Anthropic's SpaceX Lease Duration Divides Opinions

Earlier this month, xAI finalized a significant compute arrangement with Anthropic, committing billions per month for exclusive access to the Colossus cluster. The deal proved advantageous for both sides: xAI gained essential revenue, while Anthropic moved forward in the ongoing competition for computing resources.
However, this morning on X, Elon Musk played down the extent of SpaceX's commitment to the agreement.
“SpaceX has not committed to a multi-year lease of Colossus, though that could still happen,” he responded to a user. “It’s a 180-day lease with a mutual 90-day cancellation clause after that. We requested the short term, not Anthropic. We won’t abandon them and will offer a fair exit, but if computing capacity becomes extremely scarce, I mentioned we might need to reclaim it at some point.”
Musk's remarks directly conflict with SpaceX's recent S-1 filing, which acknowledges the standard 90-day cancellation clause yet describes the deal as a three-year commitment. Page F-62 of the filing states:
On May 3, 2026, the Company signed a cloud services agreement with Anthropic PBC, an AI research and development public benefit corporation, for access to compute capacity. Under this agreement, the customer agreed to pay a monthly fee through May 2029, with capacity gradually increasing starting May 2026 at a reduced rate. Either party may terminate the agreement with 90 days' notice. The customer retains ownership and intellectual property rights over its content, AI models, and related data.
The crucial detail is that Anthropic "has agreed to pay a monthly fee through May 2029" – a clear indication of a three-year lease. This phrasing appears again on page F-96 and in a slightly different version ("the customer has agreed to pay us $1.25 billion per month through May 2029") on pages 13 and 146, ruling out a simple typo.
xAI did not reply to a request for clarification.
One could argue over whether Anthropic's commitment to pay for a service equates to SpaceX's commitment to provide it, but that is not typically how a lease works. Moreover, why implement a one-sided lock-in if either side can cancel the deal with just three months' notice?
I do not have the actual contract, so I cannot confirm its terms – and neither SpaceX nor Anthropic has disclosed the duration in their public statements. Nevertheless, the facts should be clear, and making false claims during a company's quiet period is not advisable.
As usual, the SEC is unlikely to take action – and even if it did, Elon would probably not be concerned. However, this appears to be a material misrepresentation made while marketing a security, which at the very least reflects poorly.
Sean O'Kane contributed reporting to this article.
Related article
Anthropic launches Opus 4.8 featuring new dynamic workflow tool
Anthropic unveiled Opus 4.8 on Thursday, marking the latest iteration of its premier public model. Priced identically to its predecessor, this update is now accessible across all platforms.Releasing just 41 days after Opus 4.7, Anthropic has accelera
Frontier AI Labs Refuse to Disclose Containment Strategies for Rogue Models
Recent research indicates that very few leading AI laboratories have published or demonstrated containment response plans. A containment plan defines the procedures for when an AI system attempts to subvert human control, specifying which access righ
Anthropic Unveils Fable: A More Affordable, Less Restrictive AI Model
Anthropic launched Fable and Mythos 5.1 on Tuesday, offering paired iterations of its latest AI technology. Beyond performance enhancements, the updated Fable version reduces token usage and tightens safety filters to minimize false-positive restrict
Related Special Topic Recommendations
Comments (0)
0/500

Earlier this month, xAI finalized a significant compute arrangement with Anthropic, committing billions per month for exclusive access to the Colossus cluster. The deal proved advantageous for both sides: xAI gained essential revenue, while Anthropic moved forward in the ongoing competition for computing resources.
However, this morning on X, Elon Musk played down the extent of SpaceX's commitment to the agreement.
“SpaceX has not committed to a multi-year lease of Colossus, though that could still happen,” he responded to a user. “It’s a 180-day lease with a mutual 90-day cancellation clause after that. We requested the short term, not Anthropic. We won’t abandon them and will offer a fair exit, but if computing capacity becomes extremely scarce, I mentioned we might need to reclaim it at some point.”
Musk's remarks directly conflict with SpaceX's recent S-1 filing, which acknowledges the standard 90-day cancellation clause yet describes the deal as a three-year commitment. Page F-62 of the filing states:
On May 3, 2026, the Company signed a cloud services agreement with Anthropic PBC, an AI research and development public benefit corporation, for access to compute capacity. Under this agreement, the customer agreed to pay a monthly fee through May 2029, with capacity gradually increasing starting May 2026 at a reduced rate. Either party may terminate the agreement with 90 days' notice. The customer retains ownership and intellectual property rights over its content, AI models, and related data.
The crucial detail is that Anthropic "has agreed to pay a monthly fee through May 2029" – a clear indication of a three-year lease. This phrasing appears again on page F-96 and in a slightly different version ("the customer has agreed to pay us $1.25 billion per month through May 2029") on pages 13 and 146, ruling out a simple typo.
xAI did not reply to a request for clarification.
One could argue over whether Anthropic's commitment to pay for a service equates to SpaceX's commitment to provide it, but that is not typically how a lease works. Moreover, why implement a one-sided lock-in if either side can cancel the deal with just three months' notice?
I do not have the actual contract, so I cannot confirm its terms – and neither SpaceX nor Anthropic has disclosed the duration in their public statements. Nevertheless, the facts should be clear, and making false claims during a company's quiet period is not advisable.
As usual, the SEC is unlikely to take action – and even if it did, Elon would probably not be concerned. However, this appears to be a material misrepresentation made while marketing a security, which at the very least reflects poorly.
Sean O'Kane contributed reporting to this article.
Anthropic launches Opus 4.8 featuring new dynamic workflow tool
Anthropic unveiled Opus 4.8 on Thursday, marking the latest iteration of its premier public model. Priced identically to its predecessor, this update is now accessible across all platforms.Releasing just 41 days after Opus 4.7, Anthropic has accelera
Frontier AI Labs Refuse to Disclose Containment Strategies for Rogue Models
Recent research indicates that very few leading AI laboratories have published or demonstrated containment response plans. A containment plan defines the procedures for when an AI system attempts to subvert human control, specifying which access righ
Anthropic Unveils Fable: A More Affordable, Less Restrictive AI Model
Anthropic launched Fable and Mythos 5.1 on Tuesday, offering paired iterations of its latest AI technology. Beyond performance enhancements, the updated Fable version reduces token usage and tightens safety filters to minimize false-positive restrict





Home






