Japanese Startup Preferred Networks Accelerates Investment in Customized AI Chips

Published: December 20, 2023 at 4:09 am Updated: December 20, 2023 at 4:09 am

Edited and fact-checked:
December 20, 2023 at 4:09 am

In Brief

Preferred Networks increases investment in customized AI chips, aiming to secure access to hardware amid the surge in global AI investments.

Japan's Preferred Networks Accelerates Investment in Custom AI Chips

Japan-based startup Preferred Networks is increasing its investment in customized artificial intelligence (AI) chips, aiming to secure access to crucial hardware amid the surge in global investment driven by advancements in generative AI.

Preferred Networks specializes in chips optimized for various AI tasks, boasting lower power consumption and enhanced computing power achieved, in part, by transferring functions typically handled by hardware to software.

The startup has finalized the design of its second-generation AI chip, currently in production by TSMC, and set to power its upcoming supercomputer.

“We recognized the need to optimize energy consumption and minimize procurement risk, factors crucial for sustaining our business,”

said Toru Nishikawa, co-founder and CEO of Preferred Networks.

The company, with investors including automaker Toyota and robot maker Fanuc, initiated the development of its first-generation AI chip in 2016 for powering its supercomputers.

Preferred Networks intends to make its latest technology available for the creation of large language models (LLMs) and drug discovery in the coming year, with plans to offer pure computing power to customers by 2027.

Companies are Delving into In-House AI Chip Creation

There’s a lack of GPUs due to the increasing demand for generative AI, which commonly uses GPUs for training and operations. To reduce dependence on GPUs, well-funded firms, especially tech giants, are creating custom chips designed for developing and deploying AI models, sometimes offering them to customers.

Amazon and Microsoft are among those developing chips internally.

In a recent development, AWS unveiled two chip families—AWS Graviton4 and AWS Trainium2—designed in-house for training foundation models and inferencing.

Microsoft also introduced its custom AI chips, Azure Maia 100 and Cobalt 100, specifically designed for the company’s cloud infrastructure. These silicon chips are intended to drive Microsoft’s Azure data centers and are scheduled for release in 2024.

The active pursuit of custom AI chips by companies worldwide reflects a strategic response to the global demand of tech hardware.

Disclaimer

In line with the Trust Project guidelines, please note that the information provided on this page is not intended to be and should not be interpreted as legal, tax, investment, financial, or any other form of advice. It is important to only invest what you can afford to lose and to seek independent financial advice if you have any doubts. For further information, we suggest referring to the terms and conditions as well as the help and support pages provided by the issuer or advertiser. MetaversePost is committed to accurate, unbiased reporting, but market conditions are subject to change without notice.

About The Author

Alisa is a reporter for the Metaverse Post. She focuses on investments, AI, metaverse, and everything related to Web3. Alisa has a degree in Business of Art and expertise in Art & Tech. She has developed her passion for journalism through writing for VCs, notable crypto projects, and engagement with scientific writing.

Alisa Davidson