Reflection AI secures $1 billion Nebius deal for Nvidia GB300 chips

Reflection AI locked in over $1 billion in computing capacity from Nebius, securing Nvidia's latest chips through 2029.

AI startups are racing to lock down computing power as demand for AI infrastructure outpaces new data-center supply, with Reflection's $1 billion-plus Nebius deal marking the latest long-term capacity commitment.

"The need for open models is clear, and this additional compute capacity will allow Reflection to continue to build and train frontier AI models at scale," said Ioannis Antonoglou, chief technology officer and co-founder of Reflection.

The agreement runs through 2029 and gives Reflection access to Nvidia Corp.'s GB300 AI chips, the companies said Tuesday. The deal builds on Reflection's June agreement with SpaceX for computing capacity, which media reports said would cost the startup about $150 million a month through 2029. Combined, the two contracts represent one of the largest compute commitments by an AI startup this year.

The computing capacity arms race reflects a structural shift in the AI industry, where access to high-end chips and data-center space has become a competitive moat. Open-source models, which are typically easier to customize and cheaper to run than closed-weight rivals, have drawn growing interest as rising AI bills push businesses to cut costs. Last month's US curbs on Anthropic's advanced models also exposed the risks of relying on providers that can be cut off overnight.

Reflection, launched by two former Google DeepMind researchers, develops open-source models that serve as an alternative to offerings from OpenAI and Anthropic. The startup's dual compute agreements — with both Nebius and SpaceX — indicate that AI companies are increasingly treating computing capacity as long-term infrastructure rather than a pay-as-you-go utility.

For Nebius, the Amsterdam-based cloud provider that emerged from Yandex's corporate restructuring, the deal provides contracted revenue visibility that can help fund new data-center construction and Nvidia hardware purchases. The more customers reserve capacity upfront, the less operators depend on short-term utilization swings, pushing the AI supply chain toward an infrastructure business model with visible backlog.

The deal also underscores the intensifying competition among cloud providers for AI workloads. Nebius, which trades on Nasdaq under the ticker NBIS, is positioning itself as a specialized AI infrastructure provider, competing with hyperscalers like Amazon Web Services, Microsoft Azure and Google Cloud. Unlike the hyperscalers, Nebius focuses exclusively on GPU-heavy workloads, offering direct access to Nvidia's latest silicon without the bundled software services that can lock customers into a single ecosystem.

The deal is bullish for Nebius and Nvidia, reflecting sustained demand for AI computing infrastructure. Nebius gains a marquee customer with contracted revenue through 2029, while Nvidia secures another large-scale deployment of its GB300 chips. For investors, the shift toward multi-year compute commitments suggests that AI infrastructure spending is becoming more predictable and less cyclical — a dynamic that could support premium valuations for providers with contracted backlogs.

This article is for informational purposes only and does not constitute investment advice.