Singapore's first biological data centre is now operational inside the National University of Singapore (NUS) Centre for Life Sciences. Unlike traditional server farms, this facility processes data using living human brain cells grown in the lab. The project, a collaboration between Australian startup Cortical Labs, NUS, and data centre operator DayOne, went live on July 16. It marks a concrete evolution from previous research prototypes, such as the experiment that taught neurons on chips to play Doom.

Add AlexTech.ai asPreferred Source on Google

20 CL1 units and access costs

The facility currently comprises 20 biological computer units (CL1s), the world's first commercial, code-deployable biological computers. Each unit houses at least 200,000 neurons derived from reprogrammed stem cells that exchange electrical signals with a silicon chip. Access costs US$2,200 per month, roughly half the US$4,300 monthly fee charged by major cloud platforms for high-end AI chips like the H100. Plans are in place to expand up to 1,000 units, subject to regulatory approval and safety testing.

CL1 wetware computer plays Doom as its living brain cells form data ... — https://www.notebookcheck.net/CL1-wetware-computer-plays-Doom-as-its-living-brain-cells-form-data-centers-that-sip-power-unlike-Nvidia-GPUs.1246944.0.html

Energy efficiency: 30 watts vs. silicon gigawatts

The primary competitive advantage lies in energy consumption. Each CL1 unit uses 30 watts, less power than a handheld calculator, even when accounting for the life-support systems that feed cells with sugar and oxygen every three days. By comparison, a single Nvidia H100 chip can consume up to 700 watts. This difference is critical in Singapore, where traditional data centres have already forced the government to pause new facility construction in 2019 due to electricity and water demands.

Cortical Labs CL1：第一台有人類神經元的計算機 — https://tecnobits.com/zh-TW/cl1-第一台帶有人類神經元的生物計算機/

Limits and specific applications

Chong Hon Weng, CEO of Cortical Labs, clarifies that silicon remains superior for fast, repeatable calculations underpinning large language models. Biological data centres are instead suited for scenarios with limited datasets and unpredictable conditions, such as humanoid robot navigation in dynamic environments or cybersecurity anomaly detection. The approach leverages the cells' ability to learn from few examples, similar to human cognition. The NUS site will also serve as a lab to define manpower requirements and skills needed for large-scale commercialization.