Jakarta, INTI – The global dependence on U.S.-made chips, particularly those from Nvidia, has long posed a major challenge in the development of artificial intelligence (AI). While many large language models (LLMs) demand high energy consumption and powerful hardware, researchers in Beijing have introduced a new solution. They unveiled SpikingBrain 1.0, an AI system designed to mimic the human brain, operate with minimal energy, and run entirely on hardware made in China. The release was announced by the Chinese Academy of Sciences’ Institute of Automation as part of the launch of a neuromorphic research project held in mid-September 2025 in Beijing.
Inspired by the Human Brain Mechanism
SpikingBrain 1.0 adopts a spiking computation approach, a method that replicates how biological neurons function. Instead of activating an entire network as conventional AI systems do, this model sends signals only when specific inputs trigger a response. This design allows most of the network to remain silent, thereby conserving energy while accelerating data processing.
Researchers emphasized that inspiration came from the efficiency of the human brain which operates on only about 20 watts of power. As a result, SpikingBrain 1.0 can perform certain tasks up to 100 times faster, even when trained on less than 2 percent of the data typically required by conventional models.
Testing and High Performance
In testing, the research team developed two versions of SpikingBrain 1.0: a smaller model with 7 billion parameters and a larger one with 76 billion parameters. Both were trained on around 150 billion tokens of data, a relatively modest amount for models of this scale.
When tested on a dataset containing 4 million tokens, the smaller model proved to be over 100 times faster than standard systems. Furthermore, in generating the first token from a 1-million-token context, SpikingBrain 1.0 demonstrated a 26.5-fold speed increase compared to conventional transformer architectures.
Moreover, the system remained stable for weeks when deployed on hundreds of MetaX chips, a platform developed in Shanghai, showcasing its practical capability for industrial applications.
Real-World Application Potential
SpikingBrain 1.0 holds significant potential for fields requiring the processing of large and complex datasets. Its key applications include:
- Analysis of lengthy legal and medical documents
- High-energy physics research
- DNA sequencing and biotechnology
These advantages signal that the future of AI computing may no longer rely solely on Nvidia, while also paving the way for China’s technological independence in critical sectors.
Conclusion
The launch of SpikingBrain 1.0 is not just another AI innovation but also a strong indication that Nvidia’s dominance in the chip market could begin to face serious competition. With its energy efficiency, high processing speed, and reliance on domestic hardware, this brain-inspired AI offers a new direction for the future of global technology development. If widely adopted, SpikingBrain 1.0 could reshape the landscape of the AI industry worldwide.
Read More: SensAI Strengthens the National CX Ecosystem Through AI Innovation
Author: Nd