SpaceX filed an S-1 document with the SEC that describes Anthropic payments running through May 2029. Elon Musk has been publicly characterizing the same deal as short-term and cancellable. Both versions can't be correct.
The arrangement gives Anthropic access to SpaceX's computing infrastructure - the GPU clusters needed to train and serve the Claude family of models. Compute access is the most constrained resource in AI right now, and three-year commitments reflect the reality that you can't scale a frontier model company on month-to-month contracts. A deal running to May 2029 is a planned, long-term infrastructure relationship.
Musk's interest in reframing the deal's length is easy to read. He controls both SpaceX and xAI, his AI company that builds the Grok models. If SpaceX's Colossus data center in Memphis is committed to Anthropic through 2029, that capacity isn't readily available for xAI's own training runs. Describing the deal as easily terminated reduces the apparent conflict of interest and keeps his options open.
The problem is the SEC filing. S-1 documents are legal disclosures for investors, meant to accurately describe material business arrangements. TechCrunch's reporting notes the filing explicitly describes payments through May 2029. If the deal is genuinely short-term, that's a disclosure question. If the filing is accurate, Musk's public statements are misleading.
For Anthropic, the immediate concern is reliability. The company is deep into scaling its model training and can't easily swap infrastructure providers mid-cycle. Any real uncertainty about whether this contract will be honored - or quietly renegotiated under pressure from its counterparty's controlling shareholder - creates operational risk for a company competing directly against OpenAI and Google.