Elon Musk’s latest comments have thrown one of SpaceX’s most eye‑catching AI deals into doubt, raising questions about how long Anthropic can count on the company’s supercomputer resources.

Earlier this month, SpaceX disclosed in its IPO S-1 that Anthropic had agreed to a massive cloud services arrangement, paying about $1.25 billion a month for access to SpaceX’s Colossus compute cluster “through May 2029,” with capacity ramping from May 2026 and either side allowed to terminate on 90 days’ notice.¹ The filing language — repeated across several pages — clearly framed the pact as a multi‑year agreement and a marquee revenue line for investors.²

On May 28, tech media highlighted the deal as a three-year lease, describing how it would give Anthropic exclusive use of Colossus and billions in recurring payments to SpaceX.² Financial press coverage likewise characterized Musk’s arrangement as a long-term data center partnership underpinning the IPO narrative.¹

That same morning, however, Musk publicly reframed the commitment. Responding on X, he said SpaceX has “not committed to leasing Colossus for years,” calling it “a 180 day lease with 90 day notice mutual cancellation thereafter,” and stressing that the short term was SpaceX’s request because “if compute gets super tight I said we might need it back at some point.”² Business press summarized his position as SpaceX having “only committed to a 180-day compute agreement with Anthropic,” not a firm multi‑year lock‑in.³

The result is a stark discrepancy: public filings portray a multi‑year revenue stream based on Anthropic paying monthly fees through 2029, while Musk emphasizes the deal’s cancellability and short initial term.² Analysts note this tension could matter for IPO investors trying to judge the durability of SpaceX’s AI infrastructure business and for Anthropic’s own planning, even as neither company has offered further clarification.