A biological data center is a room of computers that run on living human brain cells grown on silicon chips. Cortical Labs, a company in Melbourne, Australia, grows these neurons and rents their computing power to researchers. Knowing how the machines work gives you a clear way to judge the headlines, including the striking claim that they sip power while conventional data centers strain the grid.
You may have met the idea in a news alert about brain cells playing a video game or steering a robot dog. The technology is real, and it is small. Only a few facilities exist, measured in dozens of units, and the ethical questions are arriving ahead of the products. That makes now a good time to learn the basics, before the claims get bigger.
Start by holding three questions in mind: what was actually shown, how many units exist, and who outside the company checked the work. Those three questions sort a real result from a sales pitch in every breakthrough story you read.
What a Biological Data Center Is
A biological data center gives you a plain picture of a strange machine: living neurons on a chip, wired to a computer, doing small computing jobs. The film The Matrix imagined machines farming humans for power. A biological data center turns that story around. Human neurons do the thinking for the machines, and the sales pitch is that they use very little power. Once you see how the cells receive signals and learn, the company's claims and the limits make much more sense.
Neurons on a Chip
The cells start as stem cells. Scientists take ordinary donated cells, such as blood or skin cells, and reprogram them into induced pluripotent stem cells. Those stem cells can grow into almost any cell type, including neurons. The neurons grow on a flat chip lined with tiny electrodes, called a multi-electrode array. Over several weeks the neurons reach out to each other and form connections, much as they do in a developing brain.
The electrodes work in both directions. They send small electrical pulses into the cells, and they read the electrical spikes the cells fire back. Software translates a task into patterns of pulses. The neurons respond, and the software reads that response as an output, such as moving a paddle up or down.
How the Cells Learn
Learning comes from feedback. In the Pong experiments, published in the journal Neuron in 2022, the team gave the cells a predictable signal when they hit the ball. A miss brought a burst of random noise instead. Neurons tend to organize toward predictable input, so the cells changed their firing in ways that kept the rally going. The researchers called the setup DishBrain.
Living Cells Versus Silicon
A conventional processor runs on billions of transistors that switch on and off in fixed patterns. It follows its code exactly, at enormous speed, and it draws a lot of power to do it. Living neurons work by changing their own connections. They are slower and less precise, and they adapt as they go. That is the core trade a biological data center offers: speed and certainty on one side, adaptation and low energy on the other.
We define biocomputing as the use of living biological material to process information. A related research area, organoid intelligence, grows three-dimensional clusters of brain cells called organoids and studies whether they can compute. Both fields are young, and the results so far come from lab demonstrations. Next comes the company that turned those demonstrations into a product.
Cortical Labs and the CL1
Knowing the company and its record gives you the facts behind the headlines. Cortical Labs built the machines inside the first biological data centers. Its record runs from a lab game to a walking robot, and each step sets up the bigger question of what the technology can really do.
The CL1 Computer
The CL1 is the company's biological computer, launched in March 2025. Each unit holds a chip of living neurons inside a life-support system that pumps nutrients, filters waste, and holds the cells at body temperature. The company says the cells can stay alive for months. The CL1 is code-deployable, which means a researcher can write software and run it directly on the neurons, much like deploying code to a server. Cortical Labs also offers remote access through the cloud, so a lab can run experiments without owning a unit.
From Pong to Doom to a Robot Dog
- Pong. The DishBrain cells learned to return the ball in a simple version of the game, and the work appeared in a peer-reviewed journal.
- Doom. Cortical Labs later showed neurons on a CL1 playing the classic shooter Doom, a more complex game with movement and targets.
- Robot dog. In October 2026 the company used living neurons to steer a four-legged robot dog, moving the cells from a screen into the physical world.
Prepper Life news covered the robot dog demonstration as Cortical Labs expanded its small biological data centers in Melbourne and Singapore.
Melbourne and Singapore
The Melbourne facility holds 120 CL1 units, according to the company. In Singapore, Cortical Labs is building a 20-unit prototype with the data center operator DayOne and the National University of Singapore. The partners plan to grow it to as many as 1,000 units. For scale, a large conventional data center can hold tens of thousands of servers.
The Power Claim
The headline number comes from CEO Hon Weng Chong, who says each CL1 unit uses less power than a handheld calculator. That is a striking claim while conventional data centers strain power grids. Read it as the company's own figure until outside groups test it on real workloads. The next question is what that low power actually buys.
What a Biological Data Center Can and Cannot Do
Weighing the promise against the limits gives you a fair reading of the technology, free of hype in either direction. The company and its critics both make reasonable points. Set them side by side, then decide for yourself. After that come the harder ethical questions.
The Company's Case
- Energy efficiency. A human brain runs on about 20 watts, and the company argues that neurons on a chip carry some of that efficiency with them.
- Adaptability. Neurons rewire themselves in response to feedback, so the company argues they can adjust to new tasks without new hardware.
- Learning from little data. The Pong cells improved within minutes of play, and the company argues biological systems need far fewer examples than today's AI models.
The energy point matters most if you follow the environmental impact of AI data centers. The largest facilities draw as much power as a small city and use water for cooling. Some regions have slowed new grid hookups or debated who pays for upgrades, a pattern we track as the grid threat nobody voted for. A computer that sips power would ease that load.
The Skeptics' Case
Outside researchers are more cautious. Ernst Wolvetang, a stem cell researcher at the University of Queensland, has said that flat layers of neurons on a chip remain simple systems. A CL1 holds a few hundred thousand neurons. A human brain holds about 86 billion, arranged in layers and regions. Playing a game is a narrow task. No independent test has yet shown these cells beating conventional chips on real commercial workloads.
Keeping Cells Alive
Living cells also need care that silicon never does. They need nutrient fluid, waste removal, steady temperature, and sterile conditions. Contamination can end a culture, and every culture eventually dies and must be replaced with freshly grown neurons. A full facility also runs pumps, incubators, and climate control. Ask whether any power claim covers the whole system or only the cells.
The energy idea is real and attractive. The scale is tiny, and performance on real jobs is unproven. Both things are true at once, and you can hold them together.
The Ethics and What to Watch
Understanding the ethical questions gives you the words to follow the debate as it grows. The systems are small and the questions are early, which is the best time to learn them. Then comes a simple routine for judging any breakthrough claim, one you can practice this weekend.
Consciousness
The biggest question is whether neurons on a chip could ever feel anything. Cortical Labs says it works with bioethicists to make sure its cultures do not become conscious. Many researchers consider current flat cultures far too simple for awareness. They also expect the question to grow as each biological data center scales up and as organoid intelligence work moves into three-dimensional tissue.
Consent for Donated Cells
Every neuron in a CL1 traces back to a person who donated cells. Donors usually sign consent forms for research use. Commercial computing is a newer use. Ethicists ask whether older consent covers it, whether donors should be told, and whether they should share in any profit. Medical history includes cell lines used for decades without the donor's knowledge, so the question gets close attention now.
Who Sets the Rules
Rules written specifically for computers built from human neurons are still taking shape. Oversight today comes from research ethics boards, stem cell guidelines, and laws written for medicine. As facilities spread from Australia to Singapore, the rules may differ by country. Watch for professional bodies and governments to address the field directly as it grows.
The Matrix Question
The Matrix remains the cultural shorthand for these fears: human biology locked inside a machine. The real systems are a few hundred thousand cells on a chip playing games and steering a small robot. Keeping both pictures in view lets you take the ethics seriously and keep the technology at its true size.
Check the Claim
Start practicing a three-question habit on the next breakthrough headline you see. Preparedness covers more than food, water, and power. It also covers judging information when a story moves fast. Preppers already check sources before acting, and the same skill sorts a real advance from a press release.
- What was shown? Name the actual demonstration, such as a game played or a robot steered, and set the predictions aside.
- How many units exist? Count the working machines, since 120 units in one building is a pilot and 1,000 units is still a plan.
- Who outside the company reviewed it? Look for a peer-reviewed paper or an independent test, and note when the company is the only source.
If you have kids who saw the robot dog video, walk through the questions together; a curious child makes a sharp fact-checker. If you work in tech or follow investing, add a fourth question about who benefits from the claim. If you have only a few minutes, the third question alone filters out most hype. Once the habit sticks, use it on every AI story that reaches your home, a core step in protecting your family from AI.
