Field guide

FinalSpark Neuroplatform, explained

The Neuroplatform is a lab in Vevey, Switzerland, run by FinalSpark, where human brain organoids sit on electrode arrays and researchers anywhere can stimulate and record them from a Python notebook. I went through the 2024 paper that describes it, and these are the numbers it gives.

Peer-reviewed sourcesUpdated 13 November 2026

The platform in numbers

Figures from the 2024 paper you'll find in the sources below.

0organoids tested
0of data recorded
0electrodes on each array
0lifetime in the best cases
Part 01

Who built it

The paper came out in Frontiers in Artificial Intelligence on 2 May 2024. All five authors work at FinalSpark and they say so. You should read it as the company describing its own system. The paper itself says the platform is free for research use and gives no price.

Part 02

The organoids

What you're working with are forebrain organoids grown from human induced pluripotent stem cells by way of neural stem cells. Each one is around 500 micrometres across (about the size of a grain of sand). They go onto the recording system at about twelve weeks old. The paper doesn't give a cell count per organoid.

Part 03

The hardware

PartWhat the paper reports
Electrode arrays4 MEAs, each holding 4 organoids, with 8 electrodes per organoid (32 per MEA)
Recording30 kHz sampling at 16-bit resolution, about 0.15 microvolts of accuracy
Life supportNeuronal medium pumped at 15 microlitres a minute through a 170 microlitre chamber, organoids at an air-liquid interface, two incubators
Other stimulationUV light at 365 nm to release caged dopamine onto the organoids

The electrical stimulation can be biphasic or triphasic pulses. One thing you'll notice if you read closely is that the paper gives two different current ranges in two sections, so I'd check with FinalSpark before relying on either.

Part 04

How remote access works

You log in by Remote Desktop and work in Jupyter notebooks through a Python API. The same API runs the pumps and cameras (and the UV lights too). So in principle you can run a whole experiment without anyone touching the lab. Recorded data go into a shared database while each group keeps its scripts private. In 2023, 36 academic groups proposed projects and 8 were chosen.

Part 05

What it has done so far

Over the years the platform has tested more than 1,000 organoids on about 250 array replacements. It has recorded over 18 terabytes of data and more than 20 billion action potentials. What struck me most is the lifetime. Early on the organoids lived for a few hours and the best now last up to 100 days (the abstract says more than 100 in some cases).

Part 06

The limits

The paper is honest about these. If you're thinking of using it, they matter more than the headline figures. Closed-loop learning experiments run on one organoid at a time, though parallel runs on up to 32 are planned. There's only one way of growing the organoids in use. The current needed to make an organoid fire goes up as it ages, possibly because glial cells wrap around it, and its spontaneous activity drifts from day to day. The authors also say their single dopamine experiment can't show cause and effect.

Don't read too much into the energy case either. The paper cites about 10 gigawatt-hours to train GPT-3 and about 20 watts for a human brain, but it doesn't measure the platform's own power use. I cover that comparison on brain vs GPU power. For the wider field read organoid intelligence explained, and for the best-known earlier experiment see DishBrain and Pong. If you're weighing this against the machine Cortical Labs sells, I've put the two side by side in FinalSpark vs Cortical Labs.

Part 07

Questions

What is the FinalSpark Neuroplatform?

It's a remote lab run by FinalSpark in Switzerland where human brain organoids live on electrode arrays, and researchers stimulate and record them over the internet through a Python API.

How long do the organoids live?

The 2024 paper says up to 100 days in the best cases, and more than 100 in some, up from a few hours when the project began.

Can anyone use the Neuroplatform?

The paper describes it as free for research, with access given to selected academic groups, and in 2023 eight of 36 proposals were accepted.