• SpaceX has signed a compute power agreement with open-source AI startup Reflection worth up to $6.3 billion, according to people familiar with the matter.
  • The deal is part of a broader push by SpaceX to monetize its Colossus data centers, which are powered by Nvidia GPUs, as the company pivots toward becoming a major AI infrastructure provider.
  • This follows similar multi-hundred-megawatt deals with Anthropic and Google, signaling SpaceX's aggressive expansion into AI compute services.

SpaceX Goes Big on AI Infrastructure

SpaceX, best known for its rockets, is rapidly building a presence in AI compute. The company's Colossus data centers and massive Nvidia GPU clusters are now being deployed to serve external customers, with the Reflection deal representing its largest publicly known agreement. The contract, worth up to $6.3 billion over several years, will provide compute capacity for Reflection's open-source AI models, according to people familiar with the negotiations. SpaceX declined to comment; Reflection did not respond to requests for comment.

The scale of the deal underscores a broader industry scramble for compute power. AI labs like Reflection, Anthropic, and others are contracting directly with infrastructure providers to secure the vast computing resources needed for training and inference, bypassing traditional cloud providers. “This is a land grab for compute,” said an industry executive familiar with the negotiations, speaking on condition of anonymity. “Whoever controls the GPUs controls the future of AI.”

A Strategic Pivot for SpaceX

SpaceX's push into AI compute is a strategic shift from its core aerospace business. The company has been quietly building a portfolio of data center assets and partnerships, positioning itself as a rival to hyperscalers like Amazon Web Services and Microsoft Azure in the AI infrastructure market. The Reflection deal, along with earlier agreements with Anthropic (300 MW) and Google, signals that SpaceX is serious about competing.

“SpaceX has the capital and the operational expertise to run massive data centers,” said an analyst at a major investment bank who tracks the sector. “They're leveraging their existing supply chain and energy procurement to offer competitive pricing.” The analyst noted that SpaceX's deals often include long-term power commitments, which help secure favorable rates from utilities.

Implications for Chipmakers and the Industry

The surge in large compute deals has implications for chipmakers like Nvidia and AMD, as well as for data center hardware ecosystems. SpaceX is a major customer of Nvidia, and its deals help drive demand for the latest GPU generations. The Reflection deal alone could require tens of thousands of H100 or B200 chips, providing a significant revenue boost for Nvidia.

However, the concentration of compute capacity among a few players has drawn scrutiny. Critics argue that deals like these could create an infrastructure bottleneck, where a handful of firms control access to the massive computing power needed to train cutting-edge AI models. “We're seeing a winner-take-all dynamic in compute,” said a policy researcher focused on AI governance. “That raises questions about competition and access.”

Looking Ahead

Short-term, industry watchers expect more multi-hundred-megawatt deals as AI labs rush to secure capacity ahead of a potential crunch. SpaceX is reportedly in talks with several other AI startups and cloud providers. Longer term, SpaceX's foray could reshape the AI infrastructure landscape, blurring the lines between aerospace, cloud computing, and energy.

This article was updated on [Date] to include additional context on market implications.