SpaceX announced a new compute partnership that will see open‑source AI startup Reflection AI rent Nvidia’s GB300 AI chips and supporting hardware at the company’s Colossus 2 data center outside Memphis, Tennessee. The contract begins on July 1 2026 and runs through July 2029, with Reflection paying $150 million per month – a total that could climb to $6.3 billion if the agreement remains in force.
Reflection AI, which launched in 2024, was created by two former Google DeepMind researchers. The firm describes itself as an “open‑weight” AI lab, releasing trained model parameters publicly to offer an alternative to closed‑source systems from labs such as Anthropic and OpenAI. The company highlighted the SpaceX deal as one of the largest infrastructure commitments ever announced for an open‑source AI project.
Unlike SpaceX’s earlier contracts with Anthropic and Google, which command $1.25 billion and $920 million per month respectively, the Reflection agreement is markedly smaller. Both the Anthropic and Google deals also run through July 2029, but SpaceX’s chief executive, Elon Musk, has emphasized that the contracts can be canceled at any time, underscoring the flexibility built into the new arrangement. After the first three months, either party may terminate the deal with 90 days’ notice.
The Colossus 2 facility was originally built by xAI, Musk’s own artificial‑intelligence venture, before becoming part of SpaceX’s broader AI infrastructure. As xAI’s internal projects stalled, SpaceX repurposed the data center’s high‑performance hardware for external customers, turning its chip inventory into a revenue stream.
“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,” a Reflection spokesperson said in an emailed statement. “Our deal with SpaceX AI 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.”
The partnership arrives amid heightened scrutiny of closed‑source AI models. The U.S. government recently banned Anthropic’s closed models, Fable and Mythos, citing security concerns. Reflection’s open‑weight approach, which publicly releases model parameters, is positioned as a safer, more transparent alternative.
Industry observers note that SpaceX’s move to lease compute capacity reflects a broader trend of aerospace and satellite firms monetizing their onboard AI hardware. By offering access to cutting‑edge Nvidia chips, SpaceX not only diversifies its revenue but also supports the growth of a nascent segment of the AI market that prioritizes openness and collaboration.
Este artículo fue escrito con la asistencia de IA.
News Factory APP - noticias agénticas para impulsar tu SEO y AEO.