When Cortical Labs shipped the first CL1 units in 2025, it wasn’t just a product launch — it was the opening of a new chapter in what computation can be. A box containing up to 800,000 living human neurons, interfaced with silicon, learning in real time, drawing just 30 watts. That’s not a GPU. That’s not even close.
The question that followed wasn’t can biology compute. The answer to that has been yes for three billion years. The question was: what do you actually do with it once you can buy one?
At $35,000 per unit, the CL1 sits firmly in research territory — not consumer electronics, not enterprise IT. The buyers are labs, universities, and biotech companies willing to work at the frontier. And what they’re building with living neurons is more varied — and more consequential — than most people expect.
Drug Discovery Is the Killer App — And It’s Already Running
The dominant use case for the CL1 in 2026 is neuroactive compound screening: testing how real human neurons respond to drugs that act on the brain. Teams in Singapore, Barcelona, and Melbourne are already running assays on the platform.
The pitch is straightforward. Animal models miss human-specific neurobiology. Silicon simulations can’t replicate the electrochemical complexity of a living synapse. The CL1 puts actual human cells — not mouse neurons, not a mathematical proxy — at the center of the experiment. That matters enormously when you’re trying to understand why a promising Alzheimer’s drug worked in rats and failed in Phase III.
Researchers are focusing heavily on conditions like epilepsy, Alzheimer’s, and psychiatric disorders — exactly the categories where clinical trial failure rates are highest and where animal-to-human translation breaks down most often. Compressing weeks of testing into days isn’t marketing copy. It’s a structural advantage that comes from having the relevant biology in the loop from day one.
Disease Modeling Has Gone Personal
Beyond screening libraries of compounds, the CL1 is enabling a subtler kind of science: patient-derived disease modeling. Scientists grow neurons from a specific patient’s stem cells, culture them on the chip, and watch how that individual’s biology behaves under disease conditions.
This is personalized medicine at the wetware level. A neural culture modeling epileptic firing patterns can be used to test which compound restores normal function — not for a theoretical average patient, but for the person whose cells are in the dish. Early work has already shown the system restoring learning function in epilepsy-model cultures. That’s not a simulation of recovery. That’s recovery, in living tissue.
The implications for neurological conditions with high individual variability — treatment-resistant depression, rare pediatric epilepsies, early-onset Alzheimer’s — are significant. The platform doesn’t just model the disease. It models your disease.
Neuroscience Finally Has a Live Platform
Academic neuroscience has spent decades studying the brain through proxies: slices, simulations, animal models, fMRI snapshots. The CL1 offers something different — a living neural culture that can be observed, stimulated, and interrogated continuously, in real time.
Universities and research groups using the Cortical Cloud (Cortical Labs’ Wetware-as-a-Service platform) are running experiments on fundamental questions: How do neurons form memories? What does plasticity actually look like at the network level? How does a culture learn a task, and what changes when it does?
The Melbourne facility runs 120 units. Singapore is scaling toward 1,000 units in phases, beginning with validation at the National University of Singapore’s Yong Loo Lin School of Medicine. That’s not a pilot. That’s infrastructure — the early skeleton of a biological cloud.
Robotics Researchers Are Teaching Neurons to Drive
One of the more surprising use cases: feeding simulated sensor data — camera feeds, balance signals, spatial inputs — directly into neuron cultures and watching them learn to control robotic systems.
This is embodied biological intelligence: neurons that don’t just compute, but act in the world through the feedback loop of sensing and response. What makes it interesting isn’t just novelty. Traditional deep learning requires enormous datasets and compute budgets to learn basic navigation. Biological neurons, properly interfaced, can adapt to new tasks with far less data and at a fraction of the power cost.
The research here is early. But the direction is clear — toward robotic systems with a fundamentally different kind of adaptability than anything built on silicon alone.
Hybrid AI Is the Long Game
The most speculative — and potentially most important — work happening on the CL1 involves hybrid intelligence: combining the precise, deterministic mathematics of silicon AI with the energy-efficient, adaptive learning of biological neurons.
Neither does what the other does well. GPUs are fast and exact but power-hungry and rigid. Biological neurons are flexible, efficient, and self-organizing but slow and hard to scale. Researchers exploring hybrid architectures are betting that the combination solves problems neither can solve alone — adaptive edge intelligence, low-power inference, real-time learning in unstructured environments.
Some groups have tested the CL1 for AI acceleration. Others are investigating it as an alternative computing paradigm entirely. In-Q-Tel, the CIA’s venture arm, has shown interest in the broader field. What defense agencies want from biological computing is left as an exercise for the reader.
The Fringe Is Real, But It’s Not the Story
A few inquiries have come in from music, entertainment, and art industries curious about creative collaborations with living neural networks. Someone apparently tried Bitcoin mining. These things happen at the frontier.
They’re not the story. The story is drug discovery. Disease modeling. Neuroscience infrastructure. The CL1 is a research platform first, and the research being done on it is directly tied to some of the hardest problems in medicine and computing.
Living Computation Has a Maintenance Cost Silicon Doesn’t
The CL1’s limitations are real and worth naming. Cell viability over long periods remains a challenge — neurons aren’t static components, and keeping them healthy, consistent, and performant requires biological expertise that most IT teams don’t have. Scaling to match GPU throughput for raw computation tasks isn’t the point, but it does define what the platform is and isn’t suited for.
Cortical Labs addresses part of this with the cloud offering: wetware without the wet lab. But the underlying challenge of working with living systems — their variability, their needs, their finite lifespans — doesn’t disappear. It’s managed.
The First Commercial Biocomputer Is Already Teaching Us What Biology Can Do
In 2026, the CL1 is still a research tool. That’s not a limitation — it’s a position. Every major computing paradigm started here: mainframes in university basements, GPUs in academic rendering labs, TPUs in Google Brain. The question is never whether a new platform will find its applications. It’s whether the applications it finds matter.
Living neurons screening drugs for Alzheimer’s. Patient cells modeling their own epilepsy. Biological clouds forming in Melbourne and Singapore. The CL1’s customers are not playing with a curiosity. They’re building the first evidence base for what programmable biology can actually do.
The silicon era assumed intelligence was something you built from inert materials. The biological era is discovering it was always growing.
References
- Cortical Labs. (2025). CL1 Biological Computer — Product Overview. https://corticallabs.com/cl1
- Reardon, S. (2025). World’s first commercial biological computer goes on sale. New Scientist. https://www.newscientist.com/article/2025-cortical-labs-cl1
- Herkewitz, W. (2025). Cortical Labs launches CL1, the world’s first commercial biocomputer. IEEE Spectrum. https://spectrum.ieee.org/cortical-labs-cl1-biocomputer
- Cortical Labs. (2026). Biological Data Center — Singapore Expansion Announcement. https://corticallabs.com/news/singapore
Related: What Is a Biocomputer in 2026? · Biological Data Centers: Can Wetware Solve AI’s Energy Crisis? · Organoid Intelligence: The Brain in a Dish
Feature image: AI-generated using Grok