When you think of a data centre, you probably picture rows of servers, cables, cooling systems and thousands of processors working away behind the scenes. Living brain cells probably aren't the first thing that comes to mind. But researchers in Singapore are now experimenting with using living neurons as part of the hardware that processes information.
In August 2026, NUS Medicine, DayOne and Cortical Labs unveiled a biological data centre prototype that combines living neurons with silicon chips. The idea is to develop computing systems that can learn and adapt while potentially using less energy than conventional hardware.
It's an interesting possibility, especially as AI's computing demands continue to grow. But how does biological computing actually work, and how close are we to using it outside a research lab?
What Is a Biological Data Center?
A biological data centre is an experimental computing facility that uses living biological cells alongside electronic hardware to process information. Instead of relying entirely on silicon processors, these systems use networks of living neurons that can receive electrical signals, respond to them and change their behaviour over time.
This approach is known as wetware computing, because it uses living biological material as part of the computing system. It's quite different from the artificial neural networks used in conventional AI. Those networks are mathematical models that imitate certain aspects of how the brain processes information.
Biological computing uses actual neurons, although it still needs conventional hardware and software to operate. The technology is also different from neuromorphic computing, which uses electronic chips designed to imitate how biological neurons work. For now, biological data centres are research facilities rather than replacements for conventional data centres.
But researchers are beginning to investigate what happens when these living computing systems are brought together.
How Does Computing With Living Neurons Work?
The basic idea is to connect living neurons to a computer so that information can pass between them. Cortical Labs has developed a biological computer called the CL1. It uses neurons grown from stem cells and placed on a silicon chip containing tiny electrodes. These electrodes send electrical signals to the neurons and record how they respond.
The computer then uses those responses to change what happens in a digital environment. The neurons receive new signals based on those changes, creating a continuous feedback loop. This is called closed-loop computing. The neurons aren't following programmed instructions like a normal processor.
Instead, researchers study how their activity changes as they receive information and feedback. A 2025 technical publication in Nature Reviews Bioengineering describes how the CL1 combines this process with a built-in system that keeps the cells alive. And researchers have already demonstrated some interesting results.
In a 2022 study published in Neuron, Cortical Labs researchers connected living neurons to a simulated game of Pong. The neurons received electrical feedback about the game's environment, and the researchers observed signs of learning within five minutes under certain experimental conditions.
It wasn't the same as teaching a computer to play a game using conventional AI. But it demonstrated that living neural networks could respond to information and adapt their activity during a computing task.
Inside Singapore's Biological Data Center Prototype
Singapore's biological data centre prototype takes that research beyond individual experiments. According to NUS Medicine's August 2026 announcement, the project brings together 20 Cortical Labs CL1 units in a research environment at the NUS Life Sciences Institute. Each partner has a different role.
NUS Medicine provides the neuroscience expertise needed to grow and study the neurons. Cortical Labs supplies the biological computing technology, while DayOne provides the infrastructure design and operational support.
The project is intended to investigate how biological computing could support areas such as AI research, drug discovery and neurological disease research. There's also a bigger infrastructure ambition behind it. The partners want to understand whether these systems could eventually operate at a larger scale, including within commercial data-centre environments.
But the prototype hasn't demonstrated that biological computers can replace conventional AI servers. What it does show is that multiple biological computing units can be operated together in a dedicated research setting. The next challenge is finding out whether they can deliver useful computing performance beyond controlled experiments.
Could Biological Computing Make AI More Energy-Efficient?
One of the biggest arguments for biological computing is its potential to use less electricity. Living neurons process information through biological activity rather than relying entirely on electronic circuits. Cortical Labs believes this could offer a more energy-efficient way to perform certain computing tasks.
And there's a good reason researchers are interested in alternatives. The International Energy Agency's 2026 report Key Questions on Energy and AI estimates that data centres consumed 485 terawatt-hours of electricity in 2025. It expects that figure to reach around 950 terawatt-hours by 2030.
However, lower electricity consumption doesn't automatically mean better computing efficiency. A biological computer might use less power than a conventional processor, but we'd also need to know how much useful work it completes. Its supporting equipment, including the systems that keep neurons alive, must be included in that calculation.
Researchers still need reliable comparisons involving equivalent tasks, performance, energy use and operating costs. Until those measurements are available, biological computing's energy advantages remain a possibility rather than a proven solution for sustainable AI infrastructure.
What Could Biological Data Centres Actually Be Used For?
Biological computing has several possible applications, although some are much closer to practical use than others. Neuroscience is an obvious starting point. Researchers can use living neural networks to study how brain cells respond to information, learn and adapt.
They can also investigate how different substances affect neural activity, which could support research into neurological conditions and drug development. There are also possibilities in robotics and AI research. Biological systems can change their behaviour in response to feedback, so researchers are interested in whether they could help develop more adaptable computing systems.
Some of these ideas overlap with organoid intelligence, which investigates computing using lab-grown, brain-like structures. However, using biological computers for mainstream AI training or inference is a much bigger challenge. There's currently no established evidence that they can compete with conventional processors across those workloads.
Their first useful applications may therefore come from helping researchers understand biological intelligence, rather than trying to replace the hardware already running AI systems.
What's Stopping Biological Data Centres From Scaling?
Getting living neurons to process information is one thing. Building infrastructure that organisations can depend on is another. Biological computers introduce challenges that conventional servers don't have, particularly when it comes to keeping the hardware working and producing consistent results.
Keeping living hardware alive
Unlike silicon processors, neurons need the right conditions to survive. They require nutrients, controlled temperatures and a suitable environment. Cortical Labs has developed life-support technology for the CL1, but maintaining living cells still introduces additional operational requirements.
Operators would need to manage cell health, contamination risks and the eventual replacement of biological cultures. And unlike replacing a faulty electronic component, replacing living neurons may also affect how the computing system behaves.
Making performance reliable and repeatable
Living neural networks don't necessarily develop or respond in exactly the same way. That creates problems when you're trying to build computing infrastructure that delivers predictable results. Researchers need to establish whether different neural cultures can perform the same tasks consistently, how their performance changes over time and what happens when a system fails.
They also need benchmarks that allow meaningful comparisons with conventional computing. Without that evidence, it's difficult to judge which workloads biological computers are suited to or whether they could operate reliably at a larger scale.
Managing security and ethical questions
Biological computing also raises questions about how these systems should be managed. Software interfaces and remote access still need cybersecurity controls, particularly when systems handle sensitive research information. There are additional ethical considerations because the hardware contains living cells.
These include how biological material is obtained, whether appropriate consent has been given and what responsibilities researchers have towards the systems they create. Scientists are also examining questions around consciousness and biological intelligence.
However, there's no established evidence that the neural cultures used in these computers have conscious experiences. These questions will need clearer answers as biological computing moves towards more widespread use.
Final Thoughts: Biological Computing Still Has a Lot to Prove
Singapore's biological data centre prototype offers an interesting glimpse of what computing could look like beyond conventional silicon processors. Researchers have demonstrated that living neurons can participate in computing tasks, and they're beginning to investigate how these systems might work together.
But there's still a considerable gap between experimental biological computing and infrastructure that organisations can use reliably. Perhaps the more interesting question isn't whether living neurons will replace GPUs. It's whether they can eventually perform certain kinds of work that conventional computers struggle to reproduce.
Finding those applications will take more research, better benchmarks and a clearer understanding of the technology's limitations. And as emerging technologies move from laboratory experiments towards practical applications, EM360Tech continues to examine what those developments could mean for the future of enterprise technology.
Comments ( 0 )