gpu-utilization
(3)
MemVerge
MemVerge is a Milpitas, California AI infrastructure software company that started out virtualizing DRAM and persistent memory into pooled 'Big Memory' and has since pivoted toward the AI era. Its Memory Machine products maximize GPU utilization through memory tiering, pooling and transparent checkpointing, and in 2025 it launched MemMachine, an open-source AI memory layer that gives LLMs and agents persistent episodic, profile and procedural memory across sessions, models and clouds. Founded in 2017 by CEO Charles Fan, CTO Yue Li and chairman Shuki Bruck, the company has raised about $43.5M from investors including Intel Capital, Cisco, NetApp, SK hynix and Lightspeed.
Hammerhead AI
Hammerhead AI is a Redwood City data center infrastructure startup that uses reinforcement learning agents to turn stranded power into usable AI compute. Its ORCA platform (Orchestrated RL Control Agents) coordinates power, cooling, and GPU workloads in real time so operators can raise token throughput by up to 30% within their existing power envelope - without waiting for new grid capacity. Founded in 2024 by former AutoGrid executives Rahul Kar and Rajeev Singh, the company emerged from stealth in November 2025 with a $10M seed round led by Buoyant Ventures.
Rahul Kar
Rahul Kar is the founder and CEO of Hammerhead AI, a Redwood City startup that uses reinforcement-learning control agents to squeeze stranded power out of AI data centers. After helping build and scale AutoGrid - the distributed-energy software pioneer acquired by Schneider Electric - Kar launched Hammerhead in November 2025 with a $10M seed round led by Buoyant Ventures. His thesis is blunt: the real bottleneck for AI is not chips or models, it is power, and most data centers run at 30-50% of their allocated capacity. Hammerhead's ORCA platform orchestrates power, cooling, and compute in real time to unlock roughly 30% more usable capacity without touching the grid.