2 hours ago
2
SpaceX just inked a deal worth up to $6.3 billion to supply computing power to Reflection AI, an open-source artificial intelligence startup founded by former DeepMind researchers. The arrangement will give Reflection access to Nvidia’s GB300 chips housed at SpaceX’s Colossus 2 data center, a facility originally built for xAI’s Grok chatbot.
The deal structure
The contract is structured around monthly payments of $150 million beginning July 1, 2026, and extending through 2029. The agreement includes a termination clause allowing either side to walk away with 90 days’ notice after the first three months, meaning both companies have committed to at least a quarter of payments before the escape hatch opens.
Reflection AI was founded in 2024 by Misha Laskin and Ioannis Antonoglou, both alumni of Google’s DeepMind lab. The startup focuses on building open-source AI models and automation tools for coding. The company has already raised approximately $800 million in early funding rounds, with Nvidia serving as a notable backer. That financial support has pushed Reflection AI’s valuation to as high as $25 billion.
SpaceX becomes an AI landlord
The Colossus 2 facility was purpose-built to train and run xAI’s Grok models. Reflection AI isn’t even the first external client. SpaceX has reportedly struck similar arrangements with Anthropic and Google.
Reflection AI is also involved with the US Department of Energy’s Genesis Mission and has received clearance to operate within Pentagon classified AI networks.
What this means for the broader compute market
At $150 million per month, the deal sets a concrete price benchmark for what large-scale GPU access actually costs. Decentralized compute networks like Render, Akash, or io.net have long argued that decentralized GPU marketplaces can undercut centralized providers on price. A $6.3 billion reference deal gives them a very specific target to beat.
For traders and investors, the key variable to watch is whether Reflection AI’s $25 billion valuation holds up under the weight of $150 million monthly infrastructure costs. Burning through that much cash on compute alone requires the company to either generate massive revenue from its AI products or continue raising capital at escalating valuations.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) · 