At MWC 2025, Australian startup Cortical Labs presented the world’s first desktop computer CL1, which uses living human brain cells. The unique system maintains the viability of cells and promotes their development in the process of self-learning. The computer functions autonomously, without requiring a connection to classic computing devices. It is assumed that artificial intelligence on such platforms will not only be more energy efficient, but also intellectually advanced.
Cortical Labs was founded in Melbourne in 2019 and has since been developing a hybrid cell-silicon platform in collaboration with scientists at Monash University in Australia. The human brain is inferior to traditional silicon processors when it comes to computing tasks that require complex mathematical operations. According to recent studies, the human brain is slower than a 50-year-old processor.
However, when it comes to intuitive problem solving, human intelligence still remains unrivaled. It was only in the summer of 2022 that the $600 million, 630 m² Frontier computer system outperformed humans in intuitive calculations for the first time. This proves that biological computing systems are not yet outdone.
Cortical Labs’ first major achievement was the DishBrain, a hybrid processor based on a CMOS matrix and a colony of neurons. It was trained to play Pong by growing neurons on a grid of electrodes and then stimulating them with micro-current discharges, rewarding successful hits and “punishing” misses. This method allowed the formation of stable neural connections that lead to self-learning.
Since then, the company has significantly improved its platform and is now ready for mass production of hybrid computers. The first model, CL1, presented at MWC 2025 in Barcelona, is a bioreactor – a life support system for colonies of neural tissue growing on a silicon chip. The exact number of cells and their configuration cannot be predicted in advance, but the technical description is planned to use the term CFU (colony forming units).
The developers admit that they do not fully understand what cells their biological processor consists of and in what proportions they should be present. The company currently uses two methods of growing artificial nervous tissue: obtaining induced pluripotent stem cells (iPSCs) from the blood of rodents and humans, as well as genetic modifications. However, precise control of these processes is still unachievable, which leads to variability in results.
Despite this, the technology is already ready for commercial implementation. Cortical Labs plans to begin deliveries of CL1 in the second half of 2025. The expected cost of one computer is $35,000, which is about 2.5 times cheaper than similar ones. At the end of the year, the company will launch cloud access to the CL1 cluster, which includes four sections of 30 computers each. This will be the first experience of combining biocomputers into a cluster, and its results are still difficult to predict. Such uncertainty scares off investors, but successful sales can restore the trust of financial circles.
Both individual devices and the cloud cluster are considered by the company as an experiment. The developers do not yet have a clear idea of how exactly the system should work and in which areas it will be most effective. Ideally, biocomputers will be able to consume less energy and cope with generative models faster. For example, the power consumption of a rack of CL1 will be about 1 kW – significantly less than that of classic servers.
DishBrain neurons growing on an electrode array. Image credit: Cortical Labs
Beyond computational tasks, biological computers could become a platform for testing new drugs for neurodegenerative diseases. Testing on living people is unethical, and using colonies of nerve cells in computers could be an alternative method of research. This could be the most valuable application of the new technology, since ultimately human health is what matters most.