Arm Stops Renting Out Its Genius and Starts Keeping the AI Money

For decades, the silicon architect behind NvidiaNVDA

and Qualcomm’sQCOM

chips tinkered in the shadows while others took the glory. Not anymore. Now, ArmARM

is done sharing the spoils, launching its own chip into direct competition with former customers, and investors approved it with a 16.4% surge on Wednesday.

• According to Citigroup, Arm isn’t testing the waters and “has jumped in with both feet” — debuting a full-blown data center CPU built to run AI agents efficiently at scale.
• Financially, Arm expects this launch alone to generate roughly $15B in annual revenue within five years — with total sales reaching $25B and EPS hitting $9 during that period.

Bold timing: As agentic AI shifts from one-off chatbot replies to always-on “agents” juggling multi-step tasks, CPUs have quietly come back into focus. MetaMETA

co-designed Arm’s AGI CPU, while OpenAI, CloudflareNET

, and SAPSAP

have already lined up as customers, all looking to reduce their reliance on Nvidia. Still, with Arm already trading at a premium to peers like AMDAMD

, this new chip will have to justify more than just a good first headline.