Singapore and Melbourne are preparing sites for a form of computing that sounds less like cloud infrastructure and more like experimental biology. At the center are biological computing systems built for real workloads.
Early demonstrations moved from simple game play toward harder tasks, hinting that the idea is maturing. The CL1 links lab-grown neurons to silicon interface technology, while drawing a fraction of the power consumed by dense AI racks. That changes the argument fast.
Why Cortical Labs is pairing neurons with silicon
Cortical Labs is pairing living neurons with silicon chips to build machines that learn in a less conventional way. At its Melbourne biological data center, CL1 units will sit at the core of that effort, while a second site is taking shape in Singapore with DayOne Data Centers Ltd. In materials released Tuesday, each unit is described as using neurons created through stem cell conversion so software can interact with tissue.
The chip beneath each culture manages the dialogue needed for computation. Through electrical signal exchange, it sends prompts, receives activity, and records the results for analysis. That stream supports neural response mapping, giving Cortical Labs a way to turn cellular reactions into usable output. The arrangement reflects a distinctive biotech hardware design : not rows of standard processors, but lab-grown neurons joined to silicon to test whether brain-like efficiency can be shaped into computing systems.
From Pong to Doom, the CL1 system shows steady progress
Cortical Labs has used video games as a measuring tool for learning inside the CL1 platform. An early benchmark was Pong, where the cells handled a simple task. Last month, the company said the same system had moved to Doom through game-based training, rather than by simply following a rigid software script alone.
Researchers are careful with the claim, and that restraint fits the evidence. The shift points to adaptive cell behavior under controlled conditions, with neurons altering their responses as the chip delivers prompts and reads feedback. For the team, the move from Pong to Doom signals a broader range of pattern handling, not just a repeatable reflex. Cortical Labs frames the result as progress for biological computing, while admitting that the system has modest capacity.
Low power use gives these sites a different profile from AI server farms
The sharpest contrast with mainstream AI infrastructure is power draw. Cortical Labs says each CL1 uses less energy than a handheld calculator, pointing to calculator-level consumption rather than the heavy loads tied to advanced GPUs. That profile stands apart as AI expansion raises water use concerns and pushes operators toward reduced electricity demand targets in data centers across several regions.
Hon Weng Chong, the startup’s founder and chief executive officer, said the Melbourne facility will house 120 units, while the Singapore site with DayOne will add as many as 1,000 in phases. The first Singapore deployment is planned at the National University of Singapore’s Yong Loo Lin School of Medicine, using neurons converted from human blood cells. At that scale, Cortical Labs is testing an alternative compute model for data centers, even if these systems remain far from replacing conventional AI server farms.
