Nasdaq-listed neocloud Nebius acquires Inferize, whose tech helps optimize GPU utilization and reduce AI request costs, sources say for $100M to $150M

Your AI inference costs may decrease as GPU utilization improves.

Nasdaq-listed neocloud company Nebius has acquired Israeli startup Inferize in a deal reportedly worth between $100 million and $150 million. Inferize specializes in technology designed to optimize GPU utilization and reduce the costs associated with AI inference, particularly during fluctuations in demand. Founded in early 2026 by former Granulate team members Guy Bortnikov and Lior Gorbonos, Inferize was operating in stealth and had raised a Seed round. The startup's technology addresses the challenge of idle GPU capacity by allowing AI inference to scale more closely with actual demand, tackling the inefficiencies of "cold starts" and managing demand spikes. This acquisition aligns with Nebius's strategy of enhancing its AI computing infrastructure with specialized software and technology acquisitions, following its earlier agreement to acquire Tavily. Nebius plans to integrate Inferize's capabilities into its Token Factory to improve customer demand response. The move underscores the growing importance of inference costs and speed in the AI infrastructure market as companies shift focus from model training to deployment.

The acquisition of Inferize by Nebius highlights a critical, yet often overlooked, segment of the AI infrastructure market: inference optimization. As AI models move from research and training to widespread customer deployment, the efficiency and cost of serving requests in real-time become paramount. Inferize's technology directly addresses the "cold start" problem and demand variability, which can lead to significant underutilization of expensive GPU resources. Nebius's integration of this technology signals a move to offer more cost-effective and responsive AI services to its clients, potentially setting a new standard for operational efficiency in AI cloud platforms.

This deal indicates a market trend where AI cloud providers are aggressively acquiring niche technologies to build out a comprehensive software stack around their core compute offerings. Nebius's strategy of combining substantial computing capacity with targeted software acquisitions, like Inferize and Tavily, demonstrates a clear intent to provide end-to-end solutions rather than just raw power. Companies operating at the forefront of AI deployment will likely face increased pressure to optimize their inference costs, making solutions like Inferize's highly sought after and potentially driving further consolidation in the AI infrastructure sector.