Diff — FinalSpark

{{infobox

| title = FinalSpark

| subject = The Swiss company renting living brain organoids as cloud processors

| type = Private biocomputing company — "wetware-as-a-service"

| founded = 2014 · Vevey, Switzerland

| founders = Dr. Fred Jordan & Dr. Martin Kutter

| product = Neuroplatform — remote, 24/7 API access to human brain organoids

| scale = 16 organoids live · 1,000+ hosted · 18+ TB of neural data · organoid lifespans >100 days

| price = ~$500/month academic · free for selected research institutions

| ambition = "The world's first living processor" — AI for 100,000× less energy (company claim)

| related = [[Biological Computing]] · [[Reversible Computing]]

| image = fs-vevey.jpg

| caption = Vevey, on Lake Geneva — the Swiss lakeside town where a six-person company keeps sixteen living brain organoids online, rentable by the month (JoachimKohler-HB, CC BY-SA 4.0).

}}

On the northern shore of Lake Geneva, in the Swiss town of **Vevey** — better known for Nestlé's headquarters and Charlie Chaplin's last home — a company of about six people runs the strangest cloud service on the internet. For **$500 a month**, any accredited researcher can log in from anywhere on Earth and run experiments on living human brain cells. Not simulations of brain cells. Actual, living, firing neurons — grown into tiny spherical **organoids**, wired to electrodes, kept alive in automated incubators, and addressable through a **Python API** like any other compute instance.

The company is **FinalSpark**, its product is the **Neuroplatform**, and its stated goal is nothing less than *"the world's first living processor."* It is the other half of the story [[Biological Computing]] tells — where Cortical Labs sells you the box, FinalSpark rents you the brain.

# Two Frenchmen and a counterfeiting fortune

FinalSpark exists because of fake handbags.

The company is, to this day, largely self-funded by AlpVision's profits — a counterfeiting-detection business quietly bankrolling the development of living processors. It is one of the more quietly extraordinary funding stories in technology.

# The Neuroplatform

The machine, in plain terms:

- **Sixteen live organoids** online at once — four per processing unit, each organoid ~0.5 mm across, sitting on a multi-electrode array with eight electrodes reading and stimulating it

- **>100-day lifespans** — the organoids live in automated microfluidic incubators that feed them, control their environment, and change their medium without human hands

- **A Python API** — real-time stimulation and recording through a documented library, Jupyter-friendly, with data storage, a digital lab notebook, and technical support

- **Dopamine as reward** — the training signal is the brain's own: expose the organoid's neurons to dopamine when they produce the desired response, and the tissue rewires to produce it again, the same reinforcement chemistry a skull uses

[[File:fs-organoid-dev.jpg|right|320px|A brain organoid during development — a self-assembled ball of neural tissue a fraction of a millimetre wide. Sixteen of these, on electrode arrays, are what a FinalSpark customer rents (NIH Image Gallery, public domain).]]

# The claim — honestly flagged

FinalSpark's pitch rests on one number, and it's worth flagging precisely. The company claims bioprocessors "consume a million times less power than traditional digital processors," and Jordan has stated the ambition as "artificial intelligence for 100,000 times less energy."

Both figures are **aspirational comparisons, not benchmarks** — they compare the metabolic efficiency of neurons to the total energy of training and running large digital AI, not measured compute-per-joule on a head-to-head task. A Neuroplatform organoid cannot (yet) do what a GPU does; what it *can* do is learn from remarkably little stimulation on essentially no power, which is a different and still remarkable thing. The honest version of the claim: **the physics of wetware's efficiency advantage is real; the product-level equivalence is still being built.** The distinction matters, and the field — including FinalSpark's own published paper — is careful to make it.[[cite: Company claims per Business Wire and interviews; the energy figures are aspirational comparisons as flagged in the FinalSpark Frontiers paper's own framing]]

# The questions it raises

FinalSpark runs sixteen living balls of human neural tissue as a commercial service — which raises questions the field itself is only beginning to formalize. The organoids are human neurons. They are small (~10,000 cells — a fruit fly's worth, nowhere near the ~86 billion of a human brain), they have no sensory organs, no body, no way to act except through the electrodes — but they *do* respond, adapt, and learn. Nobody currently has an instrument for whether there is anything it is *like* to be an organoid in a Vevey incubator, and nobody knows where the threshold is between *it responds* and *it experiences*.

[[File:bio-mea.jpg|right|320px|A microelectrode array — the kind of chip an organoid sits on: eight electrodes per organoid for stimulation and recording, the interface between the wetware and the Python API (Kwayyy, CC BY 4.0).]]

# What might happen

The roadmap is the company's own, flagged as such:

- **The raise** — Jordan has said FinalSpark is seeking ~**$50 million** to hire thirty scientists and build out the cloud platform — "the cloud computing application ready within seven years," as he put it

- **Scale** — more organoids per unit, more units per platform, the path from research rental to actual bio-compute service

- **The applications** — drug screening on human neurons, disease modeling, AI research that needs biological learning at biological prices

- **The far target** — "the world's first living processor" is not a metaphor in their usage; they mean a commercially useful computer whose substrate is living neural tissue

```fact The understated summary

Six people in Vevey, funded by counterfeit-detection profits, run a Python API into sixteen living balls of human neurons — and you can rent one for less than a gym membership. Whatever the future of computing turns out to be, it is probably already weird.

```

# The last word

There is a particular kind of technology company that only exists at the exact moment its idea stops being fiction — too early for anyone to believe it, too late for anyone to have founded it first. FinalSpark is one of those. Two physicists who spent twenty years spotting fakes now run the world's most honest service business: they rent you a brain that is, unambiguously, real.

# Quiz

- [ ] 1 — What does FinalSpark's Neuroplatform actually rent you?

- [ ] 2 — What unrelated business funds FinalSpark's development?

- [ ] 3 — What chemical does the platform use as the neurons' "reward" signal?

- [ ] 4 — How many organoids run live on the platform, and for how long do they stay viable?

- [ ] 5 — What is the honest caveat on the "million times less power" claim?

```hide Show answers — click to check yourself

1 — **24/7 remote API access to living human brain organoids** — real neurons on electrodes, ~$500/month, Python API.

2 — **AlpVision** — the founders' anti-counterfeiting company (~30 billion products/year protected).

3 — **Dopamine** — the brain's own reinforcement signal, released when neurons respond correctly.

4 — **16 live organoids** at once (4 per unit), surviving **>100 days** in automated incubators.

5 — It's an **aspirational comparison**, not a measured benchmark — the neurons' efficiency is real, but they can't yet do what a GPU does.

```

# Notes

1. FinalSpark's specifications (16 organoids, >100-day lifespans, 1,000+ hosted, 18+ TB data, $500/month) are the company's published figures via its Frontiers paper and launch press.

2. The "million times less power" / "100,000× less energy" claims are marketing-level comparisons, flagged honestly — aspirational, not benchmarked.

3. Founding dates: AlpVision ~2001; FinalSpark 2014; Neuroplatform public launch May 2024.

4. Organoid neuron counts (~10k/organoid) reflect published figures; the human brain comparison is order-of-magnitude.

5. FinalSpark's paper is open access — *Frontiers in Artificial Intelligence* 7 (2024), PMC11097343.

# References

- FinalSpark, Business Wire launch release (May 2024) — first remote platform for biocomputing on human neurons.

- Infinite Frontiers, "Interview: Fred Jordan" (June 2024).

- Tom's Hardware, "Human brain organoid bioprocessors now available to rent for $500 per month" (August 2024).

- FinalSpark corporate materials — finalspark.com/neuroplatform.

[[Category:Computing]]

[[Category:Companies]]

[[Category:Emerging technology]]