Cortical Labs builds brain cell data centers to tackle AI energy consumption crisis

Cortical Labs Launches Brain Cell-Powered Infrastructure

Australian biotech startup Cortical Labs is building two biological data centres that mark a significant shift in how computing infrastructure could be designed. The company announced a partnership with DayOne Data Centers to establish facilities in Melbourne and Singapore, both powered by its CL1 biological computer units3

3

. The Melbourne facility will house 120 CL1 units, while the Singapore location is set to deploy as many as 1,000 units in phases2

2

. Each CL1 unit consists of a silicon chip with approximately 200,000 lab-grown human neurons grown on top, converted initially from human blood cells2

2

. These human brain cells respond to electrical stimuli, forming networks similar to neural arrays in our own brains.

Energy-Efficient AI Alternative Addresses Power Crisis

The timing of this development comes as AI energy consumption has become a critical concern for tech companies and governments alike. Traditional AI data centers consume thousands of watts per chip and require substantial cooling infrastructure, but Cortical Labs argues that each CL1 requires around 30W1

1

. Founder and CEO Hon Weng Chong stated that each unit uses less power than a handheld calculator3

3

. Paul Roach at Loughborough University noted that when scaled up to whole rooms, there could be huge power savings compared to conventional data servers1

1

. The swift buildout of AI data centers across the planet has led to environmental concerns over their power needs and water consumption as well as shortages in silicon3

3

.

Brain Computing Services Expand Through Cloud Access

Cortical Labs has revealed that these biological data centres will allow the company to expand their cloud-based brain computing services1

1

. Michael Barros at the University of Essex, who already uses Cortical Labs' cloud services as part of his research, explained that biological computers like the CL1 are often hard to build and not easy for others to use1

1

. The company is essentially allowing its biocomputer to be accessible at a large scale, making it the first to do so1

1

. The Singapore project will start with an initial 20 CL1 units in a validation phase at the Yong Loo Lin School of Medicine at the National University of Singapore4

4

.

Programmable Wetware Platform Demonstrates Learning Capability

The neurons used by Cortical Labs, grown from stem cells, sit on a chip that sends and receives electrical signals to the cells and records how they respond3

3

. This programmable wetware platform allows the company's software to interact with the cells and interpret their responses as computing output. Cortical Labs has demonstrated that its neuron-filled chips can play Pong and Doom, with the company showing last month that the CL1 could learn to play Doom in a week1

1

. The system is equipped with on-board life support, recording, and application handling, and can keep the neurons alive for up to six months4

4

. Each unit is priced at around $35,0002

2

.

Early Stage Technology Faces Significant Challenges

Despite the promise, experts caution that the technology remains in early development. Steve Furber at the University of Manchester stated it's a very long way from production ready, noting it's a very big step from a small network playing a computer game to an LLM1

1

. The exact way in which these neurons function and how best to train them to perform tasks like machine learning is still unclear, according to Reinhold Scherer at the University of Essex1

1

. One remaining issue is that it's still unclear how to store the results of training these neurons in a form of memory or how to run actual computational algorithms on them1

1

. Another challenge involves retraining: whatever the neurons are trained on is lost when the culture ends its life, requiring complete retraining every 30 days1

1

. The proposed biological data centre will have hundreds of biological chips, compared to hundreds of thousands of GPUs seen in the largest AI data centers1

1

. These biological data centres are not being positioned as direct replacements for traditional server farms, at least not yet, but could carve out a niche where adaptability and lower energy use matter more than raw conventional performance4

4

.