Positron designs memory-optimized silicon for AI inference.
Positron's Atlas accelerator serves production models on about a third of the power a top GPU draws. Thomas Sohmers, Barrett Woodside and Edward Kmett founded the company in 2023, and in February 2026 it raised $230M at a valuation above $1B to scale production of the chips.
The binding cost of running an AI model is electricity, not silicon. Optimizing the hardware around memory rather than raw compute cuts the power drawn per token to about a third, which turns an AI company into an energy company in everything but name.
NZVC invested in Positron from NZVC Fund II. We file the company under Deep Tech on the portfolio wall.
We backed Positron because inference is becoming the dominant AI workload and the power it draws is what decides whether serving a model is affordable. Optimizing the chip around memory rather than raw compute attacks that number directly, and the Atlas accelerator already serves production models on about a third of the power a top GPU draws. Founders Thomas Sohmers, Barrett Woodside and Edward Kmett raised $230M in February 2026 at a valuation above $1B, which tells us the market agrees the power problem is the one worth solving.