 months to February 1, 2027 after repeated permit denials
- A state air-quality permit for the 2.45GW Bloom Energy fuel cell system is still pending, with a November 23 deadline
- Oracle remains the main tenant — the notice only lets it delay payments if the site misses its 2028 online target
- Read any 2027+ capacity promise as a forecast; permitting risk has no engineering workaround
- Put your provider behind one adapter with a fallback chain so a capacity slip costs you a deploy, not a rewrite
Build for the vendor you'll have in 2028, not the one you signed. We architect AI features so the provider is a swappable module with a fallback chain — not a dependency welded into your business logic. See how we build vendor-agnostic systems you own, or bring us your AI roadmap and we'll pressure-test the assumptions.
Sources: TechCrunch: Oracle sends force majeure notice on its New Mexico Stargate data center, Yahoo Finance: Oracle shares fall as company invokes force majeure.
- #oracle
- #ai-infrastructure
- #vendor-risk
- #stargate
- #compute
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
YouTube Custom Feeds: viewers now write the algorithm
YouTube's Custom Feeds let viewers describe what they want in plain language and Gemini builds the feed. Discovery just became a sentence, not a keyword.
Read itSora's API shuts down today. Vendors delete features.
OpenAI removes the Sora video API on September 24 with no replacement listed. Six months' notice, no successor — here's how to build so a deprecation is a config change.
Read it