Can Pudu's 'One Brain, Many Forms' Win the Robot Endgame? Let's Do the Math

At WAIC 2025, Pudu Robotics released its "One Brain, Multiple Forms" strategy, centered on two software platforms, PuduFM and PuduAgent, paired with hardware bodies of different forms. This world-leading commercial robotics company (by market share) began answering the question every pre-sales engineer fears most: How does the client generate continuous value after purchasing?

The era of making money selling hardware is over. Pudu's "One Brain, Multiple Forms" is essentially doing two things: lowering the usage threshold for clients and increasing the task reuse rate of robots. But the question arises—compared to another mainstream industry route, "One Form, Multiple Brains" (one general-purpose humanoid robot adapting to multiple tasks), which has higher commercial landing efficiency?

Comparison: Two Routes, Two Ways of Counting Costs

First, let's look at Pudu's route. They use PuduFM to build digital twins and task orchestration, and PuduAgent for multimodal interaction and decision-making. The underlying hardware can be delivery robots, cleaning robots, service robots, or even robotic arms. The core logic of this system is software reusing hardware—the same brain, swapping different bodies to do different jobs.

Now look at the popular "One Form, Multiple Brains" route, represented by Tesla Optimus and Figure AI. They bet on this: a general-purpose humanoid robot, by continuously updating its brain software, can do everything humans can do. Costs are concentrated in the hardware body, while software is iteratively updated.

Dimension	Pudu "One Brain, Multiple Forms"	Industry "One Form, Multiple Brains"
Hardware Cost	Low (Reuses mature chassis + modular bodies)	Extremely High (Humanoid bipedal + dexterous hands)
Software Reuse Rate	High (One brain runs multiple forms)	Low (Retraining needed for each task switch)
Client Willingness to Pay	Medium-High (Clear scenarios, calculable ROI)	Extremely Low (Currently lacks practical scenarios)
Landing Difficulty	Low (Direct replacement of manpower in existing scenarios)	Extremely High (Tech immature, costs won't come down)
Ceiling	Medium (Limited to structured scenarios)	High (Theoretically can replace all human labor)

From a sales perspective, Pudu's route can help clients calculate the ROI within a year. A hotel front-desk robot replaces one staff member. Monthly cost is 3,000 RMB, robot price is 30,000 RMB, payback in 10 months. Clients are willing to pay because this is "certain cost reduction." Whereas the general-purpose humanoid robots of the "One Form, Multiple Brains" route currently cost hundreds of thousands of RMB and cannot handle the complete workflow of any single job role—clients buy them just as display items. Who wants to pay for "display"?

What Exactly Is Pudu's "One Brain, Multiple Forms" Selling?

PuduFM and PuduAgent, if broken down, are actually two different products:

• PuduFM (Facility Manager): Targets B-end clients, selling "automated management capabilities." Clients can drag-and-drop to orchestrate tasks, e.g., "Clean lobby at 10 AM, deliver takeout to Room 301 at 11 AM." This essentially turns robots into "programmable labor resources." For clients, they aren't buying a robot, but "task execution capability without manpower."

• PuduAgent (Agent): Targets C-end users, selling "interaction experience." Hotel guests can talk to the robot to ask for directions, order food, or control room devices. This essentially turns robots into "mobile smart terminals." For clients, they are buying "tools to enhance customer satisfaction."

These two products bundled together form the core of Pudu's "One Brain." But note, Willingness to pay for PuduFM is much higher than for PuduAgent. Because the former directly saves clients money (reducing labor costs), while the latter only helps clients make money (enhancing customer experience, indirectly increasing revenue). In the current economic environment, enterprises are more willing to pay for "cost reduction" than "revenue growth."

Data Speaks: The Gold Content of Market Leadership

Pudu claims global market share leadership; this data needs to be dissected. According to third-party reports, global shipments of commercial service robots in 2024 were about 450,000 units, with Pudu shipping about 120,000 units, accounting for 26.7%. But note, Pudu's main product is delivery robots, with unit prices between 15,000-30,000 RMB, resulting in total revenue of about 2.5 billion RMB. In contrast, UBTECH's humanoid robots sell for 600,000-800,000 RMB, but shipment volume is only around 2,000 units, with revenue of about 1.5 billion RMB.

Key Data Comparison:

• Pudu: High unit count, low unit price, high client repurchase rate (hotel, dining scenarios)
• UBTECH: Low unit count, high unit price, extremely low client repurchase rate (display, research scenarios)

From a business health perspective, Pudu's cash flow is more stable. But the problem is, low unit price means low gross margin. Pudu's gross margin is roughly 35%-40%, while UBTECH's humanoid robots can achieve gross margins of 50%+. So Pudu must rely on "volume" to dilute R&D costs, and the "One Brain, Multiple Forms" strategy is precisely designed to increase multi-unit purchases per client—a hotel that previously bought only 1 delivery robot can now buy 1 delivery + 1 cleaning + 1 front desk reception, raising the average transaction value from 20,000 to 60,000 RMB.

Landing Difficulty: The Real Pain Points of Clients

As a pre-sales engineer, what I care about most is: Can clients actually use it? Pudu's "One Brain, Multiple Forms" looks beautiful, but there are several pitfalls in actual landing:

Pitfall 1: Digital twins require massive on-site data. PuduFM needs to construct a 3D model of the venue before orchestrating tasks. For large hotels and malls, modeling costs could reach 50,000-100,000 RMB, exceeding the price of the robot itself. Clients will ask: "I spend 100,000 on modeling just to save one waiter? The math doesn't add up."

Pitfall 2: PuduAgent's interaction capabilities are limited. Currently, large models still suffer from hallucinations in open-domain conversations. If a hotel guest asks "What's good to eat nearby?", the robot might give incorrect info. Once errors occur, customer satisfaction drops instead of rising. PuduAgent is currently better suited for "command-based interaction" (e.g., "Deliver a bottle of water to 801") rather than "open-ended Q&A."

Pitfall 3: Management complexity of multi-robot coordination. A venue with 5 different

Original link: https://www.qbitai.com/2026/07/453802.html