Singapore's Neuron Rack Costs More to Rent Than It Saves (julienreszka.com)

Singapore recently launched a data center rack featuring sixteen million living human neurons, designed by NUS Medicine, DayOne, and Cortical Labs. While it draws significantly less power—approximately one kilowatt—compared to traditional silicon AI servers that can consume up to one hundred kilowatts, the cost of renting this neuron-based setup raises questions. Access to the rack's twenty CL1 units costs $44,000 per month, resulting in an annual expenditure of $528,000, far exceeding the estimated $190,800 in annual savings on electricity. This discrepancy underscores the complexities of evaluating operational costs in biological computing compared to silicon alternatives. The implications for the AI/ML community are profound, as they reveal that while biological systems may offer energy efficiency, the associated operational costs involving specialized labor and maintenance can be prohibitively high. Each neuron culture must be regrown every six months by trained biologists, alongside expenses for nutrient replenishment and gas exchanges. Furthermore, the price comparison made by Cortical Labs between CL1 units and cloud AI chips lacks transparency, as there is no definitive benchmark demonstrating equivalent performance per dollar spent. As Cortical Labs navigates the challenge of scaling its technology, the long-term viability of its pricing strategy remains uncertain, raising critical questions about the future of biological computing in AI applications.