Nvidia, AMD Back Tensormesh's $20M AI Memory Fix

Tensormesh has raised $20 million from Nvidia, AMD, CoreWeave, and venture firms to tackle a core inefficiency in AI inference. The funding coincides with the launch of Tensormesh Inference, a SaaS platform using KV caching to store intermediate LLM computations and eliminate redundant GPU processing. The technology can deliver a 10-fold reduction in latency and GPU costs, with some customers achieving cache hit rates above 70%. Funds will expand hardware integrations and accelerate development of the open-source LMCache project underpinning the platform.