Anthropic has committed to pay xAI $1.25 billion per month for compute capacity through May 2029, securing the entire 300-megawatt output of xAI's Colossus 1 data center near Memphis. The deal, which surfaced in SpaceX's S-1 filing with the SEC on May 20, could deliver more than $40 billion in total revenue to xAI if it runs to term.
The contract includes a discounted rate for the first 2 months as xAI completes its ramp-up. Either party can terminate with 90 days notice, preserving flexibility as both companies' compute needs shift. SpaceX disclosed the arrangement as part of its public offering preparations, noting that the deal lets xAI monetize unused infrastructure capacity.
The transaction follows Anthropic's surprise announcement earlier this month that it had secured the full capacity of Colossus 1. At the time, the financial terms were undisclosed. The pricing now revealed makes this one of the largest cloud-compute contracts in the AI industry, dwarfing typical cloud commitments even among frontier labs.
“allows us to monetize unused compute capacity in our infrastructure”— SpaceX, SEC filing disclosure
Key facts
- 01Anthropic pays xAI $1.25 billion per month through May 2029, with a discounted rate for the first 2 months during ramp-up.
- 02The deal secures 300 megawatts of compute capacity from xAI's Colossus 1 data center in Memphis, Tennessee.
- 03Total contract value could exceed $40 billion in revenue to xAI over the full term.
- 04Either party can terminate with 90 days notice, preserving optionality for both sides.
xAI appears to have overbuilt relative to its current demand. Usage of Grok, xAI's flagship AI assistant, has declined significantly in recent months, freeing up servers that the company is now selling to a direct competitor. The deal positions xAI as both a model provider and a cloud vendor, a hybrid stance rare among AI companies.
SpaceX framed the arrangement as strategic in its filing, stating the company expects to enter into additional similar services contracts. The neocloud model—where an AI company builds infrastructure for itself but sells excess capacity externally—has emerged as a way to offset the capital intensity of frontier AI development.
Most AI companies either build data centers for their own use or operate as pure cloud providers. xAI is attempting both simultaneously. The model works only if the company can maintain high utilization across its own products or sign anchor tenants at scale. Anthropic now fills that role for Colossus 1.
The deal also surfaces the scale mismatch between xAI's infrastructure buildout and its product traction. A $1.25 billion monthly bill implies Anthropic is running workloads large enough to justify locking in that capacity rather than buying spot compute on AWS or Azure. That suggests either Anthropic's training runs have scaled beyond what public clouds can deliver reliably, or the terms are favorable enough to make the commitment worthwhile despite lock-in risk.
“We believe our dual monetization strategy provides multiple pathways to generate returns on invested capital”— SpaceX, SEC filing disclosure
For xAI, the contract provides a revenue floor ahead of a potential IPO. Monetizing stranded capacity turns a balance-sheet liability into a cash-generating asset. For Anthropic, the deal secures compute independence from hyperscalers at a moment when GPU availability remains a constraint across the industry.
The 90-day termination clause is the tell. Both sides preserved optionality. If Grok usage rebounds or xAI signs a larger customer, it can exit. If Anthropic's own infrastructure comes online or a better deal emerges, it can walk. The clause suggests neither side views this as a permanent arrangement.
The deal restructures competitive dynamics in the AI infrastructure market. xAI is now simultaneously training against Anthropic in the model race and selling it the compute to do so. That creates an information asymmetry—xAI will have visibility into Anthropic's usage patterns and scaling behavior, even if contractual walls exist.
SpaceX's disclosure that it expects additional similar contracts signals xAI intends to operate as a neocloud provider at scale, not just offload spare capacity to one customer. If that model works, it changes the unit economics of frontier AI—capital costs get amortized across both internal model development and external cloud revenue, lowering the effective cost per parameter trained.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
