Investors Ask Data Centers to Show Their Work

Reuters reports that data-center companies considering public listings face more demanding investors, while Futurism reports trouble around an $18 billion New Mexico facility associated with Project Jupiter and the wider Stargate AI buildout. According to Futurism’s account, Oracle sent lender Blue Owl Capital a force-majeure notice after the project failed to secure certain development permits, raising questions about payments and a planned 2028 completion.

The facility was planned around as much as 2.45 gigawatts of computing capacity, with Oracle expected to serve as its primary tenant. A proposed natural-gas plant was abandoned after land-use permits were not secured, and Oracle reportedly turned toward solid oxide fuel cells at a scale far beyond established deployments. Blue Owl supplies financing, while Oracle, OpenAI and SoftBank hold different roles in the broader Stargate effort. The available accounts do not provide a complete set of on-record responses from every lender, developer and permitting authority, so redesign and delay should not be treated as confirmed cancellation.

The benign interpretation

Investor caution could be routine discipline arriving late to a capital-heavy industry. Data centers require land, substations, cooling, equipment and construction long before they produce revenue. Public-market investors are entitled to demand contracted income and credible delivery dates rather than reward every issuer for standing near the phrase artificial intelligence.

On that view, a smaller project is not necessarily a failed one. Developers routinely phase campuses as permits, customers and power become available. Reducing an initial build can protect capital, match capacity to actual demand and prevent an ambitious plan from becoming an expensive empty shell.

The harder reading

The opposing case is that physical constraints are exposing assumptions that presentations kept conveniently distant. A tenant may want immense computing capacity without accepting every financing risk. A lender may fund construction only if completion and occupancy remain plausible. A grid connection, permit or fuel-cell plan cannot be replaced by enthusiasm. Yesterday’s campus was rendered at heroic scale. Today’s prospectus needs occupancy rates.

One New Mexico project does not establish an industry collapse, just as a crowded IPO pipeline does not establish durable demand. The useful comparisons are less cinematic: signed lease terms, customer concentration, available power, construction progress, completion guarantees, debt maturities and the amount of refinancing required before revenue begins. Each measure tests a different point where promised scale can detach from a bankable facility.

The boom remains credible if customers commit for long periods, power solutions survive permitting and engineering review, and facilities enter service close enough to schedule to support their debt. If projects repeatedly shrink while issuers still seek valuations based on their original capacity, investors will have their answer: the renderings expanded faster than the business.