Top Chinese companies are training their artificial intelligence models overseas to access Nvidia’s chips and bypass US efforts to prevent their development of the powerful technology, reports the Financial Times. Alibaba and ByteDance are among the tech groups training their latest large language models in data centres across south-east Asia, according to two people with direct knowledge of the matter.
These people said there had been a steady increase in training in offshore locations after the US in April moved to restrict sales of the H20, Nvidia’s China-only semiconductors.
Over the past year, Alibaba’s Qwen and ByteDance’s Doubao models have become among the top-performing LLMs worldwide. Qwen has also become widely adopted outside China by developers as it is a freely available “open” model. Data centre clusters have boomed in Singapore and Malaysia, fuelled by Chinese demand. Many of these data centres are equipped with high-end Nvidia products, similar to those used by US Big Tech groups to train LLMs.