LPU-based ultra-low-latency inference. Closed a 2026 H1 round at $5B+ post; rapid expansion of cloud and on-prem footprint.
Open source ↗