Key Takeaways
1. Cortical Labs has created the world’s first biological data centers using living human brain cells embedded in silicon chips, consuming significantly less power than traditional GPUs.
2. The core technology, called CL1, consists of 800,000 neurons from human stem cells, maintained in a nutrient-rich environment and capable of functioning for up to six months.
3. Each CL1 unit costs about $35,000 and a rack of units uses only 850-1,000 watts, compared to 6,000 watts for a single Nvidia GPU.
4. The neurons can autonomously adapt and learn from stimuli, showcasing capabilities in playing complex games through a system called Synthetic Biological Intelligence (SBI).
5. Cortical Labs is expanding its operations with a new facility in Singapore and offers access to its technology through a Wetware-as-a-Service model for researchers worldwide.
Cortical Labs, a biotech startup, has introduced what it labels as the first biological data centers in the world. Rather than using traditional power-hungry Nvidia GPUs, these centers utilize living human brain cells embedded in silicon chips, which require as little as 30 watts per unit.
The Core Technology
At the core of these data centers is the CL1, a “wetware” biological computer that was initially revealed at MWC 2025. Each CL1 unit contains around 800,000 neurons cultivated from human stem cells in a laboratory. These neurons are sustained by a nutrient-rich fluid, while an array of electrodes provides electrical input and gathers their reactions. The display of the CL1 shows the health of the neurons, with the system regulating temperature, gas composition, and waste removal to maintain the viability of the human brain cells for up to six months.
Pricing and Power Efficiency
The CL1 functions as a standalone computing system, priced at about $35,000, and an entire rack of such units consumes only between 850 and 1,000 watts. This is a striking contrast to the typical 6,000 watts consumed by a single Nvidia GPU in conventional AI data centers. With energy prices soaring and AI systems consuming vast amounts of power, the electrical grids are under pressure. Cortical Labs is marketing its biological computing as a significantly more sustainable alternative compared to traditional silicon-based GPUs.
Innovative Learning Capabilities
Interestingly, the neurons don’t require traditional training; they can adapt and rewire themselves in response to different stimuli—an ability that has benefited humanity for ages. For instance, the company’s earlier model, DishBrain, learned to play simple games like Pong autonomously. The latest CL1 has already shown the ability to tackle more complex games, such as Doom, by itself, as shown in the video below.
Cortical Labs refers to this blended method as Synthetic Biological Intelligence (SBI) and operates it through a unique Biological Intelligence Operating System (biOS). The software facilitates interaction between the neurons and the tasks assigned to them. Researchers can remotely access the system through the newly launched Cortical Cloud, which operates under an innovative model named Wetware-as-a-Service. For just $300 per week per unit, scientists around the world can deploy and evaluate their code directly on living neural networks and assess their effectiveness.
Future Developments
Cortical Labs has established its first facility in Melbourne as a proof of concept. However, it is in the process of constructing a significantly larger facility in Singapore in collaboration with the local DayOne Data Centers provider. While the company’s two-dimensional neuron layers are quite basic compared to the complex structure of an actual brain, these biological data centers represent an intriguing venture into wetware organic computing that leverages neurons instead of transistors.
Source:
Link
Leave a Reply