Cortical Labs, the Melbourne-based biotech company that made headlines by teaching neurons to play Pong, is now building two small data centers powered by lab-grown human brain cells in Singapore and Melbourne. The move represents the first known attempt to commercialize biological neural networks as computing infrastructure, raising both technical promise and ethical questions about what it means to run data operations on living tissue.
From Pong to Processing Power
The scientific foundation for this effort traces back to a peer-reviewed study published in the journal Neuron by lead researcher Brett Kagan and colleagues. That work described how in vitro neural networks, derived from both human and rodent cell lines, were integrated with silicon via high-density multi-electrode arrays and placed in a closed-loop simulated game environment. The neurons learned to play Pong, adapting their electrical activity in response to game-state feedback. The study, known as the DishBrain experiment, provided the first rigorous demonstration that disembodied biological neurons could exhibit goal-directed behavior when coupled to digital systems.
That research generated significant scientific debate. A Nature News analysis of the Kagan et al. findings questioned what “learning” actually means in this context, noting that while the neurons improved their Pong performance over time, the interpretation of that improvement as genuine learning or even sentience remains contested among neuroscientists. Some researchers argued the observed changes could reflect simpler forms of adaptation rather than anything resembling cognition. That distinction matters enormously now that Cortical Labs wants to scale the same biological platform into commercial data centers.
Two Data Centers, Two Continents
Cortical Labs is working on two small data centers that will be run by human brain cells, with facilities planned in Singapore and Melbourne. The data centers will use lab-grown neurons as their core processing elements rather than conventional silicon chips. Specific details about the scale of these facilities, their projected operational dates, and the funding behind them have not been disclosed in available primary sources.
The choice of locations is notable. Melbourne is Cortical Labs’ home base and the site of its original DishBrain research. Singapore, meanwhile, has aggressively positioned itself as a hub for both AI infrastructure and biotech research, offering regulatory frameworks and government incentives that could ease the path for experimental computing architectures. Placing one facility in each city gives the company access to distinct talent pools and regulatory environments while spreading operational risk across two jurisdictions.
The Software That Makes It Work
Running a data center on living neurons requires more than just growing cells on electrode arrays. The neurons need constant, precisely timed input signals, and their output must be captured and routed with minimal delay. Cortical Labs has addressed this challenge by developing a software interface called CL API, described in a technical paper published on arXiv. The CL API enables real-time closed-loop interactions with biological neural networks, supporting deployable code with defined inputs, timing constraints, reproducibility guarantees, and synchronization protocols.
This software layer is the bridge between wet biology and digital infrastructure. Without it, biological neurons would be laboratory curiosities rather than functional computing elements. The CL API effectively standardizes how external systems communicate with neuron arrays, making it possible to write code that treats living tissue as a programmable resource. That shift from bespoke lab setups to a reproducible software interface is what separates a research demonstration from something that could operate at data center scale.
Why Neurons Instead of Silicon
The appeal of biological computing comes down to energy efficiency. The human brain performs extraordinarily complex tasks while consuming roughly the energy equivalent of a dim light bulb. Traditional data centers, by contrast, require enormous amounts of electricity for both computation and cooling. As AI workloads have grown, so has the power consumption of the facilities that run them. If neuron-based systems can perform even a fraction of the tasks currently handled by GPUs and CPUs while using substantially less energy, the economic and environmental case becomes compelling.
But the gap between theoretical promise and practical delivery is wide. Biological neurons are fragile. They require specific temperature ranges, nutrient solutions, and sterile conditions to survive. Scaling from a few thousand neurons on a multi-electrode array to the millions or billions that would be needed for meaningful data processing introduces challenges that no research group has yet solved. The full published study by Kagan et al. details the precise conditions needed to maintain neuron viability during experiments, and those conditions become far harder to guarantee at commercial scale.
There are also open questions about how general-purpose such systems can be. The DishBrain platform excelled at a specific closed-loop control problem, keeping a virtual paddle aligned with a ball. It remains unclear whether similar neuron cultures can handle the diversity of workloads that modern data centers process, from large language models to real-time recommendation engines. Even if they can, orchestrating many biological arrays in parallel would require sophisticated control software and robust fault-tolerance mechanisms.
Ethical Tensions Around Living Compute
The original Kagan et al. paper used the word “sentience” in its title, a choice that drew immediate scrutiny. If neurons in a dish can learn, and if the company’s own published research frames that behavior as exhibiting sentience, then building data centers that run on those same neurons raises questions that go beyond engineering. Are the neurons experiencing something? Do they have interests that deserve protection? How should regulators classify living biological tissue that is being used as commercial infrastructure?
No public regulatory filings or ethical review documents from Australian or Singaporean authorities regarding these specific data center plans have surfaced in available primary sources. That absence is itself significant. The regulatory frameworks for biological computing do not yet exist in any jurisdiction, which means Cortical Labs is building into a legal gray zone. The company’s own peer-reviewed research includes conflicts-of-interest disclosures, acknowledging the commercial stakes of the scientists involved, but commercial deployment introduces a different order of ethical complexity than laboratory experimentation.
Bioethicists are likely to focus on several core issues. One is the origin of the cells themselves, especially when human-derived lines are involved. Another is the possibility that increasingly complex neural architectures could cross some threshold of experience, even if that experience is rudimentary or alien. Without agreed standards for assessing sentience or suffering in vitro, companies could end up defining their own ethical red lines, a situation that risks both public backlash and regulatory intervention once these systems move beyond proof-of-concept.
What This Means for Computing
Cortical Labs’ data center plans represent a test case for whether biological computing can move from provocative demonstration to practical infrastructure. If the Singapore and Melbourne facilities succeed even on a limited scale, they could establish neuron-based processing as a niche complement to traditional chips, perhaps specializing in certain kinds of adaptive control or pattern-recognition tasks where biological networks excel.
On the other hand, failure to deliver reliable performance or clear energy savings would reinforce the view that DishBrain-style systems are more scientific curiosity than industrial technology. The skepticism already present in the scientific commentary around the original Pong experiments underscores how much work remains to convince both researchers and customers that living compute is more than marketing hype.
For now, the two planned data centers are best understood as large-scale experiments in a frontier field. They will test not only the durability and controllability of neuron cultures under commercial conditions, but also society’s comfort with outsourcing bits of its digital infrastructure to living tissue. Whether that prospect is thrilling or unsettling, it forces a reexamination of what counts as a computer, and where the boundary lies between tool and organism in an age of increasingly blurred lines.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
