Niv-AI

About Niv-AI

Niv-AI is a Tel Aviv-based technology company that has developed an AI-driven platform to manage power consumption in data centers. Founded in May 2025 by CEO Tomer Timor and CTO Edward Kizis, both veterans of Israel's elite intelligence and technology units, the company aims to resolve critical power bottlenecks created by high-performance computing and AI workloads. In March 2026, Niv-AI emerged from stealth, announcing a $12 million seed funding round led by Glilot Capital Partners, with participation from Grove Ventures, Arc VC, Encoded VC, Leap Forward Ventures, and Aurora Capital Partners.

The company's core technology addresses the issue of volatile, high-intensity power spikes generated by GPU clusters. These millisecond-level surges force data center operators to maintain significant power capacity buffers—leaving as much as 30% of contracted power unused—to prevent outages and ensure stability. This practice results in substantial underutilization of expensive infrastructure. Niv-AI's platform introduces a control layer between the energy delivery system and the computational workloads. It uses high-frequency sensors to capture granular, real-time data on power consumption, creating what the company calls an "electrical fingerprint" of each AI workload. This data is fed into AI models that predict power fluctuations and orchestrate workloads to smooth out demand before spikes occur, effectively balancing the load.

By managing workloads in real time, Niv-AI unlocks this "trapped" electrical capacity without reducing GPU utilization or requiring additional physical hardware like batteries and capacitors. The software-driven solution is designed to integrate with common data center orchestration stacks like Slurm and Kubernetes, allowing operators to safely increase computing output and density within their existing power envelope. This approach helps clients maximize the performance of their current infrastructure, turning stranded electrical headroom into productive AI performance. The company began early deployments with design partners and planned initial rollouts in several U.S. data centers in 2026.