Cortical Labs, an Australian biotech startup, has launched what it claims is the world's first commercial biological cloud computing service, operating out of a Melbourne datacenter where the morning routine involves replenishing cerebrospinal fluid rather than checking server uptime dashboards. The service provides API access to 120 CL1 units — proprietary biological computers in which living human and rodent neurons are cultured on high-density multielectrode arrays. Users submit Python code or Jupyter Notebooks via the platform, with workloads executed on these biological neural networks (BNNs). The facility requires daily hands-on maintenance: technicians swap out the fluid medium every 24 hours as neurons deplete its oxygen and glucose, and carefully regulate the surrounding gas mixture to approximately five percent oxygen using nitrogen and carbon dioxide.
The technical lineage of the CL1 traces back to a 2022 research paper, "In vitro neurons learn and exhibit sentience when embodied in a simulated game-world," which demonstrated that BNNs could learn to play Pong through electrophysiological stimulation and feedback. Cortical Labs subsequently refined these techniques into a commercially available device and, more recently, demonstrated the machines learning to play DOOM. CEO and founder Hon Weng Chong argues that BNNs hold structural advantages over both classical silicon hardware and large language models: the ability to learn faster, generate genuinely novel ideas rather than recombining existing data, and operate at lower energy consumption. Independent validation of those claims at commercial scale doesn't yet exist.
The practical constraints of this first cloud offering are significant. Each job requires roughly one week of setup time for Cortical Labs to source appropriate cells and configure the biological environment — a stark contrast to the on-demand elasticity of hyperscale cloud providers. Most customers are expected to rent three or four CL1 units simultaneously to allow for result duplication and control groups. Chong anticipates early adopters will be scientific research labs lacking the infrastructure to run their own CL1 units in-house, along with forward-looking enterprises — he cited Australian banks making early quantum computing investments as an analogy — seeking to build organizational familiarity with the technology before it matures.
The most revealing constraint Chong named isn't the one-week setup time — it's the absence of a dedicated cell foundry. There is no biological equivalent of TSMC to supply standardized neural components at scale, which means every new workload requires bespoke cell sourcing. That's a supply chain problem, not a software one, and Chong has no near-term fix for it. Until that infrastructure exists, biological cloud computing scales the same way hand-blown glass scales: carefully, expensively, and one batch at a time.