Nexalytics Tech
The newest machine in Singapore's data-center scene doesn't just process information — it's alive. The Yong Loo Lin School of Medicine at the National University of Singapore (NUS Medicine), Singapore-headquartered data-center operator DayOne, and Melbourne startup Cortical Labs have unveiled what the partners describe as the world's first independently operated biologically integrated server rack: a "Biological Data Centre" prototype installed at the NUS Life Sciences Institute. The rack houses twenty of Cortical Labs' CL1 biological-computing units, each containing roughly 800,000 lab-grown human neurons — about 16 million living cells in total, grown under the direction of NUS neuroscientist Professor Rickie Patani. An invitation-only demonstration on August 6 drew more than 80 attendees from industry, academia and the digital-infrastructure world, and the partners say their larger ambition is Singapore's first large-scale biological data centre — the first such effort outside Australia.
How a Living Server Actually Works
The CL1 doesn't replace silicon so much as recruit biology to work alongside it. Each unit cultures stem-cell-derived human neurons on a microelectrode array bonded to a CMOS chip; the array feeds electrical signals into the neural network and records how the cells fire back. Cortical Labs' software stack — an operating system it calls biOS — translates digital state information into stimulation patterns for the neurons and maps their responses into a simulated environment, an approach the company markets as Synthetic Biological Intelligence. Because the compute substrate is living tissue, every unit carries its own life-support system that can keep the cultures viable for up to six months, according to Data Center Dynamics — a maintenance model closer to a cell-culture lab than a conventional server aisle.
What It Can — and Can't — Do
It's worth being precise about the state of the art. Cortical Labs' predecessor system, DishBrain, famously taught a culture of human and mouse neurons to play Pong, and the launch demonstration showed live neural activity running through the integrated rack and CL1/Cortical Cloud units. The partners are targeting exploratory applications where biology's sample-efficient learning could shine: drug discovery, neurological-disease research, and data-sparse adaptive learning, with CEO Hon Weng Chong also pointing to humanoid robotics, cybersecurity and fraud detection. What the rack is not doing is training or serving large language models or matching GPU throughput — there is no verified evidence of that today. The same caution applies to the headline energy claim: NUS says biological computing has the potential to run on a fraction of the power of digital computers, but the partners have not published a verified wattage, silicon benchmark or lifecycle comparison for this deployment.
The Long Road From Rack to Hyperscale
The challenges are as biological as they are computational. Cell cultures are perishable and variable; contamination, ethics review and wet-lab maintenance have no equivalent in a conventional facility, and a 20-unit research rack is a long way from a hyperscale hall. There is also, as yet, no independent quantified evaluation of the system's performance or cost. Still, the people involved are serious: Patani frames the installation as a platform for studying learning and adaptation itself, not just computation; DayOne CEO Jamie Khoo positions it as a step toward sustainable digital infrastructure; and Chong describes a deliberate transition from research toward commercial application. Cortical Labs, founded in Melbourne in 2019, has raised about $11 million from backers including Horizons Ventures, Blackbird Ventures, LifeX Ventures, Radar Ventures and In-Q-Tel. In the division of labor, DayOne supplies and hosts the infrastructure, NUS brings the neurobiology and cell-culture expertise, and Cortical Labs provides the CL1 hardware and software.
💡 The Nexalytics Take
This rack won't threaten a GPU cluster any time soon — and it doesn't have to. Its real significance is structural: for the first time, a commercial data-center operator has agreed to host living compute as infrastructure, with a medical school supplying the biology and a startup supplying the machines. If the energy-efficiency claims survive independent measurement, even narrowly, biological co-processors could one day sip power for the adaptive-learning workloads that brute-force silicon does badly. The honest reading today: a remarkable, fragile research platform with credible backers and unanswered questions about scale, reproducibility and cost. The metric to watch isn't neuron count — it's whether the partners can publish a hard performance-per-watt number and keep a culture alive long enough for a paying customer to care.
Sources: NUS Medicine — Biological Data Center prototype announcement (Aug 17, 2026) · Data Center Dynamics — Singapore's first biological data center prototype (Aug 18, 2026) · Interesting Engineering — World-first biological data center
Reporting only; performance claims are the partners' own.
