Tel Aviv-based startup Niv AI has emerged from stealth mode with innovative solutions to extract maximum power and performance from GPUs in AI data centers.
The company raised $12 million in seed funding from prominent investors including Glilot Capital, Grove Ventures, Arc Ventures, Encoded Ventures, Leap Forward Ventures, and Aurora Capital Partners.
Revolutionizing Power Management in AI Infrastructure
Founded last year by CEO Tomer Timor and CTO Edward Kizis, Niv AI addresses surging power demands in data centers packed with thousands of high-performance GPUs.
Millisecond-scale power spikes from GPU task switches and communications cause up to 30% throttling, squandering revenue on costly chips and forcing reliance on temporary energy storage.
Niv AI's rack-level sensors deliver precise, real-time GPU power measurements, enabling data centers to optimize usage and avoid inefficiencies.
Roadmap, Impact, and Industry Echoes
The startup plans initial sensor deployments with design partners and operational systems in select US data centers within 6-8 months.
Future developments include an AI-powered "copilot" to predict and synchronize power loads, evolving into an intelligence layer bridging data centers and electrical grids.
This approach helps hyperscalers overcome land-use hurdles and supply chain delays without endless new constructions.
Nvidia CEO Jensen Huang recently lamented vast power waste in AI factories, underscoring the urgency for breakthroughs like Niv AI's.
Grove Ventures partner Lior Handlesman, now on Niv AI's board, warned that traditional data center builds are no longer viable.
By maximizing existing capacity and crafting responsible power profiles, Niv AI promises to alleviate grid anxieties over sudden AI-driven surges.
CEO Tomer Timor highlighted the dual focus: empowering data centers to utilize more paid power while fostering grid-friendly operations.
As AI proliferation strains global energy systems, Niv AI's advancements could redefine scalable, efficient AI infrastructure worldwide.
Alfred Lee
