The launch of the world’s first biological data center relying on human neural cells marks the beginning of the hybrid living computing era to break the thermal and electrical limits of traditional silicon chips.
Article index:
- Launch of the first-of-its-kind biological computing model
- How live neural cells perform computational processes
- Energy consumption efficiency compared to traditional GPUs
- Research horizons and economic and operational challenges for expansion
- Frequently asked questions
Launch of the first-of-its-kind biological computing model
The National University of Singapore has unveiled what it described as the world’s first independently operating biological data center prototype; an advanced hybrid system relying on 16 million live human neurons cultivated and grown in the lab from stem cells to execute computational tasks alongside classical silicon hardware. This innovative model was officially launched on July 16 at the university’s Life Sciences Center, before being officially showcased to the public and experts on August 6.
The system consists of 20 “CL1” biological computing units developed by Cortical Labs, headquartered in Melbourne, Australia. Each computing unit houses approximately 800,000 human cortical neurons fixed and distributed over silicon-based microelectrode arrays, where these arrays electrically stimulate the neural networks and continuously record their activity and electrical impulses in real time.
How live neural cells perform computational processes
This project is the result of close collaboration between the Yong Loo Lin School of Medicine at the National University of Singapore, which oversees the cultivation and biological care of the neural cells; DayOne, the Singaporean data center operator providing infrastructure engineering expertise; and Cortical Labs, which supplied the system with biological hardware and the dedicated “BOS” operating system.
The live neural cells are grown under extremely precise laboratory conditions requiring continuous feeding every three days with a liquid solution rich in glucose, micronutrients, and controlled pH levels. A dedicated life support system pumps a precise gas mixture of carbon dioxide, oxygen, and nitrogen at strictly regulated temperatures. The estimated operational lifespan of these cells is approximately six months before new neural clusters must be substituted.
Energy consumption efficiency compared to traditional GPUs
Each “CL1” biological computing unit consumes between 25 and 30 watts only, including the power required to operate the life support system. When assembling a full server rack with a capacity of 20 units, the total electricity consumption ranges between 800 and 1,000 watts only, which represents an extremely minor fraction of the massive power demanded by traditional GPUs; NVIDIA’s single H100 card alone consumes over 700 watts under full operational load.
Research horizons and economic and operational challenges for expansion
This achievement builds upon Cortical Labs’ previous laboratory experiments, most notably the famous “DishBrain” experiment which proved the ability of on-chip cultured neurons to learn to play the digital game “Pong.” The Singapore facility marks the company’s first location outside Australia and a turning point from laboratory experiments toward practical infrastructure. Targeted research applications include developing brain-inspired artificial intelligence, biomedical modeling, drug discovery, and neurological disease research. DayOne, whose valuation surged to $20 billion after raising $4.5 billion in June, is studying the expansion of the facility to house 1,000 biological units, while the most prominent question remains regarding the long-term feasibility and commercial reliability of running live cells.
Frequently asked questions
Question: How does the biological data center launched by Singapore work?
Answer: It operates via 16 million live human neural cells cultured on silicon chips and electrodes to electrically stimulate them and record their signals for data processing.
Question: What is the energy consumption of the biological data center?
Answer: A full rack comprising 20 units consumes between 800 and 1,000 watts only, which is a very minimal rate compared to traditional GPUs.
Question: What is the operational lifespan of the neural cells and how are they cared for?
Answer: The cells live for about 6 months and require continuous feeding every 3 days with sugar solutions and nutrients along with a life support system to regulate gases and temperature.
