Arm Holdings plans to launch AI chips in 2025, Nikkei reports

[Daniel Nenni]

FILE PHOTO: Illustration shows Arm Ltd logo

(Reuters) -SoftBank Group's Arm Holdings plans to develop artificial-intelligence (AI) chips, seeking to launch the first products in 2025, Nikkei Asia reported on Sunday.
UK-based Arm will set up an AI chip division and aim to build a prototype by spring 2025, the report said. Mass production will be handled by contract manufacturers and is expected to start in the autumn of 2025, Nikkei Asia said.

Arm will pay for initial development costs, which may go up hundreds of billions of yen, with SoftBank also contributing, the report said.

Once a mass-production system is established, the AI chip business could be spun off and placed under SoftBank, the newspaper said, adding that SoftBank is already negotiating with Taiwan Semiconductor Manufacturing Corp (TSMC) and others over manufacturing, looking to secure production capacity.

Arm and SoftBank declined to comment, while TSMC did not immediately respond to requests for comment.

The British chip designer, which licenses its chip designs and earns funds through royalties, has been expanding into the data-centre market, where operators are looking to build their own chips to power new AI models and reduce their reliance on dominant supplier Nvidia.

SoftBank is expected to slip back into the red when it reports earnings on Monday. Investors are eagerly awaiting clues about new growth investments as the company has ample liquidity and can monetise its huge holding in Arm, whose share price roughly doubled in February on excitement over AI.

Arm Holdings plans to launch AI chips in 2025, Nikkei reports

UK-based Arm will set up an AI chip division and aim to build a prototype by spring 2025, the report said. Mass production will be handled by contract manufacturers and is expected to start in the autumn of 2025, Nikkei Asia said. Arm will pay for initial development costs, which may go up...

www.yahoo.com

 

[Daniel Nenni]

Competing with customers is never a good thing. Another reason to use RISC-V.

 

[blueone]

Competing with customers is never a good thing. Another reason to use RISC-V.

Agree, but for reasons I don't really understand (though I have theories), RISC-V seems to taking off nowhere I can find. I thought the relatively high cost of embedded Arm core licenses would at least visibly turn that market to RISC-V, but several people I've talked to on different industry projects say they are sticking to Arm. I've long suspected that open source hardware would have a more shallow adoption curve than open source software (for a lot of good reasons), but I'm thinking even my pessimism was too optimistic.

 

[Daniel Nenni]

Agree, but for reasons I don't really understand (though I have theories), RISC-V seems to taking off nowhere I can find. I thought the relatively high cost of embedded Arm core licenses would at least visibly turn that market to RISC-V, but several people I've talked to on different industry projects say they are sticking to Arm. I've long suspected that open source hardware would have a more shallow adoption curve than open source software (for a lot of good reasons), but I'm thinking even my pessimism was too optimistic.

Quite the opposite for me. I see RISC-V everywhere. Mostly in support roles and on the edge but I do know of RISC-V server chips in process.

Andes, Synopsys, and MIPS are now supporting RISC-V replacing custom architectures.

It really is something the entire world can get behind. A RISC-V tsunami is coming, absolutely.

 

[blueone]

Quite the opposite for me. I see RISC-V everywhere. Mostly in support roles and on the edge but I do know of RISC-V server chips in process.

Andes, Synopsys, and MIPS are now supporting RISC-V replacing custom architectures.

It really is something the entire world can get behind. A RISC-V tsunami is coming, absolutely.

Edge ASICs make sense. Less chance of legacy. I know about some server chips, but nothing especially impressive. The big cloud companies are still all Arm, and everything else is a rounding error by comparison.

 

[tooLongInEDA]

Competing with customers is never a good thing. Another reason to use RISC-V.

I'm not sure this necessarily is competing with customers.

Let's roll the clock back to the mid 1990s. Mobile phone companies like Nokia are just starting to integrate DSP, MCU and ASIC onto a single wireless application processor chip. TI (who produce the chip for Nokia) learns from engagements like this and spins off mass market generic wireless application processor chips for the rest of the market who don't have the resources to do it all themselves. Nokia is fully ware of all this but is happy they're getting a technical and time to market lead.

This feels very similar. The high end data centre customers (the usual suspects) continue to do their own thing. The mass market (which may be small here), needs bigger off the sheld building blocks. Who else but ARM could provide these ? ARM is climbing the food chain from IP to sub-systems just as EDA is currently moving upwards from chips to electronic systems.

And from a PR and stock market viewpoint, now ARM has tied itself to the AI mast. Or watch the stock price drop without the current AI premium.