No chips, but living cells: biocomputing in Milan

While Singapore has unveiled the world’s first data centre rack powered by living neurons, Reply and the University of Milan are also among the pioneers exploring the possibilities of biological computing. The underlying concept is radically different from that of traditional computers: the computer of the future could consist of living cells.

Biological computing is a form of computing that uses living biological cells, such as human neurons, to process information. Instead of relying solely on transistors and chips, a biological computer system uses neurons that receive, process and transmit electrical signals. The cells can also learn and adapt, allowing them to perform certain tasks in different ways and potentially with far greater energy efficiency than traditional computers.

Since early 2026, AI specialist Enrico Bersano Begey of Reply has been working closely with scientists, including Dr Linda Ottoboni, to explore this technology. Ottoboni is a researcher in the Department of Pathophysiology and Transplantation at the University of Milan and is affiliated with the Fondazione IRCCS Ca’ Granda Policlinico Hospital and the Associazione Centro Dino Ferrari. The research uses the CL1 system developed by Australian company Cortical Labs, one of the first commercially available systems of its kind.

Two research tracks: from Pong to drug development

The research follows two main directions. The first focuses on computation. Here, Begey and his colleagues are attempting to replicate the famous DishBrain experiments, in which neurons grown in a petri dish are trained to play the classic video game Pong. The ultimate ambition? To discover a “natural” form of intelligence that can automate processes as an alternative to today’s energy-intensive AI.

“The main advantage we are aiming for is efficiency,” Begey explains. “While a traditional AI setup using GPUs consumes hundreds of watts, an organoid—a three-dimensional tissue culture—may need only a few microwatts.”

The second research track focuses on translational medicine and is led by Linda Ottoboni. She studies neurodegenerative and neuromuscular disorders, including spinal muscular atrophy (SMA) in children. By stimulating cultured neurons and measuring their electrical activity, she hopes to understand how disease affects the cells’ ability to learn and respond.

“If we can demonstrate that neurons affected by a particular condition learn more slowly or differently from healthy neurons, that would provide crucial functional information that we currently cannot obtain from patients,” Ottoboni says. This could open the door to new pharmacological applications, such as testing candidate drugs directly on human neural networks and improving patient stratification for clinical trials.

The challenges: resolution, reproducibility and lifespan

The researchers still face a number of technological and biological obstacles on the road towards a fully fledged “biological computer”. Begey emphasises that an organoid is not a miniature brain. “It lacks the complex, regionally organised architecture of our own brains.”

Current measurement equipment also has limitations. The microelectrode arrays (MEAs) used to stimulate the cells and measure their activity still offer limited spatial resolution. At the same time, higher-density measurement systems are advancing rapidly. The limited resolution and relatively simple architecture of organoids are important reasons why the system is not yet particularly accurate. Its accuracy is expected to improve as the hardware continues to develop.

Ottoboni points to two further major challenges. “First, there is reproducibility. We can grow 96 organoids under exactly the same conditions, but they will always differ slightly from one another. Second, we struggle to interpret the data. We record enormous numbers of electrical ‘spikes’, but we do not yet have a reliable method for decoding these signals. What exactly are the neurons trying to tell us?”

Then there is the practical reality: organoids do not live forever, although researchers can currently keep them alive for up to several years. When they die, all the knowledge and “training” they have accumulated is lost. This makes the technology expensive and logistically challenging for the time being.

Looking ahead

The researchers are optimistic, but a biological supercomputer remains a long way off. “A biological computer capable of carrying out mathematical calculations for us is still a decade away,” Ottoboni says.

For now, the focus is on developing advanced assembloids. These are created by combining different brain structures to investigate whether electrical stimulation can improve communication between them. The researchers expect the most concrete near-term applications to emerge in this area.

“Developing experimental setups that give us insights we do not currently have—that is the real value in the short term,” Ottoboni concludes. “Hopefully, that is something we can achieve within the next four to five years.”