Key facts
- AI startup Reflection secured a computing capacity deal valued at over $1 billion with Nebius.
- The agreement includes access to Nvidia's latest chips.
- Reflection aims to use this capacity to train and scale its open-source AI models.
- The startup was founded by former Google DeepMind researchers.
- Reflection develops open-source models as an alternative to closed-source offerings from OpenAI and Anthropic.
AI startup Reflection has announced a significant computing capacity deal valued at over $1 billion with Nebius, a European AI infrastructure company formerly part of Yandex. The agreement grants Reflection access to Nvidia's latest chips, crucial for its mission to train and scale open-source AI models as an alternative to closed-weight models from competitors like OpenAI and Anthropic.
The burgeoning demand for AI computing power, driven by rapid AI adoption and limited data center supply, has led AI firms to actively secure necessary resources. Reflection, founded by former Google DeepMind researchers, focuses on open-source models, which offer greater customization and cost-effectiveness, appealing to businesses concerned about rising AI expenses and potential restrictions on proprietary models.
Ioannis Antonoglou, CTO and co-founder of Reflection, stated that the compute capacity will enable the company to continue building and training frontier AI models at scale, highlighting the market's need for open models. This deal follows a similar compute access agreement Reflection signed with SpaceX.