Arm CEO: $15 Billion AI Chip Revenue Target May Be Achieved Ahead of Schedule as ByteDance Joins Customer Roster
Taylor Wilson
Arm CEO Rene Haas said at Computex that AI demand is outpacing forecasts, and the company's $15 billion chip sales target could be reached ahead of schedule; ByteDance and Oracle joined as new customers the same day, signaling Arm's pivot from licensor to chipmaker is gaining real traction.
Why might the $15 billion target come early?
Haas previously said he was "very confident" of hitting it by the end of the decade. Now he's saying he "hopes to get there sooner."
This means → the upgrade from "on track" to "ahead of schedule" reflects data-center and AI infrastructure buildouts running faster than Arm's own models predicted.
His words were blunt: "Demand is better than we expected."
Why is Arm making its own chips?
Arm's legacy business was selling blueprints — licensing chip architectures to Qualcomm, MediaTek, and others, never touching manufacturing.
In March, Arm announced its first proprietary chip, the AGI CPU: up to 136 cores, 300 watts, manufactured by TSMC.
In plain terms = a cookbook publisher decided to open its own restaurant — the revenue ceiling is completely different, but so is the risk.
The company expects chip sales to eventually surpass its existing IP-licensing revenue.
Who is buying this chip?
The first major customer is Meta. On the same day, Arm announced two more: ByteDance and Oracle.
This means → in eight weeks the roster went from one marquee buyer to three, spanning top-tier players in both the US and China. The pace of pipeline expansion is itself a demand signal.
Haas said data-center CPU demand is stronger than it was eight weeks ago, which is why the company could announce back-to-back major customers.
Why is regulating AI CPU exports so hard?
Haas told Reuters that a US ban on AI-capable CPU exports to China would be nearly impossible to enforce.
The reason: GPUs can be gated by performance thresholds and memory bandwidth — that is how Nvidia was restricted. But CPUs are used in too many applications to draw a clean line.
His analogy: "CPUs are like oil — they permeate every use case." Enforcing a ban would mean "restricting everything," which he called "a pretty heavy-handed blunt instrument."
This reflects a natural boundary in chip-export controls: the more general-purpose the compute unit, the harder it is to regulate precisely.
Content is for reference only, not financial advice.