Singapore just switched on a server rack powered by living human neurons
AI data centers now consume enough electricity to strain national grids — and that number keeps rising. On August 17, researchers at NUS Medicine, data center operator DayOne, and Melbourne-based Cortical Labs unveiled what they call the world's first independently operated biological server rack, running 20 CL1 modules filled with living human neurons in Singapore. The pitch is simple: biology does computing at a power cost that silicon struggles to match.
The hardware
Each CL1 unit houses neurons grown from human stem cells, connected to silicon via 59-electrode arrays. The cells process information in a closed-loop system that requires no external compute support. A single module draws around 25W; the full 20-unit rack uses 800–1000W total. For comparison, a high-density AI rack from Nvidia or similar can pull over 100kW. That is a significant gap on paper.
There is a catch. Much of that low wattage goes not toward computation but toward keeping the biology alive — pumps, gas mixing, temperature control, and filtration are all built into each unit. Cortical Labs has not released throughput or latency benchmarks, so how the system performs against conventional servers on actual workloads remains an open question.
?NUS Medicine, DayOne Cortical Labs unveiled a Biological Data Centre prototype which uses living neurons to support AI systems, marking a key milestone in biological computing. ? pic.twitter.com/9beeGYtkbT
— Yong Loo Lin School of Medicine (@NUSMedicine) August 17, 2026
What it costs and who's behind it
Cortical Labs shipped the first commercial CL1 units in 2025 at $35,000 each, with a cloud rental option also available. The company has raised $11 million and counts In-Q-Tel — the CIA's venture arm — and Horizons Ventures among its backers, per DCD. The neurons themselves stay viable for up to six months, per Cortical Labs' official specs, after which modules presumably need refreshing.
The Singapore deployment is positioned as a move from lab research toward commercial use. Cortical Labs founder Hon Weng Chong described it as translating biological computing "from research to commercial application." Target use cases include drug discovery, humanoid robotics, and fraud detection — all areas where pattern recognition matters more than raw speed.
Early days
Cortical Labs isn't alone in the field. Swiss startup FinalSpark runs a competing platform using 16 brain organoids and claims its bioprocessors consume one million times less energy than digital chips — though that figure comes from the company itself and has not been independently verified.
No UK regulatory approval process or NHS partnership has been announced. The prototype launched just days ago, and real-world performance data is still pending. For now, the CL1 rack is a compelling proof of concept, but it needs published benchmarks before it can be taken seriously as an alternative to silicon at scale.
