Qualcomm threw itself into the humanoid robotics race on Thursday, unveiling a dedicated computing platform at Computex in Taipei designed to do for robots what the chipmaker spent a decade doing to smartphones and cars: own the platform layer.
Chief executive Cristiano Amon used the keynote to position the move as the logical next step in what Qualcomm calls "agentic AI" — intelligence that acts in the physical world. Smartphones, PCs, connected vehicles each demanded a new computing paradigm, he argued. Robots are next.
Why Robots Are a Harder Problem
Amon was direct about the engineering challenge. A functional robot cannot simply be a language model in a moving chassis. It needs the perceptual sophistication of a smartphone, the AI compute density of a PC and the real-time control expertise the automotive industry took decades to build — running all at once.
He described three computing layers every capable robot requires: a base layer for reflexive functions such as balance; a middle layer for task execution and environmental awareness; and a top layer for reasoning and planning. "Robots are pushing the limits of computing," Amon said. "Unlike a phone or laptop, a robot must not only understand its environment — it must physically interact with it."
The Dragonwing IQ10: One Platform, No Assembly Required
The centrepiece of the launch is the Dragonwing IQ10 Robotics Reference Design — a full-stack platform that bundles compute, sensor interfaces, connectivity and real-time control into a single architecture. The pitch to developers: stop integrating mismatched components and start building applications.
The processor delivers up to 700 TOPS of AI performance, backed by an 18-core Oryon CPU, a multi-core neural processing unit and a GPU built for perception and inference. It supports up to 12 GMSL2 cameras natively, integrates LiDAR, time-of-flight sensors and IMUs, and carries the industrial protocols — EtherCAT, CAN-FD — that factory environments demand.
From Arms to Humanoids: A Broad Ambition
Qualcomm is not targeting one robot category. Its portfolio spans autonomous mobile robots, industrial arms, quadrupeds, humanoids and drones — all intended to run on the same platform foundation.
In a post-keynote briefing, Nakul Duggal, the company's general manager for automotive, industrial and robotics, said Qualcomm has begun developing technology specifically for arm control, end effectors and humanoid systems. A robot, he noted, is more complex than a car: one system must simultaneously manage perception, navigation, balance, manipulation and fine motor precision. "We are investing in models, full embodiment systems, and different form factors of robots," Duggal said.
The Automotive Playbook, Applied to Robotics
Launch partners — including NEURA Robotics, Advantech, Innodisk and Thundercomm — signal the same ecosystem-first strategy that drove Qualcomm's automotive expansion. The company built its automotive business not by selling chips to carmakers but by creating a platform that Tier 1 suppliers could build on. It is running the same play here.
Amon also signalled a deeper partnership with Figure AI, the closely watched U.S. humanoid startup, without disclosing specifics.
The Gap Between Lab and Factory Floor
Amon argued that the industry's binding constraint is no longer demand — logistics, manufacturing and services are hungry for automation. The real question is whether robots can be deployed reliably and at commercially viable cost. Qualcomm is betting the Dragonwing IQ10 closes that gap. With generative AI accelerating robotic perception and reasoning, the sector is moving from proof-of-concept toward early commercialisation — and Qualcomm wants to be the platform it runs on. （Related： ZHITAI at COMPUTEX: Targeting Gamers and High-Performance Creators with Flagship SSD Launches Across the APAC Region ｜ Latest ）
