Singapore Launches a Biological Data Center Powered by Living Human Neurons

The line between biology and computing is becoming increasingly difficult to define.

In Singapore, researchers and technology companies have launched an experimental biological data center that uses living human neurons as part of its computing infrastructure. The project brings together NUS Medicine, Singapore-based data center operator DayOne, and Australian biotechnology company Cortical Labs.

At the center of the project is a rack containing 20 CL1 biological computers developed by Cortical Labs. Unlike conventional computers that depend entirely on silicon processors, these systems combine traditional electronics with networks of living human neurons grown from stem cells.

The result is a new type of computing platform often described as biological computing, or “wetware.”

While it is far too early to suggest that biological computers are about to replace GPUs, CPUs, or traditional AI infrastructure, the Singapore project represents an important step toward taking biological computing out of the laboratory and putting it into a real operating environment.

What Has Been Built in Singapore?

The Biological Data Centre prototype was officially unveiled by NUS Medicine, DayOne, and Cortical Labs in August 2026.

The system consists of 20 CL1 biological computing units installed together in a server rack. According to the organizations behind the project, it is the first independently operated biologically integrated server rack of its kind.

Each CL1 contains a network of living neurons that grows directly on a specially designed silicon chip.

These neurons are not simply stored inside the machine. They are electrically connected to the computer.

The hardware can send signals to the neural network and record the electrical responses produced by the cells. Software then creates an environment in which the neurons can interact with digital information.

In other words, the biological component is actively participating in the computing process.

Cortical Labs describes the CL1 as a system where silicon and biological neural networks operate together rather than as two completely separate technologies.

Around 800,000 Living Neurons Per Computer

Each CL1 system can contain approximately 800,000 neurons.

With 20 machines operating in the Singapore installation, the complete rack could therefore involve roughly 16 million living neurons.

The cells are derived from human stem cells and are maintained in a controlled environment that supplies the nutrients and conditions necessary for them to remain alive.

According to Cortical Labs, the CL1 includes its own life-support infrastructure and can maintain biological neural networks for periods of up to approximately six months.

The company has also developed an operating environment called Biological Intelligence Operating System, or biOS, which allows developers and researchers to communicate with these biological networks.

Instead of training an artificial neural network that mathematically imitates certain characteristics of the brain, biological computing takes a very different approach.

It uses actual neurons.

Why Scientists Are Interested in Biological Computing

Modern artificial intelligence was heavily inspired by neuroscience.

Artificial neural networks contain interconnected mathematical units known as artificial neurons. These systems have become extraordinarily powerful, particularly with the development of large language models, computer vision systems, generative AI, and autonomous technologies.

But artificial neurons remain mathematical abstractions.

Biological neurons operate differently.

The human brain is extraordinarily efficient at learning, adapting, recognizing patterns, and processing complex information while consuming remarkably little energy compared with modern computing infrastructure.

This has created a fascinating research question:

Instead of continuously building larger artificial systems that attempt to reproduce some characteristics of biological intelligence, could living neural networks themselves become part of computing systems?

Cortical Labs has been exploring that question for several years.

The company previously demonstrated biological neurons interacting with a simplified version of the video game Pong. The experiment showed that networks of cultured neurons could receive feedback from a digital environment and modify their activity based on that feedback.

The CL1 platform attempts to transform that type of research into a programmable computing system.

The Energy Question

One of the most interesting aspects of biological computing is energy efficiency.

According to figures reported by Cortical Labs, an individual CL1 consumes approximately 25 watts, while a fully populated rack can operate at approximately 800 to 1,000 watts depending on the configuration and supporting infrastructure.

For comparison, modern high-density artificial intelligence infrastructure can require dramatically more electricity.

NVIDIA documentation indicates that a DGX GB200 NVL72 rack can consume approximately 120 kilowatts of power.

That difference immediately attracts attention, especially as the rapid expansion of artificial intelligence creates growing demand for data centers, electrical infrastructure, cooling systems, and energy generation.

However, the comparison needs an important qualification.

A biological computer rack and an NVIDIA AI rack do not currently provide equivalent computing capabilities.

The NVIDIA system is designed to train and run massive artificial intelligence models at enormous scale. Biological computing remains an emerging experimental technology whose practical performance is still being studied.

Therefore, comparing 1 kW with 120 kW does not mean that biological computing already delivers the same AI performance while using one hundredth of the electricity.

It does not.

What the comparison demonstrates is the potentially enormous difference in energy requirements between biological neural systems and today's highest-performance silicon computing infrastructure.

Whether that biological efficiency can eventually translate into commercially useful computing at scale remains one of the biggest unanswered questions.

Drug Discovery Could Be One of the First Major Applications

Interestingly, the most immediate applications may not involve replacing conventional computers at all.

Medicine and pharmaceutical research could become some of the first areas where biological computing provides significant advantages.

Researchers could potentially study how living human neural networks react to new compounds without relying exclusively on animal models.

A biological computing platform could help scientists observe changes in neural activity, learning, adaptation, or cellular responses after exposure to experimental drugs.

Because the neurons are human-derived, the resulting data may eventually provide additional insights that are difficult to obtain from traditional laboratory models.

NUS Medicine believes the technology could contribute to neurological disease research and accelerate parts of the drug discovery process.

This may ultimately prove more important in the short term than attempting to build a biological version of ChatGPT.

Robotics and Adaptive Machines

Robotics represents another potential application.

Traditional robots generally depend on software models that must be programmed or trained using substantial amounts of data.

Biological neural networks naturally adapt to changing signals.

A future robotic system could theoretically use biological computing for tasks where rapid adaptation to unfamiliar environments is valuable.

Cortical Labs has specifically identified humanoid robotics as one area it intends to explore.

The concept could eventually create hybrid systems in which conventional processors handle predictable calculations while biological neural networks deal with learning or adaptive behavior.

That remains experimental, but it demonstrates why researchers see biological computing as something that could complement traditional artificial intelligence rather than simply compete with it.

Cybersecurity and Fraud Detection

Cortical Labs has also identified cybersecurity and fraud detection as potential commercial applications.

These fields frequently involve detecting unusual behavior in environments where patterns continuously change.

Conventional AI systems can perform these tasks extremely well, but they often require significant amounts of historical data and repeated retraining.

Biological neural networks may eventually offer advantages in situations where a system needs to learn from relatively limited data or rapidly adjust as conditions change.

That possibility remains to be demonstrated at commercial scale, but it is one of the areas the Singapore installation is expected to help researchers investigate.

Biological Computing Will Not Replace Traditional AI Tomorrow

The headlines surrounding this technology can easily create the impression that servers powered by human brain cells are about to replace conventional computers.

The reality is much more complicated.

Today's semiconductor industry has decades of engineering development behind it.

Modern CPUs and GPUs are extremely reliable, reproducible, scalable, and precisely controllable. They can operate continuously for years and can be manufactured in enormous quantities.

Living biological systems introduce an entirely different set of challenges.

Cells need nutrients.

Their environment must be carefully controlled.

Biological networks change over time.

Individual cultures may behave differently.

Their useful lifespan is limited.

Maintaining millions of living neurons inside a commercial data center therefore requires technologies and procedures that traditional computing infrastructure simply does not need.

This makes biological computing both fascinating and difficult.

The Singapore project is important precisely because it begins testing what happens when the technology moves beyond isolated laboratory experiments and into something resembling real computing infrastructure.

New Ethical Questions Will Follow

As biological computing develops, technological challenges will probably not be the only issues society will need to address.

Using living human neurons inside commercial computing systems raises ethical questions that do not exist with conventional silicon processors.

At what level of biological complexity should special ethical protections apply?

Could increasingly sophisticated networks of neurons eventually develop forms of activity that researchers need to treat differently from ordinary cell cultures?

How should biological computing systems be regulated?

Who owns neural material derived from human cells?

These questions may seem premature when discussing systems containing hundreds of thousands of cultured neurons, but the technology is likely to become more sophisticated.

The ethical framework surrounding biological computing will therefore need to evolve alongside the technology.

A Different Direction for the Future of AI

For decades, progress in computing has largely followed one direction: build faster processors, add more transistors, connect more machines, collect more data, and increase computing power.

Artificial intelligence accelerated that trend dramatically.

The world's largest technology companies are now building enormous AI data centers requiring gigawatts of electrical capacity and billions of dollars of infrastructure.

Biological computing proposes a completely different possibility.

Instead of using increasingly powerful machines to imitate biological intelligence, engineers may eventually combine machines directly with biological intelligence.

It is unlikely that one technology will completely replace the other.

A more realistic future may involve hybrid computing architectures.

Traditional processors could continue handling enormous deterministic workloads.

GPUs could remain responsible for large-scale AI training and inference.

Biological neural networks could potentially be introduced for specialized tasks involving adaptation, biological modeling, learning from limited information, or extremely energy-efficient processing.

That future is still speculative.

But the 20-unit biological computing rack now operating in Singapore demonstrates that the idea is beginning to move from theoretical research toward real infrastructure.

The question is no longer whether living neurons can interact with computers.

Researchers have already demonstrated that they can.

The more important question is what happens when biological computing begins to scale.

And if the technology succeeds, the next major revolution in computing may not come from a smaller transistor or a faster GPU.

It may come from combining silicon with living cells.

Sources

National University of Singapore, Yong Loo Lin School of Medicine — “NUS Medicine, DayOne and Cortical Labs Unveil Biological Data Center Prototype in Singapore,” August 17, 2026. NUS Medicine source

National University of Singapore, Yong Loo Lin School of Medicine — “Biological Data Centre prototype established at NUS Medicine.” NUS Medicine background source

Cortical Labs — CL1 biological computer product and technology information. Cortical Labs CL1

Cortical Labs — Biological computing and company technology overview. Cortical Labs

The Next Web — “A data centre rack running on living neurons is now operating in Singapore,” August 20, 2026. The Next Web source

NVIDIA — DGX GB Rack Scale Systems documentation, including approximate 120 kW rack power consumption. NVIDIA documentation