Biocomputing platforms, compared honestly
Two biocomputing platforms are realistically accessible today: the Cortical Labs CL1, a local desktop unit, and the FinalSpark Neuroplatform, a remote bio-cloud. They differ less in raw capability than in how you reach the tissue and what you must maintain yourself.
Spec sheets for living-tissue systems invite the wrong comparison. There is no clock speed, no cache, no clean transistor count. The numbers that matter are biological and electrophysiological, and most of them drift as a culture ages. The table below tracks the figures against vendor documentation; treat every value as a starting point to confirm, not a guarantee, because these platforms change faster than any datasheet.
The platforms side by side
| Cost | Electrode interface | Rack draw | Status | |||
|---|---|---|---|---|---|---|
| Cortical Labs CL1 | Local hardware | Inquire | ~800,000 | Planar CMOS MEA, ~1,024 channels | ~850-1000 W | Commercial, 2026 |
| FinalSpark Neuroplatform | Remote node | Subscription | multi-organoid | 3D organoid MEAs, few channels per chamber | Hosted | Open research access |
Power and neuron figures are operational estimates and vary with culture age and configuration. Confirm against current vendor documentation before relying on any value.
The metrics that actually matter
Forget gigahertz. The first real metric is how many neurons are active and addressable, and how they are arranged. A planar array such as the CL1 puts cells on a dense, two-dimensional electrode grid, which buys precise single-cell targeting at the cost of the three-dimensional connectivity that real tissue depends on. A three-dimensional organoid, as on the Neuroplatform, recovers that native architecture and far higher synaptic density, but its interior is electrically shielded from the surface electrodes, so most of the network is seen only indirectly.
The second metric is the electrode interface. Research-grade high-density arrays such as the Maxwell Biosystems MaxOne reach into the tens of thousands of electrodes for fine spatial mapping;10 microfluidic chambers instead offer a handful of channels per well but many wells in parallel, which suits screening more than precision. Neither number means much without the third metric, signal quality: usable spike amplitudes are on the order of tens of microvolts against a noise floor of a few microvolts, and channels drift out of that range as electrodes foul or cells die. A platform is only as good as the fraction of its electrodes still reporting clean signal this week.
Local hardware or remote bio-cloud?
The real procurement decision is not which platform is faster. It is whether you want to run a cell-culture laboratory. A local CL1 gives you the lowest possible latency, which matters for sub-millisecond closed-loop control, but it obliges you to keep an incubator, a carbon dioxide supply, sterile media handling, and a trained technician running indefinitely. A remote node such as the Neuroplatform turns all of that into someone else's problem and exposes the tissue through a network API, at the price of latency and of never quite owning your experiment.
If your work needs tight real-time closed-loop control and you can staff a culture lab, buy local. If you are exploring, screening, or validating ideas and would rather not own an incubator, rent remote. Most groups starting out should rent first; the biology is harder than the code.
How platforms keep the substrate trustworthy
Because every culture is different and degrades over time, a credible platform runs daily electrophysiological baselines: signal-to-noise per channel, electrode impedance, and cross-correlation across electrodes to confirm the network still bursts and synchronizes like healthy tissue. Channels that fail are dropped from the active routing matrix, and cultures that fall below baseline are retired. When you read a platform's neuron count, the number you should want is not the total seeded but the fraction still passing this QC today.
Frequently asked questions
Which biocomputing platforms can I actually use right now?
The Cortical Labs CL1, a local desktop unit available commercially in 2026, and the FinalSpark Neuroplatform, a remotely accessible bio-cloud for researchers.
Is local hardware or a remote node better?
Local hardware gives the lowest latency for real-time closed-loop control but requires running a cell-culture lab. A remote node removes that burden at the cost of network latency. For most teams starting out, remote is the safer first step.
What is the single most important spec?
The fraction of electrodes still reporting clean signal on a healthy, bursting culture today, not the total number of neurons seeded or electrodes fabricated.
Why do the numbers in vendor sheets keep changing?
Because living tissue ages and varies. Neuron counts, usable channels, and signal quality all drift with culture age and configuration, so any figure is a snapshot to confirm, not a fixed datasheet value.
References
- Kagan BJ, et al. In vitro neurons learn and exhibit sentience when embodied in a simulated game-world. Neuron. 2022;110(23):3952-3969. doi:10.1016/j.neuron.2022.09.001. Accessed 2026-06-12.
- Jordan FD, et al. Open and remotely accessible Neuroplatform for research in wetware computing. Frontiers in Artificial Intelligence. 2024;7:1376042. doi:10.3389/frai.2024.1376042. Accessed 2026-06-12.
- Mueller J, et al. High-resolution CMOS MEA platform to study neurons at subcellular, cellular, and network levels. Lab on a Chip. 2015;15(13):2767-2780. doi:10.1039/C5LC00133A. Accessed 2026-06-12.