Reflection AI closes billion-dollar deal with SpaceX for compute capacity
Startup Reflection AI is developing an open-source AI model. Starting in July, it will use compute capacity from SpaceX's AI data center in Memphis.
(Image: Fotogrin / Shutterstock.com)
Reflection AI thinks with SpaceX: The AI startup Reflection AI will use SpaceX's infrastructure in the future. This is a good business for SpaceX.
Reflection AI will use SpaceX's Colossus-2 data center in the future. Reflection AI will pay 150 million US dollars per month for this. It is the startup's first contract for the use of compute capacity. This is reported by various media, including the US online magazine Techcrunch and the US business news agency Bloomberg.
The deal runs until 2029 and has a total value of up to 6.3 billion US dollars, with both sides able to terminate the contract with three months' notice. It starts in July.
Colossus was built by xAI
According to its information, the Colossus-2 data center in Memphis, Tennessee, has 200,000 Nvidia H100 GPUs. It was built by Elon Musk's AI company, xAI. Through the acquisition of xAI by SpaceX earlier this year, it belongs to Musk's space company.
Anthropic and Google also use the capacities of the data center. However, they pay significantly more than Reflection: Anthropic transfers 1.25 billion US dollars per month to SpaceX, Google 920 million US dollars.
Reflection AI was founded in 2024 by two former employees of Google's AI division, Deepmind. The startup has developed an open AI model that is intended to compete with the closed systems of OpenAI, Anthropic, or Google. The company is currently valued at 25 billion US dollars.
Videos by heise
“Recent events highlight how important open source is to the AI ecosystem, with more nations and enterprises recognizing the risks and costs associated with exclusively depending on closed models,” says a statement from Reflection AI. “Our deal with SpaceXAI signals Reflection’s strategic importance within the frontier AI ecosystem, and more compute means more runway to build the world’s best open models at scale.”
(wpl)