A Startup’s Radical Bet: Data Centers Powered by Living Neurons?

A biotechnology startup is attempting to rethink the future of computing by building machines powered not by silicon chips but by living human neurons — an approach that its developers say could drastically reduce the energy demands of artificial intelligence infrastructure.

The Rise of Biological Computing

In the global race to expand computing power for artificial intelligence, most technology companies are investing billions in increasingly powerful silicon chips and energy-hungry data centers. But one biotechnology startup is pursuing a markedly different path: building computers from living human brain cells.

Researchers at the Australian biotech company Cortical Labs unveiled what they described as the world’s first code-deployable biological computer, known as the CL1. Unlike conventional processors made from silicon wafers, the system is composed of roughly 200,000 living human neurons derived from blood stem cells.

The concept represents a new category of computing sometimes referred to by the company as “wetware,” blending biological systems with electronic hardware. In this arrangement, computers send electrical signals to living neurons, while embedded chips record the neurons’ responses as computational output.

For decades, scientists have explored ways to harness biological processes for computing tasks. But the emergence of artificial intelligence — with its enormous appetite for processing power — has renewed interest in alternatives to traditional chip-based systems.

Teaching Neurons to Compute

Last year, Cortical Labs demonstrated the capabilities of its biological computing platform by showing how the neurons could be trained to interact with the video game Doom. The demonstration followed earlier experiments in which similar neural systems had been trained to play Pong in 2022.

The Doom experiment, according to researchers involved in the project, represented a more complex proof of concept. By sending electrical signals to clusters of neurons and monitoring their responses, scientists could effectively train the biological network to respond to stimuli in ways that resembled learning.

The experiments illustrate the unusual nature of biological computing. Instead of relying solely on programmed algorithms executed by silicon processors, these systems interact with living cells whose responses can evolve through exposure to stimuli.

While the practical applications of the technology remain uncertain, the demonstrations are intended to show that biological neurons can perform tasks that resemble computational processes.

A New Kind of Data Center

Cortical Labs is now preparing to scale its technology beyond laboratory experiments. According to company officials, the startup is working with DayOne Data Centers to develop facilities designed specifically to host biological computing systems. The first installations are planned in Melbourne, Australia, and Singapore.

The Melbourne facility is expected to house approximately 120 CL1 units, while the Singapore site could eventually deploy as many as 1,000 units. Unlike conventional AI data centers packed with racks of graphics processing units, these facilities would contain arrays of biological computing nodes built around living neural cells.

The company says the biological systems require only minimal power compared with conventional processors. According to the company’s chief executive, Hon Weng Chong, each CL1 node uses less electricity than a handheld calculator — orders of magnitude lower than the energy requirements of modern AI chips.

The Environmental Question

The effort comes at a time when the rapid expansion of AI infrastructure is raising concerns about the environmental costs of data centers.

Large AI facilities consume vast amounts of electricity, often comparable to the energy demand of small cities. They also require substantial water resources for cooling systems and can produce significant noise, factors that have drawn increasing scrutiny from regulators and local communities.

Researchers say biological computing could, in theory, offer a partial response to these challenges. By using living neurons instead of silicon processors, the systems may require only a fraction of the power used by conventional AI processors.

Still, key questions remain about the technology’s practical applications. It is not yet clear what tasks biological computing systems will ultimately perform or whether they can match the performance of the most advanced data center chips.

For now, the technology represents an experimental alternative in a field dominated by silicon — an attempt to rethink what a computer might look like in an era defined by the growing demands of artificial intelligence.