Singapore has unveiled the world’s first independently run data centre that integrates living human neurons within a server rack setting.

The Yong Loo Lin School of Medicine at the National University of Singapore (NUS) created the prototype supercomputer, which is made up of 20 CL1 biological computing units.

Overall, each CL1 computing unit contains at least 200,000 lab-grown human neurons on an electrode-fitted silicon chip, and it is the world’s first independently operated biological server track.

As previously noted, the data centre setup consists of 20 CL1 units, all housed within a single server rack at the NUS Life Sciences Institute.

Each module holds about 800,000 lab-cultivated human neurons, giving the full rack an estimated total of 16 million neurons.

These neurons function as the system’s biological processing units. The system became operational on July 16 and was officially introduced on August 17, following an August 6 demonstration attended by more than 80 guests.

It is being promoted as a potential new paradigm for AI and computing.

The system’s underlying technology differs fundamentally from that of a traditional server, even though it still relies extensively on silicon-based hardware.

Cortical Labs cultivates neurons derived from human stem cells and integrates them into a silicon platform equipped with microelectrode arrays.

Technicians at NUS provide nutrients to the cell cultures every three days, while the CL1 device maintains and controls their surrounding conditions. According to Cortical Labs, these neurons can stay viable for up to six months.

The importance of this project stems from its effort to move biological computing from mere lab experiments toward practical computing infrastructure.

Notably, the system also delivers oxygen, carbon dioxide, and nitrogen to sustain the neurons. According to Cortical Labs, such systems will be valuable in situations where there is insufficient data to train conventional AI models.

They may be especially beneficial for training humanoid robots and for applications in cybersecurity and fraud detection, where advanced computers must learn from human-generated data and adapt rapidly.