Overview
- The acquisition was announced Thursday, Oct. 1, 2026, and Nebius confirmed it will integrate Inferize into Token Factory while not disclosing financial terms.
- Industry sources estimate the deal at roughly $100–$150 million, though Nebius has not confirmed that figure.
- Inferize, founded in January 2026 and run by former Granulate engineers, built software to reduce GPU idle time during model cold-starts and traffic spikes so operators do not need to keep costly warm spare GPUs.
- This purchase continues Nebius’s strategy of buying small optimization teams — following Tavily and Eigen AI — to raise GPU throughput and lower the cost of serving models on Token Factory.
- If integrated as planned, Inferize’s tech could let Nebius run hardware closer to real demand, improving unit economics for customers and supporting the company’s expansion of Israeli data-center capacity and inference services.