Chinese AI Labs Develop Custom Chips to Cut Costs, Face High Investment Risks

Chinese AI labs are ramping up their investments in developing their own chips, aiming to cut costs in the long run and enhance how well their hardware and software work together, which is a trend seen globally. Still, experts warn that this strategy comes with significant financial risks due to the hefty upfront costs involved. Arisa Liu, who is the chief director at Taiwan Industry Economics Services, pointed out that the main objective is to boost the synergy between hardware and software while lowering expenses over time. Paul Triolo from DGA-Albright Stonebridge Group highlighted in his AIStackDecrypted newsletter that leading Chinese developers now view silicon as a key part of their models. Meanwhile, Hangzhou's startup DeepSeek has been quietly recruiting chip-design talent without announcing job openings, as two insiders revealed. According to Reuters on Tuesday, DeepSeek's custom AI inference chip project started about a year ago and is still in its early phases, allowing for optimization tailored to their specific architecture, the DeepSeek-R1 model.