Deep Tech · 2024
Purpose-built compute for real-time AI inference
Lumen
$9M
ARR
3.5×
Latency cut
Pre-seed
Our stage
Overview
We wrote Lumen's first check to rethink the hardware stack under low-latency AI.
The company
Lumen designs inference-optimized compute and the software layer around it, cutting the cost and latency of running large models in production for companies that serve millions of requests a day.
Our thesis
The frontier bet was that inference, not training, becomes the dominant cost of AI at scale — and that a focused team could win a slice of it. We backed a rare hardware-plus-compiler founding team pre-product, with capital patient enough for real silicon timelines.
The outcome
Lumen shipped its first developer platform and reached $9M ARR with a waitlist of enterprise design partners. We led the pre-seed and followed on in the seed.
Keep looking
