Singapore Opens World's Second Living Brain-Cell Data Centre Inside NUS

A data centre unlike any other has quietly gone live inside the National University of Singapore's Centre for Life Sciences — one that needs to be fed every three days. Rather than silicon chips, it runs on living human neurons, making it the second commercial biological computing facility in the world and Singapore's first. Australian biotech startup Cortical Labs, which built and operates the system, partnered with NUS and data centre operator DayOne to launch the facility on July 16, 2026.

The system currently houses 20 units of Cortical Labs' CL1 biological computers, each containing at least 200,000 lab-grown neurons mounted on electrode-fitted silicon chips. Those neurons — derived from human blood cells reprogrammed into stem cells — exchange electrical signals with conventional computers, translating their neural firing patterns into usable computing output. Every 72 hours, lab technicians deliver a life-support cocktail of sugar, micronutrients, and pH buffers, while a gas mixer continuously supplies carbon dioxide, oxygen, and nitrogen. The plan is to eventually scale the NUS site to 1,000 CL1 units, pending regulatory approval and safety testing.

The energy math is striking. Each CL1 unit, life-support systems included, consumes just 30 watts — less than a handheld calculator. By contrast, Nvidia's H100 SXM GPU can draw up to 700 watts under load, and a standard eight-GPU server can pull around 10,200 watts total. For Singapore, where data centres already consumed 7% of national electricity by 2020 — prompting the government to pause new facility construction in 2019 — the biological approach could offer a meaningful pressure valve on energy and water demand.

Cortical Labs CEO Chong Hon Weng is clear-eyed about the technology's limits: it is not a replacement for the fast, precise, repeatable computation that powers large language models like ChatGPT. Instead, biological computing targets scenarios where training data is scarce and conditions are unpredictable — humanoid robotics navigating real-world environments, for instance, or cybersecurity anomaly detection without massive labeled datasets. The company draws an analogy to human cognition: people can generalize from just a few examples in ways that current AI cannot without enormous datasets.

The Melbourne facility, the world's first commercial biological data centre, currently runs 120 CL1s with roughly 20 paying customers — corporate R&D teams and universities — experimenting in robotics and gaming. Access costs US$2,200 per CL1 per month, roughly half the ~US$4,300 monthly fee major cloud platforms charge to rent a single high-end AI training chip. The Singapore site will serve as a research and development base to determine optimal neuron compositions, workforce requirements, and scaling pathways for broader commercial deployment across Asia-Pacific.