Photonic neuron prototype promises energy‑efficient AI as US‑China rivalry intensifies

Researchers from Lund University and the University of Copenhagen have created a photonic neuron that combines several brain‑like functions in a single component. The study, led by Professor Anders Mikkelsen and Joachim E. Sestoft, notes that the human brain operates on about 20 watts—comparable to a LED bulb—while today’s large AI models require massive data‑centre power. If scaled, the light‑driven neuron could enable a new generation of low‑power AI hardware that processes information where it is generated.

At the same time, the global AI landscape is shifting as Chinese companies narrow the gap with the United States. Moonshot AI’s newly released Kimi K3 model matches or exceeds the performance of leading U.S. systems such as Anthropic’s Claude on key benchmarks, while using far less compute and cost. The development has prompted U.S. officials, including President Donald Trump, to threaten restrictions on access to Chinese AI models and consider sanctions over alleged intellectual‑property theft. The rapid progress of Chinese open‑source AI platforms such as DeepSeek, Alibaba’s Qwen and Xiaomi’s MiMo underscores the escalating competition between the two superpowers.