Business & Enterprise | 3 min read

Anthropic Is Paying xAI $1.25 Billion Per Month for Supercomputer Access

Anthropic is paying Elon Musk's xAI $1.25 billion per month for access to the Colossus supercomputer in Memphis — a $45 billion, three-year compute commitment revealed in SpaceX's S-1 SEC filing.

Hector Herrera
Hector Herrera
A modern corporate office featuring contract, related to an AI safety company Is Paying xAI $1.25 Billion Per Month f from an unusual angle or perspective
Why this matters Anthropic is paying Elon Musk's xAI $1.25 billion per month for access to the Colossus supercomputer in Memphis — a $45 billion, three-year compute commitment revealed in SpaceX's S-1 SEC filing.

Anthropic Is Paying xAI $1.25 Billion Per Month for Supercomputer Access

By Hector Herrera | May 21, 2026

Anthropic is paying Elon Musk's xAI $1.25 billion per month for access to the Colossus 1 supercomputer in Memphis — a deal worth up to $45 billion over its three-year term. The figure, which surfaced in SpaceX's S-1 SEC filing, is the largest known compute contract in AI history and underscores how severely Anthropic's GPU access constrained the company even after its $30 billion funding round.

Background

Anthropic has been capacity-constrained since its founding. Unlike OpenAI, which has a preferred compute arrangement with Microsoft Azure, or Google DeepMind, which runs on Google's TPU infrastructure, Anthropic has relied on Amazon Web Services (AWS) and spot-market GPU access. The company's rapid growth — it now claims over $3 billion in annualized revenue — has outpaced its ability to secure enough compute through conventional channels. The xAI deal is the result: a massive, multi-year commitment to a competitor's hardware to keep Claude training and inference running at scale.

The Numbers

According to TechCrunch, the deal structure includes:

  • $1.25 billion per month through May 2029 — a 36-month commitment totaling approximately $45 billion
  • Access to the Colossus 1 supercomputer in Memphis, Tennessee, which houses over 220,000 NVIDIA GPUs across 300+ megawatts of power capacity
  • Expansion into Colossus 2, xAI's next-generation cluster, as it comes online
  • Reported interest from Anthropic in exploring space-based compute development with xAI

The financial terms emerged from SpaceX's S-1 SEC filing — SpaceX is the parent entity that built and operates the Colossus facility for xAI's use.

Why This Deal Exists

The AI compute market is running at near-total constraint. NVIDIA's top-tier GPUs — the H100 and H200 series, and now the Blackwell B200 — are allocated years in advance by hyperscalers (Microsoft, Google, Amazon, Meta). AI labs that don't have a preferred cloud partnership or didn't lock in orders two years ago are effectively rationed.

Anthropic's deal with xAI is, in part, a symptom of that shortage. Colossus 1 is one of the only large concentrations of NVIDIA GPU compute outside the major cloud providers. xAI built it rapidly — in roughly 120 days — specifically to train Grok models. With Grok 3's training complete, that hardware has spare capacity Anthropic can pay to use.

The arrangement is strategically unusual. Anthropic and xAI compete directly in the frontier AI market. Anthropic's Claude and xAI's Grok both target enterprise customers, and both companies are raising money at overlapping valuations. That Anthropic is paying xAI $1.25 billion per month — effectively subsidizing a competitor's infrastructure — reflects how desperate the compute situation is, not alignment of interests.

What It Means for the Industry

Compute is the bottleneck. The deal makes explicit what insiders have known: frontier AI development is constrained not by talent, capital, or algorithms, but by physical GPU availability. The company that controls compute infrastructure controls the pace of AI development.

xAI gains a massive, stable revenue stream. At $1.25 billion per month, the Anthropic contract alone funds xAI's operations at a scale that rivals its AI revenue. This transforms Colossus from a cost center into a profit engine while xAI continues developing Grok models.

The hyperscaler model is under pressure. If AI labs are going directly to competitors' supercomputers rather than AWS, Azure, or Google Cloud, that signals the hyperscalers can't meet frontier AI demand at any price. Watch for AWS and Google to announce expanded GPU capacity commitments in response.

Anthropic's $30 billion valuation now has context. The company's recently disclosed $900 billion valuation raised eyebrows. A $45 billion compute commitment over three years, on top of ongoing AWS costs and its workforce, clarifies the cash burn profile. Revenue growth needs to be exceptional to sustain this model.

What to Watch

The SpaceX S-1 disclosure is likely a preview of more structural compute deals between AI labs. Watch for whether this arrangement influences Anthropic's IPO timeline — a $45 billion compute liability is a major line item for any prospective public investor. Also watch for whether xAI's Colossus 2 expansion attracts additional third-party tenants, effectively turning xAI into a compute-as-a-service provider alongside its AI product business.


Hector Herrera covers AI business and infrastructure for NexChron.

Key Takeaways

  • By Hector Herrera | May 21, 2026
  • $1.25 billion per month
  • Colossus 1 supercomputer
  • space-based compute development
  • Colossus 1 is one of the only large concentrations of NVIDIA GPU compute outside the major cloud providers.

Did this help you understand AI better?

Your feedback helps us write more useful content.

Hector Herrera

Written by

Hector Herrera

Hector Herrera is the founder of Hex AI Systems, where he builds AI-powered operations for mid-market businesses across 16 industries. He writes daily about how AI is reshaping business, government, and everyday life. 20+ years in technology. Houston, TX.

More from Hector →

Get tomorrow's AI briefing

Join readers who start their day with NexChron. Free, daily, no spam.

More from NexChron