China's new 'brain-like' AI server to consume 90% less power than others

China’s Guangdong Institute of Intelligence Science and Technology (GDIIST) has unveiled the BI Explorer computing system (BIE-1), a ‘brain-like intelligent computer’ that reportedly uses 90 per cent less power than traditional supercomputers, according to the South China Morning Post.

 

Chinese scientists said the BIE-1, about the size of a mini fridge, can match the power of a room-sized supercomputer. It achieves high training and inference speeds using a brain-like artificial intelligence (AI) system and an intuitive neural network.

Power shortage concerns

The announcement comes amid growing concerns over the massive energy demand driven by the rapid adoption of artificial intelligence across industries. As AI-powered data centres consume increasing amounts of electricity, technology giants are reportedly exploring nuclear energy as a sustainable solution to meet future power needs. 

 

 

AI data centres can require over 2.2 gigawatt per year, enough to power a million homes. The spike in power consumption is already straining electricity grids in the US and Europe.

Key features of BIE-1

The BIE-1 is seen as a major breakthrough, cutting power consumption by nearly 90 per cent, a key challenge in high-performance computing.

• Compact design: Roughly the size of a mini single-door refrigerator, making it easy to deploy in homes, offices, or even mobile environments.
• Energy-efficient: Uses just one-tenth the power of a traditional supercomputer and operates on a standard household socket.
• Quiet operation: Runs with minimal noise, keeping CPU temperature below 70°C even during intensive tasks.
• High performance: Equipped with 1,152 CPU cores and 204 TB of storage.
• Advanced neural network: Features a brain-inspired AI algorithm capable of efficient learning, interpretable reasoning, and pattern recognition from limited data.
• Multimodal processing: Can simultaneously handle text, images, and speech inputs.
• Versatile applications: Suitable for home health monitoring, personalised tutoring, office AI assistance, and other intelligent computing uses for non-professionals.

Demand for nuclear energy

Amid growing concerns over power shortages and environmental commitments, leading technology companies, including Amazon, Microsoft, Google, and Meta, are exploring more reliable energy sources. While solar and wind power have expanded rapidly, they remain insufficient to meet the 24×7 energy demands of AI-driven data centres. As a result, several tech giants are turning to nuclear energy, which offers carbon-free power with greater stability and reliability.

 

According to a report by Nikkei Asia, Meta, Google, Amazon, and Microsoft plan to procure more than 14 million kilowatts of electricity from nuclear power plants by 2040.