Anthropic commits $15B/year for SpaceX's Colossus AI compute

SpaceX's S-1 IPO filing dropped a bombshell: Anthropic is paying $1.25 billion per month—$15 billion annually—for access to the Colossus data centers in Memphis, TN. The deal runs through May 2029 with a 90-day termination clause for either party.

This is nearly double SpaceX's total 2025 revenue of $18.7 billion. The agreement covers Colossus I and II, SpaceX's AI training centers. Anthropic's fees reduce during the initial capacity ramp-up this month and next.

Why AI companies are desperate for compute

The exit clause reflects the fast-moving AI industry. Anthropic's Claude competes with X's Grok, yet Anthropic is paying Musk's company for infrastructure. It's a sign of how desperate AI firms are for compute capacity, especially as data center builds face local opposition.

SpaceX CEO Elon Musk posted on X that SpaceX is "offering AI compute as a service at significant scale" and stands ready to offer similar deals to other AI companies.

SpaceX's AI spending spree

SpaceX merged with Musk's xAI earlier this year. The filing shows SpaceX spent $12.7 billion in capital expenditures on AI in 2025 (61% of total spend). In Q1 2026, it spent $7.7 billion on AI versus just $1 billion on its space division. SpaceX's AI division lost $6.3 billion on $3.2 billion revenue in 2025, and lost $2.5 billion on $818 million revenue in Q1 2026.

Anthropic's financial picture

Meanwhile, Anthropic expects its first quarterly operating profit. Revenue is projected to reach at least $10.9 billion, more than double the $4.8 billion from the just-ended March quarter.

Technical implications for developers

For developers using Anthropic's Claude API, this deal means more training capacity, potentially leading to faster model releases and lower latency. However, the dependency on a competitor's infrastructure raises questions about long-term pricing and reliability.

# Example: Checking Claude API response times
curl -X POST https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-3-opus-20240229",
    "max_tokens": 100,
    "messages": [{"role": "user", "content": "Hello"}]
  }'

With the Colossus deal, Anthropic can scale training on massive GPU clusters. This could mean larger context windows, better reasoning, and faster fine-tuning for custom models.

The bigger picture

This deal underscores the shift toward "compute-as-a-service" for AI. SpaceX is pivoting from rockets to data centers. For developers, it means more options for training infrastructure, but also consolidation of power among a few players.

The 90-day exit clause protects both parties if the AI landscape shifts. It's a pragmatic approach in a volatile market.

What developers should watch

  • Pricing changes: If Anthropic's costs go up, API prices may follow.
  • Model improvements: More compute often leads to better models.
  • Competition: SpaceX may offer deals to other AI companies, increasing competition in compute services.
  • Regulatory scrutiny: Such large deals may attract antitrust attention.

For now, the deal is signed. Developers can expect Anthropic's models to keep improving, but keep an eye on the fine print.