Cortical Labs, a Melbourne-based biotech startup, has opened a cloud computing service built on 120 CL1 units — computers powered by living human and rodent neurons cultured on high-density multielectrode arrays. The service, reported by The Register on March 14, exposes an API and Jupyter Notebook interface that allows customers to upload Python code and execute it against biological neural networks, with billing handled via standard credit card. The technology traces back to a landmark 2022 paper, "In vitro neurons learn and exhibit sentience when embodied in a simulated game-world," which demonstrated neurons learning to play Pong, and was subsequently refined into the CL1 product after a high-profile demonstration of the hardware learning to play DOOM.
Operating a biological datacenter bears little resemblance to running conventional cloud infrastructure. Cortical Labs technicians begin each day by replenishing a cerebrospinal-fluid-like liquid that neurons deplete of oxygen and glucose within 24 hours, and by adjusting surrounding gas mixtures to maintain roughly five percent oxygen — optimal conditions for neuronal activity. Each customer job requires approximately a week of preparation to source the appropriate cell line and condition the biological environment, a stark contrast to the near-instant provisioning offered by hyperscalers. CEO Hon Weng Chong told The Register that most customers will rent three to four CL1 units simultaneously to enable result duplication and control groups, reflecting the experimental nature of current workloads.
Chong positions biological computing as distinct from both classical computing and large language models, arguing that neurons can genuinely learn within simulated environments and devise novel solutions rather than recombining existing information, while consuming less energy than conventional datacenters. The company's target early adopters are scientific research labs without the resources to operate their own CL1 units, and forward-looking enterprises seeking to build institutional knowledge in the technology before it matures — analogous, Chong told The Register, to banks making early exploratory investments in quantum computing. A critical supply-chain bottleneck remains: the industry lacks a high-volume cell foundry equivalent to TSMC, limiting how broadly accessible biological computing can become in the near term.
The commercialization of the service raises questions that have no current regulatory answer. The foundational paper uses the word "sentience" — not "plasticity" or "adaptation" — to describe the behavior of the neurons, a philosophically loaded framing that now underpins a for-profit API accessible to anyone with a credit card. No existing bioethics framework governing human tissue was written with cloud compute workloads in mind, and no statutory obligation currently requires Cortical Labs to formally address welfare questions before scaling. Chong himself acknowledged the ambiguity half-jokingly to The Register, noting he is "just a little uncomfortable with giving biological computers the chance to control their own destiny" — which is why human technicians, rather than automation, still manage the daily fluid and gas maintenance. Cortical Labs is operating on a genuinely different substrate for computation, one whose practical near-term applications remain speculative but whose ethical and regulatory implications are already outpacing the institutional frameworks meant to govern them.