Artificial intelligence (AI) startup Reflection AI has signed a landmark computing agreement worth more than $1 billion with cloud provider Nebius Group NV, securing critical access to NVIDIA Corp.’s highly coveted GB300 AI chips through 2029.
The multiyear deal announced Tuesday is the latest high-priced move in an industrywide race to lock down the massive processing power required to train and deploy next-generation AI models, where demand heavily outpaces data-center supply.
Founded by two former Google DeepMind researchers, Reflection develops open-source AI models designed as cost-effective, customizable alternatives to closed-weight systems from OpenAI and Anthropic.
“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,” Ioannis Antonoglou, Reflection’s co-founder and chief technology officer, said in a statement.
The agreement builds on another massive infrastructure play by the startup. Last month, Reflection signed a multibillion-dollar agreement with SpaceX reportedly valued at roughly $150 million a month through 2029 to secure the same generation of NVIDIA processors.
Backed by NVIDIA and other prominent investors, Reflection is also reportedly in discussions to raise $2.5 billion in fresh funding at a $25 billion valuation, according to The Wall Street Journal.
The partnership represents a significant win for Amsterdam-based Nebius Group. Since spinning off from Russian internet provider Yandex in 2024, Nebius has emerged as a premier neocloud provider renting high-performance AI compute, counting tech giants Microsoft Corp. and Meta Platforms Inc. among its clients.
As corporate AI expenditures skyrocket, interest in open-source models has intensified.
Beyond lower operating costs, industry experts note that recent U.S. regulatory curbs on Anthropic’s advanced models have exposed the geopolitical risks of relying on closed-door providers that can be restricted overnight, further accelerating the transition toward open-source alternatives.