The Line Between Machine and Tissue Is Getting Blurry
Computing has meant silicon for seventy-odd years. Transistors etched into wafers. Electrons herded through circuits. But a small, odd corner of research is asking a different question. What if the most efficient computer we could build isn’t silicon at all? What if it’s living brain tissue instead?
That’s the premise behind what scientists now call wetware computing organoid intelligence, or OI for short. Researchers are growing clusters of human neurons out of stem cells — structures known as brain organoids — and wiring them up to electrodes and silicon chips. The result is a hybrid biocomputer. It sounds like something out of a novel, but it isn’t. It’s early, narrow, genuinely-happening lab science, and it hints at a strange possible future for computer hardware.
Show Image
What Is a Brain Organoid?
A brain organoid is a three-dimensional clump of neurons, grown from human stem cells in a petri dish. Calling it a “mini-brain” oversells it. There’s no consciousness here, no senses, nothing you’d call thought. What it does have is the ability to self-organize into neuron-like structures, fire electrical signals, and build synaptic connections. In a small way, it echoes how an actual developing brain wires itself up.
Here’s why that catches a computer scientist’s attention and not just a neuroscientist’s: biological neurons are absurdly efficient at processing information. The human brain runs on about 20 watts — roughly a dim bulb’s worth of power. Yet it does things no silicon system can touch without burning through a data center’s electricity bill. Even a sliver of that efficiency, harnessed computationally, would be a huge deal given how much power AI now eats.
How Wetware Computing and Organoid Intelligence Actually Work
Wetware computing setups generally stack three layers on top of each other. There’s a biological layer: the organoid itself, kept alive in a nutrient bath, sometimes for months. There’s an electrode layer: dense microelectrode arrays positioned around or beneath the tissue, able to both listen in on its electrical chatter and stimulate it with targeted pulses. And there’s a silicon layer on top, which does the translating — turning the organoid’s firing patterns into usable data, and turning digital instructions into stimulation the organoid can respond to.
Put those together and you get a crude feedback loop. Some early experiments have gotten organoids to shift their firing patterns in response to electrical feedback. It’s a small, slow echo of how real brains learn through reinforcement, nothing more. One of the more talked-about demonstrations hooked an organoid up to a simple video game and watched its activity adapt as it received feedback signals over time. Nothing close to “playing” — but a real, measurable shift in behavior.
Johns Hopkins researchers, who helped coin the term organoid intelligence, have laid out a similar roadmap: use standardized brain organoids and brain-machine interface tools to explore what these tissue cultures can compute, store, and learn (Frontiers in Science, 2023).
The Real Draw Behind Wetware Computing Organoid Intelligence: Energy
Ask anyone in this field why it exists, and the answer comes back to power. Training one large AI model can burn through as much electricity as thousands of households use in a year. Data centers already account for a growing slice of the world’s power draw. Biological neurons sidestep most of that. They lean on ion channels, neurotransmitters, and synaptic plasticity — mechanisms silicon has no real analog for — and they do it at a fraction of the energy cost.
If organoid systems ever get reliable and scalable enough for real use, the pitch is simple. They’d offer a much cheaper path to certain kinds of computing, especially pattern recognition and adaptive learning: the tasks biological neural tissue already happens to be good at. Researchers at Johns Hopkins have framed this directly as a response to the physical and energy ceilings that conventional silicon-based computing is starting to hit (JHU Chemical & Biomolecular Engineering, 2023).
Where the Science Actually Stands
Worth being blunt here: organoid intelligence is a research field, not a product. Today’s organoids are tiny. Most are just a few millimeters wide, holding maybe a few hundred thousand to a couple million neurons. Set that against the roughly 86 billion neurons in a human brain, and the gap is obvious. They can’t survive long outside carefully controlled lab conditions. Whatever “computing” they’ve managed so far is nowhere near what an ordinary machine-learning system does on a laptop.
The people actually working on this are spread across neuroscience, bioengineering, and computer science departments, plus a handful of biocomputing startups. Right now they’re still wrestling with the basics: keeping tissue alive longer, reading its signals more reliably, scaling organoids up, and building training protocols that produce consistent results. A 2025 Johns Hopkins study found that lab-grown organoids show real signs of synaptic plasticity — the strengthening and weakening of neural connections that underlies learning and memory — which the researchers describe as a foundation for future organoid intelligence work (Johns Hopkins Bloomberg School of Public Health, 2025).
The Ethics Nobody Can Fully Answer Yet
This field runs straight into ethical territory that neither computer science nor bioethics has a ready-made rulebook for. Current organoids show no sign of anything like consciousness. But the direction of travel is toward bigger, more complex neural tissue, built specifically to process information. That trajectory has scientists asking hard questions they don’t have full answers to yet.
Where does moral consideration begin, if it ever does, for tissue this complex? What about the stem cells themselves? Most trace back to donated human tissue, which brings in the usual consent and downstream-use questions bioethics has dealt with for decades. And should institutions build dedicated oversight for this kind of research, given how little precedent exists? To their credit, most researchers in the space treat these questions as central to the work, not a box to check afterward. That’s not always true of fast-moving scientific fields.
What Wetware Computing and Organoid Intelligence Might Make Possible
Real-world wetware computers are probably years, maybe decades, away. Still, people in the field point to a few directions this could realistically head.
Ultra-low-power edge computing is one, for situations where saving energy matters more than raw speed. Drug testing and disease modeling is another — organoids are already used this way, somewhat separately from the computing angle, to study how neurons react to drugs or to model conditions like Alzheimer’s. That’s a nearer-term payoff from the same underlying lab technique. Further out, there’s the idea of self-repairing computing systems that borrow biological tissue’s knack for adapting and healing itself, something silicon simply can’t do.
(Related reading: see our internal guide to brain-computer interfaces and our explainer on how neural networks compare to biological neurons.)
Why It’s Hard Not to Find This Fascinating
What’s genuinely strange about this field isn’t just the biology. It’s what it does to the idea of “hardware” itself. Hardware has always meant something built. This field is asking whether it could instead mean something grown. That’s a harder distinction to hold onto than most people assume. It’s also a humbling reminder: for all the progress silicon has made, evolution solved the energy problem in computing a very long time ago. We’re only just starting to figure out how to borrow its tricks.
Keeping Up With the Field
This is still very much academic-lab territory: university neuroscience departments, government-funded research initiatives, and a small number of biocomputing startups. If you want to follow along or eventually work in the space, here’s where to start.
Track published research from the university groups actively working on wetware computing organoid intelligence; a lot of it gets published openly in neuroscience and computational biology journals, including Frontiers in Science. Build a foundation in both directions at once — neuroscience (neural signaling, stem cell biology) and computer science (machine learning, signal processing) — since this field doesn’t really tolerate specialists in only one. And don’t skip the bioethics literature on organoid research. Taking the ethical side seriously is part of doing the work responsibly, not an optional extra.
Final Thoughts
Wetware computing organoid intelligence is still squarely in foundational-research territory, not anything you’ll be buying anytime soon. But it’s one of the genuinely novel directions computing has taken in a long while. Whether it ever becomes mainstream or stays a lab curiosity, the research is already changing how scientists think about biology, energy, and intelligence together. It’s worth keeping an eye on, if only because it refuses to sit neatly inside either “biology” or “computer science.”