In mid-2026 a server rack went live inside the National University of Singapore's Life Sciences Institute. It does not look like ordinary data-centre hardware. Twenty CL1 units, each the size of a shoebox, sit in a standard rack. Inside them, lab-grown human neurons rest on multielectrode arrays. Technicians feed the cultures a nutrient mix every three days. A closed system maintains temperature, gas balance, and waste removal. The cells remain viable for roughly six months before they must be replaced. Across the twenty units the installation is said to contain on the order of 16 million living human neurons. Power draw for the whole rack is reported in the range of 850–1,000 watts.
This is not a secret prototype confined to a single lab. Cortical Labs, the Melbourne company behind the CL1, already sells the units commercially and offers cloud access described as wetware-as-a-service. A larger collection of roughly 120 units operates in Melbourne. Partners have spoken of scaling the Singapore effort toward a thousand units once further regulatory, efficiency, and safety conditions are met. The neurons themselves are not taken from living adult brains. They are differentiated from induced pluripotent stem cells, typically reprogrammed from adult blood or skin cells donated under ordinary research or commercial cell-line arrangements. The same class of culture has been used for years in neuroscience. What is new is the packaging, the commercial availability, and the explicit framing as computational infrastructure.
The technical claim is straightforward. Electrical signals are sent into the culture; the neurons respond; the responses are read out and can be shaped by feedback. Earlier research versions learned simple closed-loop tasks such as Pong. The company presents the approach as potentially more energy-efficient than pure silicon for certain adaptive or learning workloads, and as a platform for drug discovery and disease modelling. Relative to a human brain's roughly 86 billion neurons, or even to a modern GPU rack, the current systems are minute. They require continuous biological maintenance and have a finite lifespan measured in months.
None of this requires the language of monstrosities to be worth examining. The systems exist. They are small. They are still primarily research tools. Yet the trajectory is clear enough to raise concrete questions that are not resolved by the present scale.
First is provenance and consent. Public statements from the company indicate the cells come from consenting adult donors and standard iPSC lines. Independent reviews have noted that detailed, publicly available documentation tracing original donor consent specifically to commercial biocomputing use is thin. Broad research consent on a cell line is not the same as informed agreement that derivatives will be sold as computing substrates and trained on arbitrary tasks. When the product moves from academic collaboration to paid cloud access, the gap between original donation and downstream use widens. This is a governance issue more than a science-fiction one.
Second is the question of what, if anything, these cultures experience. Company scientists and collaborating bioethicists state that current cultures show no evidence of consciousness or the capacity for suffering, and that design choices aim to keep them below any such threshold. That is the responsible position given present evidence. It is also provisional. Neural cultures can exhibit goal-directed adaptation in closed-loop environments. As density, organisation, and training complexity increase, the empirical claim that "nothing is happening subjectively" will need continuous, independent measurement rather than repeated assertion. The fact that the substrate is living human neurons, rather than silicon or even mouse cells, does not by itself create consciousness; it does change the ethical stakes if evidence of experience ever appears. Treating the issue as permanently closed because today's 800,000-neuron cultures are simple is an assumption, not a demonstration.
Third is the regulatory vacuum. Existing frameworks for human tissue, stem cells, and medical devices were not written for devices whose purpose is computation and whose customers may be outside traditional biomedical institutions. Purchasers are required to have appropriate laboratory conditions and ethics approvals for generating cell lines, yet the broader questions: acceptable-use limits, decommissioning standards for trained cultures, data governance, and liability if a culture is later shown to have unexpected properties, lack clear, public rules. Commercial diffusion lowers the barriers that once kept such work inside a handful of specialised labs.
None of these points requires believing that today's Singapore rack is already a thinking entity or that data centres are about to be replaced by living tissue. They require noticing that a commercial product now places living human neurons into a computing role, that the product is being scaled in partnership with data-centre operators, and that the supporting institutions of consent, oversight, and measurement have not kept pace with the commercial step. Energy efficiency and research utility are real attractions. They do not automatically settle the provenance, monitoring, or boundary questions.
The technology is early. Its present capabilities are modest. Its maintenance burden is high. Those facts limit near-term impact. They do not eliminate the need to decide, while the systems are still small and countable, what documentation, measurement, and use restrictions will apply if the numbers grow by another order of magnitude and the customer base moves further beyond research institutions. The rack in Singapore is operational. The standards that should govern what happens next are not.
https://www.thefocalpoints.com/p/big-tech-has-officially-turned-data