Reflection AI Signs Over $1 Billion Nvidia Compute Deal With Nebius Through 2029

Reflection AI, an open-source artificial intelligence startup founded by former Google DeepMind researchers, has signed a computing agreement worth more than $1 billion with cloud provider Nebius, securing access to Nvidia GPU capacity through 2029. The deal, reported by Reuters and Bloomberg on July 14, 2026, comes just weeks after the company committed up to $6.3 billion to SpaceX for similar compute capacity.

The agreement highlights a defining dynamic of the 2026 AI landscape: well-funded startups are spending billions on compute before shipping public products. For comprehensive coverage of the AI infrastructure arms race, follow AI Buzz Wire for breaking developments.

Founded by DeepMind Veterans

Reflection AI was founded in March 2024 by Misha Laskin and Ioannis Antonoglou, both formerly of Google DeepMind. Laskin worked on reward modeling for Google's Gemini project, while Antonoglou co-created AlphaGo, the DeepMind system that defeated world champion Lee Sedol at the game of Go in 2016.

The Financial Times reported that Reflection raised $2 billion at an $8 billion valuation in October 2025, with backers including Nvidia, Sequoia, Lightspeed, GIC, DST, Eric Schmidt, and Citi. The company has since returned to investors seeking a valuation above $20 billion, according to the same reporting.

Despite this fundraising firepower and a team of roughly 60 people, Reflection has not yet released a public model. The company's spending pattern — locking in massive compute capacity before products exist — reflects the belief among frontier AI labs that GPU supply is the scarcest and most decisive asset in the industry.

The Nebius Deal in Context

Nebius, the Amsterdam-listed cloud provider created after Yandex split from its Russian assets, has aggressively positioned itself as a major AI infrastructure player in 2026. In March, Nvidia invested $2 billion in Nebius as part of a broader partnership to expand AI cloud capacity.

The Wall Street Journal recently described Nebius as a serious challenger to CoreWeave, aided by a $27 billion agreement with Meta and a separate $17.4 billion deal with Microsoft. While Reflection's contract is smaller than those, it demonstrates that Nebius is building a diversified customer base rather than depending on a single anchor tenant.

The deal also signals that the market for AI compute is expanding beyond the largest technology companies. Reflection is buying capacity at a scale that a startup without the revenue profile of Google, Microsoft, or Meta could not previously sustain — an opening that neocloud providers have been waiting for.

The SpaceX Connection

The Nebius deal follows an even larger commitment. According to Axios, Reflection agreed in June to pay SpaceXAI, SpaceX's AI division, approximately $150 million per month for capacity at Colossus 2, a data center near Memphis, Tennessee, using Nvidia GB300 chips. Run through 2029, that arrangement totals roughly $6.3 billion.

For perspective, MarketWatch reported that Anthropic's SpaceXAI arrangement is worth about $1.25 billion per month, and Alphabet's is about $920 million per month. Both carry the same 90-day termination structure after an initial commitment period. Reflection's spending, while extraordinary for a 60-person startup, remains a fraction of what the largest AI companies are committing.

Government Ties

Reflection has also positioned itself close to government demand. Axios reported that the company is partnering with the Department of Energy's Genesis Mission, a federal AI science program launched in November 2025. The Guardian reported in May that Reflection is one of eight AI companies with Pentagon agreements for classified military work, alongside SpaceX, OpenAI, Google, Nvidia, Microsoft, Oracle, and Amazon Web Services.

These government relationships give Reflection's compute spending a clearer strategic rationale. The company is not only chasing developer attention but trying to demonstrate utility to Washington, large enterprises, and investors who want a Western open-source model company capable of competing with DeepSeek, Qwen, and the closed labs.

The Missing Product

The central question hanging over Reflection AI is straightforward: when will the public see a model? A research-heavy startup does not need more than $7 billion in committed compute for ordinary experiments. That level of capacity is required to train at frontier scale — or to convince the next round of investors that the company can.

Laskin and Antonoglou's DeepMind pedigrees lend credibility to the spending. The team behind AlphaGo has demonstrated the ability to achieve breakthroughs that reshape entire fields. But until Reflection ships a public model, the compute deals are the product — and the industry will be watching closely for what emerges from that investment.

What is certain is that GPU capacity, once secured, cannot easily be replicated by competitors. In a market where chips are the limiting factor, Reflection has made sure it has somewhere to train. Twice in a single month, the company has placed enormous bets on that proposition.

Stay Ahead of AI

The AI infrastructure race is reshaping the technology landscape. For daily updates on compute deals, model releases, and startup funding, visit AI Buzz Wire.

Read more AI news →