Singapore's First Biological Data Centre Runs on Living Brain Cells
Source: The Straits Times
Singapore's data centre industry has spent years learning to work around heat, water and power constraints — a 2019 pause on new builds made efficiency the price of admission. So it is fitting that the country's newest computing facility has no silicon servers to cool at all.
Singapore's data centre industry has spent years learning to work around heat, water and power constraints — a 2019 pause on new builds made efficiency the price of admission. So it is fitting that the country's newest computing facility has no silicon servers to cool at all. Tucked inside NUS's Centre for Life Sciences is a data centre with a pulse: 20 units of living brain cells doing the processing.
The facility went live on July 16, the product of a partnership between Australian biotech startup Cortical Labs, NUS and data centre operator DayOne that was first announced in March. Each CL1 unit houses at least 200,000 lab-grown neurons — derived from blood cells reprogrammed into stem cells — sitting on an electrode-fitted chip, with lab technicians feeding the cells a cocktail of sugar, micronutrients and pH buffers every three days. Cortical Labs plans to scale the setup to 1,000 units, pending regulatory approval and energy-efficiency and safety tests.
The pitch is power. A CL1 draws about 30 watts including its life-support systems — less than a handheld calculator — against up to 700 watts for a single Nvidia H100 chip and roughly 10,200 watts for a typical eight-chip server. That is a compelling number in Singapore, where data centres consumed about 7 per cent of electricity in 2020 and where the 2019 building pause reshaped how new capacity gets approved. Founder and CEO Hon Weng Chong argues Singapore's fibre connectivity makes it a natural data centre hub, but its electricity and water constraints are exactly what biological computing sidesteps.
Biological computers are not trying to replace silicon for everything. Chong is candid that traditional chips remain superior for the fast, precise, repeatable calculations behind large language models. His argument is that living neurons excel where training data is scarce and conditions are unpredictable — teaching humanoid robots to navigate real-world spaces, or spotting cybersecurity anomalies without massive datasets. The Melbourne facility already runs 120 CL1s with about 20 paying customers in robotics and gaming research, at roughly US$2,200 a month per unit, about half the cost of renting a high-end AI chip.
Why it matters for Singapore: This is a classic frontier bet — turning a constraint into a first-mover advantage. NUS neuroscience professor Rickie Patani is using the facility to identify which neurons and support cells work best together, building a scientific case for manufacturing these cell types at scale. Whether biological computing becomes mainstream infrastructure or stays a niche research curiosity, Singapore now has a live testbed to find out, and a head start on the skills and manpower questions that commercialising it would raise.
