TL;DR
- Lease Clash: Elon Musk says Anthropic’s Colossus access is a 180-day lease, while payment terms point to May 2029.
- Customer Planning: The duration affects whether Claude users can treat higher limits as durable capacity or temporary headroom.
- Capacity Scale: Colossus includes more than 220,000 GPUs, more than 300 megawatts, and a reported $1.25 billion monthly cost.
- Supply Strategy: Anthropic’s CoreWeave deal shows SpaceX is one lane in a broader external-compute buying pattern.
Anthropic agreed to a partnership with SpaceX to substantially increase compute capacity, and now Musk is challenging how long that access will last. His claims about a 180-day lease claim now sits against payment terms tied to May 2029.
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
For Claude customers, a short lease would make Colossus a temporary bridge, while a longer term would make it part of Claude’s supply base and change how customers read Anthropic’s capacity promises.
Before the public argument, SpaceX and Anthropic created a cloud services agreement for compute capacity. Anthropic connected the supply to higher usage limits for Claude across Pro, Max, Team, and seat-based Enterprise plans, along with higher Claude Code usage limits and higher Claude API rate limits.
For developers, the Claude capacity expansion gave customers a product reason to care about because compute capacity is the processing power needed to run and serve AI models at higher volume.
A replacement plan for SpaceX compute would need matching GPUs, power, network capacity, and deployment work, not only another vendor name.
Why Duration Affects Claude Planning
Under the payment structure, Anthropic’s obligation runs as a monthly fee through May 2029, while the first two months used an unspecified reduced fee. Either side can end the deal with 90 days’ notice, so the same arrangement contains both a long payment path and a near-term exit mechanism.
If either side used that notice right, Anthropic would need to move workloads, renegotiate access, or replace a large block of supply while keeping Claude service commitments stable for enterprise customers. Moving production workloads at that scale is different from simply scaling down a small pilot.
At Colossus scale, the end date becomes operationally important: more than 220,000 NVIDIA GPUs, all Colossus 1 capacity in Memphis, Tennessee, and more than 300 megawatts of compute capacity. Megawatts translate to the power footprint behind how much hardware can run continuously.
With monthly costs at stake, Anthropic would pay $1.25 billion per month through May 2029 once the full rate took effect. Anthropic had already locked in outside compute through a multi-year CoreWeave infrastructure deal earlier in 2026, so the SpaceX fight is about the durability of one important lane inside a broader capacity-buying pattern.
