- Blockchain is a digital ledger that records every transaction on a public network.
- Once a transaction is recorded, it cannot be changed or deleted.
- It is the foundation of Bitcoin, Ethereum, and thousands of other cryptocurrencies.
For three and a half decades, Arm's business model was simple: design the blueprints, let everyone else build the chips, and collect royalties on every processor sold. That's over.
At an event in San Francisco on Tuesday, CEO Rene Haas unveiled the Arm AGI CPU — a data center processor with up to 136 cores, built specifically for AI inference workloads. Taiwan Semiconductor will manufacture it. Meta is the launch customer.
The chip will be sold at roughly 50% gross margin, according to CFO Jason Child. Haas projected it will generate $15 billion in annual revenue by 2031, bringing total company revenue to $25 billion with $9 in earnings per share. For context, Arm made $4 billion in revenue in all of 2025.
"We may be under-calling that number," Haas said at the event.
Arm has long been called the "Switzerland of semiconductors" — a neutral party that licensed its chip architecture to everyone, including companies that compete with each other, without ever picking sides. Qualcomm, Apple, Amazon, and Nvidia all build on Arm designs.
Now Arm is making its own chip and selling it directly. Citi called it "the most significant shift in the company's history." The forecasts are "well above even the highest of speculated estimates," the bank said.
The risk: Arm is now, at least partially, a competitor to the very companies it licenses to. The company framed it as giving customers "choice" — they can still buy a license, or now they can just buy the chip. Whether that reassures its customer base remains to be seen.
Arm stock jumped more than 20% Wednesday. HSBC doubled its price target to $205. The stock had been trading around $135 before the announcement.
The AGI CPU taps into a real shift in demand: as agentic AI requires more continuous compute, CPUs are seeing a resurgence alongside GPUs. Haas predicted a fourfold increase in CPU demand from agentic AI alone.
