ARM chips are moving from smartphones into humanoid robots, where they handle real-time AI inference without relying on cloud connections. Figure AI's recent livestream demonstrated four humanoid robots sorting over 47,000 packages in 38 hours using local neural network processing on NVIDIA Jetson Thor modules, which contain ARM Neoverse V3AE CPU cores. This shift signals a major new revenue stream for ARM beyond traditional smartphone royalties, as industrial robotics scales up to meet labor shortages.

Why Can't Humanoid Robots Use Cloud AI Like Smartphones Do?

The technical answer reveals why ARM's role in robotics differs fundamentally from its smartphone business. A humanoid robot balancing on two legs has roughly 10 milliseconds before a perturbation causes it to fall, but wireless latency to a cloud data center ranges from 20 to 200 milliseconds in optimistic conditions. This gap makes cloud inference impossible for closed-loop physical control systems.

Beyond latency, bandwidth creates a second barrier. Stereo cameras at production framerate generate gigabits per second of raw visual data. Streaming that to a data center in real time would saturate any warehouse's wireless network, and compressing the stream hard enough to fit erases the resolution the AI model needs to function. Reliability finishes the argument: warehouses lose connectivity routinely, and a robot that stops working when the network drops is worse than hiring a human worker.

How Are ARM Chips Powering the Next Generation of Robot Hardware?

The NVIDIA Jetson AGX Thor module, which shipped commercially in 2026, represents the reference platform major humanoid programs are designing around. Inside the package sits 14 ARM Neoverse V3AE CPU cores running up to 2.6 gigahertz, paired with a Blackwell-class GPU delivering 2,070 FP4 teraflops, all within a 130-watt power envelope. This architecture mirrors the data-center pattern already familiar to investors: ARM cores handling control logic, while NVIDIA silicon manages parallel inference.

Figure AI's Helix-02 neural network demonstrates this in practice. The unified network controls walking, manipulation, and balance from raw sensor data, replacing over 109,000 lines of hand-coded locomotion logic with a single set of weights. Motor control runs at 200 hertz, while scene understanding runs at 7 to 9 hertz, with both loops staying within millisecond latency budgets.

• CPU Role: ARM Neoverse cores handle real-time control logic and system coordination, replacing traditional hand-coded robot behaviors with learned neural network weights
• GPU Role: NVIDIA Blackwell-class processors execute parallel inference for vision, manipulation, and balance calculations at production speed
• Power Efficiency: The entire Jetson Thor module operates within a 130-watt power envelope, making it practical for robots that must work for extended shifts
• Latency Guarantee: Local processing eliminates the 20 to 200 millisecond cloud round-trip delay that would cause physical robots to fall or fail

What Does This Mean for ARM's Business Model?

The investment relevance is straightforward. A warehouse running ten humanoid robots in 2027 needs ten Jetson Thor-class modules, plus spares, plus development kits for testing. A logistics network operating ten thousand humanoids needs ten thousand. The math compounds quickly as pilot programs transition to fleet deployments, and the demand path is real. The US industrial sector is projected to need 3.8 million new workers by 2033, with nearly 1.9 million roles at risk of going unfilled.

Current consensus models for ARM and NVIDIA value the chips business off data-center capital expenditure and smartphone royalty volumes. The edge inference tier inside physical AI is not in those models in any material way, because the deployments have not yet reached scale. Figure's livestream is the first commercial-grade signal that deployments are imminent, not theoretical.

"Every Jetson Thor module pays ARM a per-unit royalty on the V3AE cores. Every alternative robotics system-on-chip built on Neoverse pays the same. The volume curve is not the slow replacement cycle of phones; it is the build-out curve of an industrial labor substitute," explained Jon Markman, Forbes contributor analyzing the chip stack.

Jon Markman, Forbes Contributor

For ARM specifically, this opens a royalty line that compounds differently than the smartphone business. Every Jetson Thor module sold generates a per-unit royalty on the V3AE cores. Every competing robotics system-on-chip that uses Neoverse cores does the same. The volume curve is not the slow replacement cycle of phones; it is the build-out curve of an industrial labor substitute.

Is ARM's CPU Strategy Shifting Across the AI Pyramid?

The architectural pattern extends beyond robotics. The same CPU-plus-GPU pairing that powers data-center training now powers edge inference in physical AI systems. The Vera Rubin platform packages the data-center version, while Jetson Thor packages the edge version. The compute layer is converging on the same two companies, ARM and NVIDIA, at both ends of the AI pyramid.

Tesla represents the visible exception. The AI5 chip Tesla taped out in April is custom silicon designed in-house for both Optimus and Cybercab. Musk has benchmarked a single AI5 die against an NVIDIA H100 for Tesla's specific workloads, with mass production targeting mid-2027. Tesla's vertical-integration play means Optimus is unlikely to ship on Jetson Thor. But Tesla is also the only credible humanoid program with the silicon capability to leave the standard stack. Everyone else is buying.

The companies positioned in the edge inference stack for physical AI are not new names. They are the same names sitting at the top of the AI pyramid's chips layer. NVIDIA is the platform owner, extending the same architectural advantage from data-center training into onboard robot inference. ARM is the CPU layer, with every Jetson Thor sold paying a royalty on Neoverse cores. Cadence sits underneath both, providing the chip design tools used to build Jetson Thor and competing alternatives.

Watch the Figure announcements on commercial fleet sales over the next two quarters. Watch the pace of Jetson Thor adoption disclosed by humanoid platforms that are not Tesla. Watch the ARM royalty mix in earnings reports as the edge tier shows up. The signal will come from chip royalties and shipment commentary, not from headline counts of robot deployments.