Japanese startup Preferred Networks is increasing its investment in custom artificial intelligence chips to secure crucial hardware access amid the global investment boom driven by advancements in generative AI.
Backed by investors such as automaker Toyota and robot maker Fanuc, the company initiated the development of its first-generation AI chip in 2016 for its supercomputers.
“We knew we needed to optimise energy consumption and minimise procurement risk, which could make it difficult to sustain our business,” Preferred Networks co-founder and CEO Toru Nishikawa said in an interview.
With the growing enthusiasm for AI technology, companies globally are rushing to obtain AI chips. Giants like Amazon and Microsoft are among those developing chips in-house.
Preferred Networks, in response to this trend, has finalized the design of its second-generation AI chip. TSMC is handling the manufacturing process, and this chip is set to empower the startup’s upcoming supercomputer.
Preferred Networks has fine-tuned its chips for AI applications, achieving lower power consumption and enhanced computing capabilities. This optimization is achieved, in part, by shifting functions typically executed by hardware to software.
The startup aims to roll out its cutting-edge technology for crafting extensive language models and facilitating drug discovery in the coming year. By 2027, it plans to provide customers with substantial computing power.