Reflection seals $1B AI compute deal with Nebius
▼ Summary
– Reflection AI signed a compute deal worth over $1 billion with Nebius for Nvidia GB300 chips, running through 2029.
– The startup was founded in 2024 by ex-Google DeepMind researchers, builds open-source models, and is valued at about $8 billion.
– AI startups are racing to lock in hardware as demand outpaces data center supply, driving up chip prices.
– Nebius is a neocloud that rents AI computing capacity, backed by Nvidia, with deals including Meta and Microsoft.
– Open models are gaining traction due to lower cost and customization, plus risks from potential restrictions on closed providers.
Reflection AI has secured a massive infusion of computing power. The US-based startup struck an AI compute deal valued at over $1 billion with Nebius, the Amsterdam-based cloud provider, as reported by Bloomberg. The agreement extends through 2029 and grants Reflection access to Nvidia’s latest GB300 chips.
Nebius confirmed the transaction. Its stock has more than doubled this year amid the AI spending boom. Shares initially rose on the announcement before pulling back later in the trading session.
This marks Reflection’s second major capacity agreement in recent weeks. Last month, the company signed a multibillion-dollar deal with SpaceX for the same class of Nvidia hardware. That arrangement is reportedly worth approximately $150 million per month through 2029.
Reflection was founded in 2024 by two former Google DeepMind researchers. The company develops open-source AI models, positioning itself as a cheaper and more customizable alternative to the closed systems offered by OpenAI and Anthropic.
The startup carries a valuation of roughly $8 billion. It has raised close to $2.6 billion from investors including Nvidia, Sequoia Capital, and Lightspeed Venture Partners. Reflection is also in discussions to raise an additional $2.5 billion at a $25 billion valuation, according to the Wall Street Journal.
“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 co-founder and chief technology officer Ioannis Antonoglou.
AI startups are scrambling to secure hardware. Business demand for AI technology is rising faster than new data centers can be built. Compute capacity has become the primary bottleneck, and chip prices continue to climb.
Open models are gaining traction for two main reasons. They are generally cheaper to operate and easier to customize. Additionally, last month the Trump administration pressured Anthropic and OpenAI to restrict some of their most powerful models, according to TechCrunch. That move exposed the risk of depending on a provider that could be cut off overnight.
Nebius operates as a neocloud provider, renting out AI computing capacity rather than building its own models. The company separated from the Russian internet group Yandex in 2024 and now trades on Nasdaq.
Nvidia has backed Nebius multiple times, from an early $700 million round to a later $2 billion investment. Nebius has also expanded up the stack by acquiring the inference startup Eigen AI.
The company’s client roster is impressive. Nebius signed a five-year deal with Meta worth up to $27 billion and an earlier agreement with Microsoft valued at up to $19.4 billion. The Reflection deal adds another high-profile name to the list.
The agreement also highlights how much of the open-model movement now depends on rented infrastructure. The models themselves may be open. The compute power behind them is anything but free.
(Source: The Next Web)
