AI-pilled Arm CEO teases mystery products that will turn it into a money machine

Summary

Arm CEO Rene Haas used the Arm Everywhere keynote to hint at undisclosed products he says could expand the company’s total addressable market (TAM) to $1 trillion by the end of the decade. The event put the spotlight on Arm’s newly announced AGI CPU family — including a 136-core part developed with Meta — which is intended to run agentic AI workloads and let Arm sell branded datacentre silicon directly to customers instead of relying solely on IP licensing and royalties.

Arm believes agentic AI frameworks will significantly increase CPU core demand because agents run on CPU cores to orchestrate and execute model-generated code, creating more persistent, multi-agent workloads. Haas predicts Arm’s datacentre TAM could reach about $100 billion as a result.

The AGI CPU has already attracted high-profile customers — Meta, OpenAI, SAP, Cloudflare and SK Telecom — but the market remains intensely competitive with Nvidia, Intel and AMD all pushing into CPUs and dense core counts. Arm plans a cadence of new AGI CPUs, with follow-ups arriving as early as next year and a third-generation design already in development.

Key Points

Rene Haas teases unnamed products and projects a $1tn TAM for Arm by the end of the decade.
Arm unveiled an AGI CPU (136 cores) co-developed with Meta, aimed at running agentic AI workloads in datacentres.
Arm expects agentic AI to multiply CPU core demand because agents consume CPU and memory to run and coordinate tasks on top of specialised model accelerators.
Arm predicts its datacentre TAM could grow to about $100bn as agentic systems scale.
Early AGI CPU customers include Meta, OpenAI, SAP, Cloudflare and SK Telecom.
Competition is fierce: Nvidia, Intel and AMD are all expanding CPU offerings and core counts, and many AI shops use multiple silicon suppliers.
Arm will continue rapid chip development — updates as soon as next year and a third-generation AGI CPU already in the works.

Context and relevance

This matters because Arm is signalling a structural shift: from being primarily an IP licensor to also selling branded datacentre silicon. That pivot could reshape revenue streams and partnerships across cloud and AI supply chains. Agentic AI — systems that orchestrate other models and code — creates a different workload profile: specialised accelerators will still run the heavy model inference, but CPUs will shoulder the orchestration, execution and memory-intensive parts of agent deployments. That combination could materially change how datacentre capacity is bought and architected.

For cloud operators, hyperscalers and companies building AI infrastructure, Arm’s move is a reminder that silicon choice is diversifying. Pre-launch customer commitments suggest traction, but broad market adoption will depend on performance-per-watt, ecosystem support, and how enterprises balance multi-vendor strategies versus consolidation around a few suppliers.

Author’s take (punchy)

Arm’s CEO just waved a very big flag: sell direct, design for agentic AI, and scale the datacentre play. It’s bold — and it forces hyperscalers and incumbents to think beyond legacy licensing models. If Arm delivers efficiency and ecosystem proof, it won’t just get royalties; it’ll be competing for hardware spend at scale.

Why should I read this?

Want to know who’s angling to grab a chunk of the next wave of AI cash? This is the move. Arm’s pivot to branded datacentre silicon and its bet on agentic AI could change who makes the money when AI scales — and who you pick for future infrastructure. We’ve skimmed the keynote and pulled out the bits that matter, so you don’t have to sit through the marketing fluff.

Source

Source: https://go.theregister.com/feed/www.theregister.com/2026/03/24/arm_ceo_boasts_new_silicon/