Arm is getting ready to flog its own AI processor to Meta and OpenAI
The SoftBank-backed outfit believes the chip will drive a fivefold jump in revenue over the next five years.
Shares in Arm popped after chief executive Rene Haas unveiled the forecasts in San Francisco, shifting the firm from chip designer-for-hire to building its own “AGI CPU”.
That puts Arm on a collision course with data centre incumbents Intel and AMD, plus a few of its own licence-paying mates, including Nvidia, Google and Amazon.
Arm expects $25bn (€21.6bn) in annual revenue in five years, roughly five times today’s sales, Haas told the FT. He said the chip was a $15bn (€13.0bn) “revenue opportunity” by 2031, with growth helped by higher royalty rates and chunky licensing fees.
The stock rose as much as 7.5 per cent in after-hours trading following the forecast, after slipping 1.4 per cent during the day.
Haas said: “With the expansion into delivering production silicon with our Arm AGI CPU, we are giving partners more choices, all built on Arm’s foundation of high-performance, power-efficient computing.
The long-anticipated Arm CPU is a big step away from playing a “neutral” platform, where its IP gets baked into chips designed by US tech giants.
Arm says the part is aimed at the power bill problem in AI data centres, promising billions in savings compared with traditional CPUs.
Despite the AI label, it is not trying to take Nvidia’s GPU crown. It is pitching itself for “orchestration” work, running fleets of AI agents such as Claude Code and OpenAI Codex, as well as other cloud AI jobs.
The chip is being made by Taiwan Semiconductor Manufacturing Company, the same supplier used by Nvidia, the Fruity Cargo Cult Apple and other Arm licensees, with shipments due at the end of this year.
Meta is the “lead partner” with early customers including OpenAI, Cloudflare, SAP, SK Telecom and Cerebras, which signed a $10bn (€8.6bn) infrastructure deal with OpenAI in January.
Analysts reckon the move rewires Arm’s business model because selling chips should bring in far more cash than today’s licensing and royalty take.
The snag is margins. Arm’s gross margin hit 98 per cent in its most recent quarter, and hardware tends to ruin that sort of party, though Haas says the AGI CPU still clears more than 30 per cent operating margin.
Arm is expected to deliver about $4.9bn (€4.2bn) in revenue for the year to end-March, according to CIQ estimates.
Arm said dozens of companies were “supporting the platform expansion”, including Amazon Web Services, Google and Nvidia. Haas said Arm had reached out to customers, got “no pushback” and reckoned broader data centre adoption helps their software businesses.
Data centre CPUs have been dominated by Chipzilla and AMD on x86, while Arm’s architecture grew up in mobiles and now sits inside more than 325bn devices, from cars to servers.
Investor enthusiasm for AI infrastructure has cooled this year amid worries about whether the capex bonfire is sustainable, while geopolitical tensions between Washington and Beijing have messed with Nvidia’s China plans.
Haas said: “There is no reason, as far as we can tell, the new Arm CPU could not be sold in China… They don’t fall under any export control restriction, so there isn’t any issue there… although we don’t have any customers yet in China.”
He said it would “be a shame” if the UK’s AI infrastructure build-out ignored CPUs developed by its national champion and hinted he has been bending ministers’ ears about it.
