Open Source For You reports that SpaceX has signed a compute-capacity lease with open-source startup Reflection AI worth up to $6.3 billion through 2029, granting the firm priority access to Nvidia GB300 'Blackwell Ultra' GPUs in SpaceX’s Colossus 2 data center. Reflection AI will pay around $150 million per month starting July 1, 2026, with a three‑month trial and a 90‑day termination clause.
This article aggregates reporting from 1 news source. The TL;DR is AI-generated from original reporting. Race to AGI's analysis provides editorial context on implications for AGI development.
This is one of the clearest signs yet that AI compute has become its own asset class. A $6.3 billion lease for Nvidia Blackwell Ultra capacity, brokered by SpaceX rather than a traditional cloud, shows how quickly the supply chain is fragmenting: rockets and satellites on one side, open‑source frontier models on the other, with billions of dollars of GPU time in between.
Strategically, the deal cements Reflection AI as a flagship buyer for SpaceX’s Memphis‑based Colossus 2 complex, alongside incumbents like Google and Anthropic. If the reported numbers hold, Reflection goes from being just another open‑source lab to a compute superpower with guaranteed access to thousands of GB300s. That could tilt the open‑weights race, especially if the company truly positions itself as a US‑based, open alternative to China’s DeepSeek and GLM models at scale.
For the broader race to AGI, the story underlines that the bottleneck is shifting from model architecture to capital‑intensive infrastructure. Whoever can marshal tens of billions into power, land, and advanced GPUs will dictate the cadence of training new frontier systems. SpaceX, Nvidia and a new cohort of 'compute‑rich' labs like Reflection are building a parallel universe to the hyperscalers—one where open‑source foundations could plausibly train AGI‑class systems if the economics pan out.