Singapore Launches First Biological Data Center, Challenging Silicon Dominance

**Wetware: The Data Center Built From Living Neurons** We talk constantly on this show about the infrastructure bottleneck — the power, the land, the financing, the chips. Jensen Huang is co-signing hundred-billion-dollar loans. Nvidia is becoming a landlord.

The numbers are so large they've started to lose meaning. So today, let's talk about a group of researchers in Singapore who looked at all of that and asked a genuinely different question: what if we stopped using silicon entirely? This week, DayOne, Cortical Labs, and researchers at NUS Medicine switched on Singapore's first biological data center prototype.

Instead of chips, it runs on wetware — real neurons, grown from stem cells, wired into a rig that processes information like a tiny brain in a box. And before you dismiss this as science fiction, consider one number: your brain runs on roughly twenty watts. A single AI training run can consume the equivalent power of thousands of homes.

The gap between biological and silicon computing efficiency isn't incremental — it's civilizational. **The Technical Picture** Here's what's actually happening at the hardware level. Cortical Labs has been growing human and mouse neurons on multi-electrode arrays — essentially flat dishes studded with sensors that can both stimulate neurons and record their responses.

The neurons self-organize into functional networks. You feed them electrical signals, they process, they respond. It's not a metaphor for a neural network.

It is a neural network, a literal one, made of biology. The key advantage is energy. Biological neurons operate on electrochemical gradients that are extraordinarily efficient compared to transistor switching.

They also exhibit plasticity — they physically rewire based on experience — which means they can adapt without the massive retraining cycles that make modern AI so expensive. The Singapore prototype is still small, and the computational tasks it can handle are far below what a GPU cluster manages. But the efficiency-per-computation ratio is the entire point.

This isn't about matching GPUs today. It's about a fundamentally different curve. The challenge is interfacing.

Getting data in and out of a biological system requires translation layers — converting digital signals to electrochemical ones and back — and that conversion has overhead. Keeping neurons alive, maintaining temperature, pH, nutrient supply — this is a lab environment problem, not a server room problem. Yet.

**The Financial Case** Let's put some numbers on the problem this is trying to solve. Data center power costs have risen sixty-six percent due to infrastructure inflation. The Ohio project Nvidia just backstopped could top five hundred billion dollars total.

Goldman Sachs has estimated the AI industry needs to spend a trillion dollars on infrastructure over the next several years just to meet projected demand growth. Meanwhile, power is becoming the single binding constraint. Utilities can't build transmission lines fast enough.

Permitting timelines for new generation run five to ten years. The companies that can crack energy efficiency don't just save money — they unlock capacity that no amount of capital can otherwise buy. A biological system that processes at even one-tenth the energy cost of silicon — for suitable workloads — isn't a curiosity.

It's a competitive weapon. The companies investing now are buying optionality. Cortical Labs has already raised venture funding.

The Singapore government is clearly signaling strategic interest. When governments start treating a technology as infrastructure, the commercialization timeline accelerates. **Market Disruption** The incumbent at risk here is obvious: Nvidia.

The entire AI infrastructure stack is currently built around GPU parallelism, and Nvidia's dominance — evidenced by the very financing deal we covered in headlines — rests on that architecture being the only viable path. Wetware doesn't threaten Nvidia tomorrow. But it threatens the assumption that underpins Nvidia's leverage: that silicon is the only substrate.

This also reshapes the data center geography story. Biological systems don't need to be near cheap power in the same way GPU clusters do. They need stable environments, reliable biology supply chains, and specialized maintenance.

Singapore's play here is deliberate — they're positioning as the biotech-meets-AI hub for Southeast Asia, much like they did with semiconductor manufacturing decades ago. For hyperscalers — Amazon, Google, Microsoft — wetware represents both a threat and an opportunity. The threat: a new compute substrate they don't control.

The opportunity: a potential path out of the power bottleneck that is increasingly their existential constraint. Expect quiet acquisition conversations around Cortical Labs and similar firms within eighteen months. **Cultural and Social Impact** There's something genuinely strange about this moment that deserves acknowledgment.

We are now building computers out of human neurons. The philosophical and ethical surface area here is enormous. Cortical Labs has published work showing their neuron cultures exhibit rudimentary learning behavior — they play Pong.

They adapt. At what point does a biological computing substrate acquire moral status? Who regulates it?

Under what framework? These aren't hypotheticals anymore. Singapore has launched a prototype.

The EU AI Act covers software systems. It says nothing about wetware. The FDA regulates medical devices involving human tissue.

Does a biological data center qualify? Nobody knows, because nobody wrote the rules for this. For the general public, the framing matters enormously.

"AI trained on living brain cells" will generate headlines that make this technology radioactive in the court of public opinion if the industry doesn't get ahead of the narrative. The researchers involved need ethicists, communicators, and policymakers in the room now — not after the first viral outrage cycle. **Executive Action Plan** If you're running an AI-dependent business, here's how to think about this practically.

First, add wetware to your technology radar — formally. This isn't a five-year story anymore. Singapore's prototype is running today.

The commercialization curve for biological computing is likely ten years to meaningful scale, but the investment decisions that position you for that window are made in the next two years. Put someone on your team whose job is to track this specifically, not as a footnote in a broader AI report. Second, engage on the regulatory question now, before the rules are written.

The companies that shaped the EU AI Act were the ones that showed up early. Wetware regulation is a blank page. Industry consortia, standards bodies, and government advisory roles are wide open.

The frameworks that get built in the next three years will either accelerate or strangle this technology. You want a seat at that table. Third, rethink your infrastructure assumptions for long-horizon planning.

If you're making ten-year data center commitments right now — and given the Ohio deal, some of you are — build in explicit review gates tied to alternative compute milestones. A contract signed today that locks you into silicon for two decades is a contract that may look very different if biological or photonic computing reaches commercial viability by 2032. Optionality has real value.

Don't trade it away for the comfort of a fixed price per teraflop. The era of assuming AI infrastructure means GPU clusters and massive power draws may be shorter than anyone currently models. Singapore just turned on a data center that runs on biology.

That's not a headline. That's a starting gun.