Elon Musk posted on X that SpaceX’s lease of Colossus compute to Anthropic is a 180-day deal with a 90-day mutual cancellation clause, not the multi-year commitment implied by SpaceX’s own SEC filing. The discrepancy between Musk’s public statements and the S-1 language raises questions about the stability of Anthropic’s compute infrastructure at a moment when both companies are preparing for IPOs.
What Musk Said vs. What the S-1 Says
In a post on X, Musk responded to a user who suggested the SpaceX-Anthropic agreement was a three-year deal: “SpaceX has not committed to leasing Colossus for years, although it’s possible that may be what happens. This is a 180 day lease with 90 day notice mutual cancellation thereafter,” Gizmodo reported.
Musk added: “The short term was our request, not Anthropic’s.”
SpaceX’s S-1 filing with the SEC describes the arrangement differently. The document states Anthropic “has agreed to pay us $1.25 billion per month through May 2029, with capacity ramping in May and June 2026 at a reduced fee.” The filing does reference a mutual 90-day termination clause, but the overall language frames a three-year commitment, per Gizmodo.
Musk’s Stated Reasoning
Musk signaled he may want the compute back. “We won’t leave them hanging and will provide a reasonable off-ramp, but if compute gets super tight I said we might need it back at some point,” he wrote on X, according to Gizmodo.
One interpretation: the short lease serves as a proof of concept for investors ahead of SpaceX’s IPO, demonstrating that leasing surplus compute is a viable revenue line. SpaceX is targeting a $2 trillion valuation and aims to raise over $75 billion in what would be the largest IPO in history. Showing demand for Colossus compute at $1.25 billion per month strengthens that narrative.
Anthropic’s Compute Portfolio
The timing is notable. Anthropic closed a $65 billion Series H round on the same day, pushing its valuation to $965 billion. The company’s announcement listed its compute agreements: up to five gigawatts with Amazon, five gigawatts of next-generation TPU capacity with Google and Broadcom, and GPU access in SpaceX’s Colossus 1 and Colossus 2. AWS remains Anthropic’s primary cloud provider and training partner.
The SpaceX compute represents one piece of a multi-provider strategy, but the scale ($1.25 billion per month) makes it a significant one. If the lease terminates after 180 days with 90 days’ notice, Anthropic could face a 270-day countdown to relocate petaflop-scale workloads.
What Changes for Agent Infrastructure
Anthropic’s $47 billion revenue run rate depends on Claude availability. Claude Code drives the majority of enterprise adoption. Dynamic Workflows, the subagent orchestration feature released with Opus 4.8 today, requires sustained training compute to iterate on. Any disruption to Anthropic’s compute access cascades through the agent toolchain: Claude Code, Claude Cowork, Managed Agents, and every third-party application built on the Claude API.
Gizmodo reached out to Anthropic for clarification. Neither company has reconciled the discrepancy between Musk’s public statements and the S-1 language.