Reflection Secures $1bn AI Compute Deal With Nebius

Reflection AI has signed a deal worth more than $1 billion with Nebius to secure computing capacity through 2029, giving the startup access to Nvidia’s latest AI chips as competition for computing infrastructure intensifies.

The agreement expands Reflection’s access to the hardware needed to train and run advanced artificial intelligence models. It also highlights the growing demand for AI computing power as businesses adopt generative AI faster than new data centres come online.

The deal follows a separate computing agreement that Reflection signed with SpaceX in June. Media reports said that the contract could cost the startup about $150 million a month through 2029.

Reflection was founded by two former Google DeepMind researchers. The company develops open-source AI models that compete with offerings 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,” said Ioannis Antonoglou, Reflection’s co-founder and chief technology officer.

AI developers have been signing large infrastructure agreements to secure scarce computing resources. Training advanced models requires thousands of high-performance chips and the data centres needed to run them.

Demand has continued to rise as more companies deploy AI tools. However, new data-centre construction has struggled to keep pace, making long-term computing agreements increasingly valuable.

Reflection says its focus on open-source models gives businesses more flexibility than closed AI systems. Open models are often easier to customise and can cost less to operate. They have also gained attention as companies look to reduce AI spending.

Interest in open models has grown following recent U.S. restrictions on Anthropic’s advanced AI models. The measures highlighted the risks businesses face when they rely heavily on a single AI provider.

The Nebius agreement also strengthens Reflection’s position as it expands its AI infrastructure. Last month, the startup reached a multibillion-dollar computing agreement with SpaceX to access Nvidia’s GB300 AI chips.

Nebius, an Amsterdam-based AI infrastructure provider, separated from Russian internet company Yandex in 2024. The company rents AI computing capacity and already counts Microsoft and Meta among its customers.