The 20-unit system reportedly uses lab-grown brain cells integrated with silicon chips to generate computing power
Singapore has launched the world’s first independently operated data center using living human brain cells for some computing tasks, the medical school of the National University of Singapore (NUS Medicine) has said. The system consumes a fraction of the power of traditional computing hardware. The neurons are fed every three days and sustained by dedicated life-support equipment.
NUS Medicine announced the prototype earlier this month in partnership with Singapore-based data center operator DayOne and Australian biotech firm Cortical Labs. Housed at the university, the system is described by its developers as the world’s first independently operated biologically integrated server rack.
The facility comprises 20 CL1 biological computers, each containing at least 200,000 lab-grown human neurons on an electrode-fitted silicon chip. Derived from blood cells reprogrammed into stem cells, the neurons exchange electrical signals with conventional hardware, turning their activity into computing power.
Cortical Labs founder and CEO Hon Weng Chong said the prototype moves biological computing “from research to commercial application,” arguing it could complement AI where “data is sparse” and conditions change. He pointed to potential uses in drug discovery, humanoid robotics, cybersecurity and fraud detection.
Unlike conventional data centers packed with power-hungry servers and cooling systems, biological computers could operate on a fraction of the electricity by harnessing the natural processing capabilities of living neural networks.
Keeping the “computers” running looks more like maintaining a laboratory than a server room. Technicians feed the cells a mixture of sugar, micronutrients and pH buffers every three days, while a gas system supplies carbon dioxide, oxygen and nitrogen.
Each CL1 draws around 30 watts including its life-support equipment, according to Chong. Nvidia’s H100 SXM AI processor can consume up to 700 watts under heavy workloads, while a server fitted with eight such chips can draw around 10,200 watts including supporting hardware.
The power difference is particularly relevant for Singapore, which temporarily halted construction of new data centers in 2019 because of concerns over their electricity and water consumption. Data centers accounted for around 7% of the country’s electricity use in 2020.
The technology is not intended to replace silicon, which Chong said remains “far superior” for the fast, repeatable calculations behind large language models such as ChatGPT. Biological systems, however, can learn from fewer examples and adapt to changing conditions.
