ASUS, the Taiwanese PC and server maker, has shown Gallop, its first humanoid robot, which runs on a single Nvidia Jetson AGX Thor computer and is learning to pack motherboard accessories at an ASUS group factory, according to a post on the ASUS Pressroom published on 13 November. Co-CEO S.Y. Hsu introduced the robot on CNBC, the post says.
The target ASUS sets is 99.5% success within 20 seconds per board, the standard it holds a person to. At 20 seconds a board, one station handles 180 boards an hour and 1,440 in an eight-hour shift, so 99.5% still leaves about seven missed boards a shift, by our arithmetic.
Business Pill · Physical AI
An evergreen Business Engineer explainer on physical AI, when AI gets a body. It teaches the general idea only and does not discuss ASUS or Gallop.
The key insight: As we read the post, ASUS is defining success the way a factory does, by a pass rate and a cycle time at one real station, rather than by a range of demonstrated tasks; the post gives the target, not a current result.
ASUS says customer pilots are planned for 2027 and market availability is targeted for the second half of 2028. The programme started in early 2026, and over the coming year ASUS plans to add its own dexterous hand, put Gallop on wheels and show it to customers at Computex 2027, the post says.
Gallop comes from the Humanoid Robotics business unit of the company’s Physical AI Solutions Business Group. Today it works from a fixed base with two arms, and its strength is pick-and-place work at a station, according to the post. The post gives no price.
The first station packs 15 small parts, many in thin, clear bags, and takes about 20 seconds per board. That is about 1.3 seconds per part, by our arithmetic. ASUS says it chose the station because small, thin and soft parts are exactly what a simple gripper cannot handle.
Missed boards in one eight-hour shift at the 20 seconds per board that ASUS states for Gallop’s first station, at different success rates; 99.5% is the ASUS target. Shift and miss counts are our arithmetic and assume continuous running.
How Gallop Learns
ASUS says it does not build its own large AI model and starts from open models. Experienced workers wear a small head camera, with their consent, simulation generates variations, and an operator in a VR headset and gloves guides the robot on the real parts. Once deployed, what the robot records feeds back into training.
The post describes four layers: a planner, a supervising second AI that checks every step, a skills layer and a control layer. The layers use fixed interfaces so a better model can be swapped in without rebuilding the robot. ASUS designs the body and the hand in-house; motors and standard components come from partners.
Electronics manufacturing comes first, the post says, then hospitals and service sites over time. ASUS says Gallop is not meant to replace workers and is designed for repetitive, physically demanding stations that are hardest to staff.
The Structural Read
As we read the post, ASUS is positioning its own factories as its advantage: it says most robotics companies have to rent the kind of test site it already has, and that its edge is the right data rather than more data.
As we read the design, ASUS is buying the brain and building the hands: it starts from open models and Nvidia Jetson AGX Thor, but designs the body and the dexterous hand in-house, which the post calls the hardest part of any humanoid.
ASUS Pressroom post on Gallop, 13 November 2026
“A demo shows that a robot can do something once. A factory needs a robot that does the job every day.”
Three Implications
FOR ASUS The post places Gallop in a Physical AI Solutions Business Group, alongside the motherboards, PCs and servers ASUS already builds; customer pilots are planned for 2027.
FOR NVIDIA Gallop runs on one Jetson AGX Thor computer inside the robot, according to the post, which adds a PC and server maker to the robot builders using that platform.
FOR FACTORY BUYERS ASUS sets 99.5% success within 20 seconds per board as the bar, the standard it holds a person to; the post gives no price.
The framework says: “The Container Moat — built by closed data loops inside a customer’s environment, nascent but the deepest position in the stack.”
As we read it, ASUS is describing that kind of loop inside its own plants: workers’ head-camera video, operator guidance and on-the-job recordings from a real station feed back into training. The framework predates Gallop and does not discuss ASUS.
Gallop at a glance, as described in the ASUS Pressroom post of 13 November 2026.
What Is Not Established
The post does not report a current success rate for the station, only the 99.5% target, and it gives no price, unit cost or production volume. The shift figures above are our arithmetic from the stated 20 seconds per board and assume continuous running; they are not ASUS figures.
We read the ASUS post in full. CNBC’s interview with S.Y. Hsu is referenced in the post; we did not rely on it for any figure. We did not contact ASUS for this piece.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
ASUS is entering humanoid robots with a factory job rather than a demo: a two-armed robot on Nvidia Jetson AGX Thor, a 99.5% target at 20 seconds a board, pilots planned for 2027 and availability targeted for the second half of 2028, according to its post.
A note on sourcing. Every fact about Gallop comes from the ASUS Pressroom post of 13 November 2026, which we read in full; the post refers to an interview with S.Y. Hsu on CNBC, which we did not rely on for any figure. Boards per hour and per shift, missed boards at each success rate and seconds per part are our arithmetic from the stated 20 seconds per board and 15 parts. ASUS was not contacted for this piece. This piece is not investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
