Wetware: neurons playing Doom, the 20-watt brain, organoids for rent
So, as of March 2026, there's a clump of about 200,000 living human brain cells hooked up to a silicon chip, actively learning how to play the 1993 video game Doom. Hey, I'm Salt. I'm Grace. Okay, Grace, living brain cells playing a vintage shooter. Why are we building this?
To save the power grid, mostly. Right now, as of August 2026, global data centers eat up roughly two percent of the world's electricity. And Penn State extension data projects that footprint is going to triple by 2030. Just to run AI models. Right.
I mean, training these large language models takes an unbelievable amount of electricity. We're building bigger and hotter server farms just to keep up. But a human brain does all its processing on about 20 watts of power, which is roughly what a dim lightbulb uses. Wait — what? Yeah.
Wetware, they call it. Biological processors consume up to a million times less power than silicon chips. So researchers are trying to harness that efficiency to solve the energy crisis. Back in October 2022, a Melbourne startup called Cortical Labs built a system called DishBrain. DishBrain.
They took a silicon chip and layered it with 800,000 living human and mouse neurons. And they taught that tissue to play Pong. Okay, walk me through that. How do you actually teach a layer of cells to play a 1970s arcade game? They used something called the Free Energy Principle.
The core idea is that biological systems naturally want to minimize unpredictable environments, so the scientists would feed electrical signals into the cells to indicate where the paddle was and where the ball was, and then just wait to see how the network responded. So they just watch it. Right. And then if the cells fire the right way to hit the digital ball, they get a nice, predictable electrical pulse. It's stable.
But if they miss the ball, the software sends back randomized, chaotic static. The cells hate the chaos, so they they physically restructure their own neural networks to get better at hitting the ball, just to stop the noise. That is wild. They just reorganise themselves to avoid the static. Exactly.
And the concept proved out. So fast-forward to March 2026, Cortical Labs scales the whole thing up. They build a code-deployable biological computer called CL1. Is that the one playing Doom? It's the, uh, it's the one.
CL1 uses 200,000 living human neurons. And those specific cells were actually grown from the blood of the startup's CEO. You're kidding. He donated his own blood to make the stem cells that eventually became the neurons in the machine. So now they have this living human tissue navigating a 3D environment, dodging enemies in Doom.
So they've got this guy's neurons in a lab in Australia playing old games. Where does this go next, Grace? Are companies going to start selling access to this kind of wetware? We're already there. In 2024, a Swiss startup named FinalSpark launched the Neuroplatform.
It's the world's first biological cloud-computing service. Meaning I can rent a brain in the cloud? You can rent 24/7 remote access to 16 living human brain organoids. They keep the tissue alive in life-support fluid, hook it up to electrodes, and researchers anywhere in the world just just log in online to run code on human cells. I mean, um, how long do the organoids even live inside those machines?
Because they're biological. They have to need food or something. They need a constant supply of nutrients, yeah. And even with that, they only live a few months, usually. Then the tissue degrades, it dies, and the lab techs swap in a fresh organoid to keep the server running.
Okay, that feels dark. We're renting human tissue by the hour and zapping it with chaotic noise until it does our math. Is anyone regulating this at all? There are no binding international frameworks for organoid intelligence right now. The industry just operates on loose guidelines.
We really don't have the ethical vocabulary yet for human biology becoming a tech commodity. Do people care, though? Because on one hand, it's, it's just lab-grown cells, you know, it's not a person. But on the other, they're learning to avoid chaos. The journal Neuroscience published a study on public opinion in July 2026.
And they found this, they found this really weird paradox. What kind of paradox? You'd think that if people believed the organoids were conscious, they'd want stricter rules to protect them, right? But the study found the opposite. When respondents perceived the biological systems as having higher consciousness, their support for using them actually went up.
Huh. Yeah. People seemed more willing to accept the technology when they thought of it as sort of aware. I guess if it's conscious, we kind of relate to it more? Or maybe we just think it makes for a better computer.
But we're building an industry where the core processor could theoretically feel the static we use to train it. And the power demands of AI are forcing the issue. If silicon data centers keep draining the grid at this pace, the pressure to shift to 20-watt biological servers is only going to grow. So we're transitioning from renting server space to renting lab-grown human brain tissue by the hour. And we're left wondering if we're building the future of sustainable computing, or just a new category of biological commodity that we have no idea how to govern.
Thanks a lot for listening to Daybrain.
Leave your email and we will let you in as soon as Daybrain opens up.