Elon Musk says SpaceX only agreed to 6-month Colossus AI lease with Anthropic despite $1.25B deal
Elon Musk is dialing back expectations around SpaceX’s blockbuster AI infrastructure deal with Anthropic.
Just weeks after reports surfaced that Anthropic had secured access to one of the largest AI compute clusters in the United States, Musk said on Thursday that SpaceX only agreed to a six-month lease for its Colossus AI training systems, raising fresh questions about how long the partnership may actually last.
“SpaceX has not committed to leasing Colossus for years, although it’s possible that may be what happens,” Musk wrote in a post on X.
Musk added, “This is a 180 day lease with 90 day notice mutual cancellation thereafter. The short term was our request, not Anthropic’s. 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.”
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. The short term was our request, not Anthropic’s.
We won’t leave them hanging and will provide a…
— Elon Musk (@elonmusk) May 28, 2026
Inside SpaceX’s $1.25B AI infrastructure deal with Anthropic and why Musk may want it back
The clarification comes after SpaceX disclosed in its IPO filing that Anthropic would pay the company roughly $1.25 billion per month to access compute capacity from its Colossus and Colossus II clusters in Memphis through May 2029. The filing stated that both companies could terminate the agreement with a 90-day notice, though it made no reference to the six-month structure Musk later described publicly.
Musk said the arrangement was intentionally short-term on SpaceX’s side.
“The short term was our request, not Anthropic’s,” he said on X.
He added that the company wants flexibility in case demand for AI compute spikes further.
“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,” the billionaire wrote.
The comments offer a rare look into how strained the AI infrastructure market has become. Access to large GPU clusters has turned into one of the biggest bottlenecks in artificial intelligence, with companies racing to lock in Nvidia hardware and data center capacity before supply tightens further.
SpaceX’s Memphis facility, known as Colossus 1, has become one of the most closely watched AI infrastructure projects in the country. The site houses more than 220,000 Nvidia GPUs and supports large-scale AI model training workloads.
The company has been quietly building a business around AI compute services at a time when demand from model developers is surging. Last week, Musk posted on X that SpaceX was in talks with other companies about “offering AI compute as a service at significant scale.”
That business is still burning cash.
According to SpaceX’s IPO filing, the company’s AI segment posted a $2.5 billion operating loss during the March quarter on revenue of $818 million, Reuters reported. The numbers show how expensive the AI infrastructure race has become, particularly for companies building massive GPU clusters from scratch.
The filing arrived just days after SpaceX officially moved ahead with plans to go public, setting up one of the most anticipated tech IPOs in years.
For Anthropic, the deal gave the OpenAI rival access to one of the largest compute environments available outside the traditional cloud giants. For SpaceX, it opened a new revenue stream tied to the AI boom.
Musk’s latest comments suggest the agreement may stay more flexible than the original filing implied.

