Newshot

Smart computing takes the baton: physical AI's next opportunity lies in computing infrastructure

AI summary

The momentum behind physical AI is spreading from robot hardware to the rest of the industrial chain. As embodied AI R&D moves through data collection, simulation, model training, reinforcement learning, inference, and real-machine validation, computing demand shifts by stage, making on-demand, elastically scalable cloud computing an increasingly realistic choice for robotics companies.

Translation(model translation, not the source's own text)

Title: Intelligent Computing Takes the Baton: The Next Wave of Physical AI Lies in Computing Infrastructure

The excitement around physical AI is spreading from robot bodies to the upstream and downstream of the industry chain.

Over the past year, humanoid robots and quadruped robot dogs have appeared frequently in factories, warehouses, and stores. The questions the industry cares about are changing: from "what actions can a robot perform" to "how can it get more accurate over time."

The answer lies not only in the robot body but also in the computing system behind it: where data comes from, where simulation runs, how often models iterate, and how computing power matches the R&D cadence. None of these can be solved by simply buying more servers.

英伟达 proposed the concept of "three computers": one for training models, one for training and validating skills in a digital twin and simulation environment, and a third deployed on the robot body for execution. Platforms such as Omniverse, Isaac, and Cosmos are designed to connect training, simulation, and edge computing into one chain. Its latest quarterly financial report shows that its data center business revenue grew 117% year over year, and physical AI is becoming a new growth market for computing platforms.

But simulation cannot fully replace the real world. Neura Robotics founder and CEO David Reger said in an interview with McKinsey that there is still a gap between digital twins and the real environment, and robots need to be trained in real physical environments and collect multimodal data such as motion, touch, and sound. These feedbacks then feed into the next round of training and simulation.

As a result, physical AI R&D has formed a continuously iterating closed loop: data collection, simulation, model training, reinforcement learning, inference, and real-machine validation. Each stage has different computing power requirements, and they change dynamically with R&D progress, making fixed-configuration computing power hard to adapt.

Changes in the R&D model are reshaping how computing power is supplied and serviced. Compared with one-time construction and long-term ownership, cloud computing power that can be called on demand and elastically scaled has begun to become a new choice for robotics companies.

Chinese robotics companies feel this more directly.

Unitree Technology founder 王兴兴 said at the Industrial Fair in October this year that robot bodies are limited by space, battery, and heat dissipation, so the computing power that can be deployed is limited, and computing tasks suitable for the cloud should be moved to the cloud as much as possible. Cloud computing power can be allocated in real time and is more flexible to manage. He also emphasized that real-machine data is an indispensable closed loop for model iteration. The smarter the robot, the more it depends on the cloud computing power behind it.

AgiBot co-founder 彭志辉 revealed in an interview that the company's training cluster this year will approach the scale of ten thousand GPUs to cope with growing computing demand. Co-founder 姚卯青 added that computing power is mainly obtained through leasing to control asset investment and retain scaling flexibility.

The computing power ledger of leading robotics companies is shifting from "how many GPUs to buy" to "whether usable computing power can be called up at any time."

The reason is that embodied intelligence R&D has obvious phases: model training, simulation, and reinforcement learning push up computing demand, while demand may fall back during the validation stage. Self-owned hardware has high costs and is prone to idling, while temporary procurement makes it difficult to ensure R&D continuity.

Therefore, intelligent computing cloud services that offer on-demand invocation, elastic scaling, unified scheduling, and adaptation to the multi-scenario needs of physical AI are becoming a more realistic choice.

Intelligent Computing Takes the Baton: The Next Wave of Physical AI Lies in Computing Infrastructure
Intelligent Computing Takes the Baton: The Next Wave of Physical AI Lies in Computing Infrastructure

(2026 ranking of the fastest-growing AI neocloud vendors for physical AI)

The capital market has already priced this demand. According to industry estimates, CoreWeave's revenue in the first half of 2026 grew 112% year over year, and its full-year capacity has been sold out in advance; domestic AI infrastructure and intelligent computing cloud service provider Jiuzhang Cloud, in the first half of 2026, saw its cloud business revenue grow about 420% year over year. The growth of AI neoclouds reflects that the market is continuously increasing its willingness to pay for "always available" computing resources.

However, there is another way to calculate this account: the unit price of on-demand rental may be higher than self-building, and under long-term high-load operation, the total cost may not necessarily be lower; there is also no industry consensus on whether the computing demand of physical AI can grow as rapidly as that of large models. At present, the training scale of robotics companies is still significantly smaller than that of general large-model companies. But the demand for multi-step computing has become clear, and the growth pace still needs to be verified.

Shang Mingdong's view points to another dimension: whether computing resources are actually used. Jiuzhang Cloud co-founder and COO Shang Mingdong previously said in an interview with CCTV Finance that measuring the computing power industry cannot rely only on construction scale, but more importantly on whether resources are truly used and whether they can serve more industrial entities. Computing power is transforming from a scarce technical resource into a general basic means of production.

Jiuzhang Cloud's approach is to undertake distributed computing demand in layers rather than relying on a single centralized solution. At the bottom layer, through heterogeneous chip adaptation, computing power pooling, technical solutions to prevent computing power fragmentation, and unified scheduling, it enables computing resources of different architectures to operate collaboratively and provides a centralized large-scale computing foundation for robot foundation model training tasks; the middle layer relies on cloud-native cluster management, high-performance storage, and reinforcement learning runtime environments to support automated orchestration of multi-round training and validation tasks; the upper layer connects model training and inference consistency capabilities and extends to the edge inference layer, dynamically allocating resources according to task requirements. What is delivered is no longer isolated computing resources but computing services that can be scheduled, scaled, and delivered through distributed collaboration.

Intelligent Computing Takes the Baton: The Next Wave of Physical AI Lies in Computing Infrastructure
Intelligent Computing Takes the Baton: The Next Wave of Physical AI Lies in Computing Infrastructure

(Global AI neocloud vendors are actively positioning themselves in physical AI)

This is also where Jiuzhang Cloud differentiates itself. Domestic robotics companies face a more complex computing environment: chips from different vendors differ in architecture, drivers, and software stacks, and training, simulation, and inference also have different resource requirements. Cross-architecture adaptation and unified scheduling capabilities directly affect resource utilization and overall computing costs. This work is not conspicuous, but it is the key to truly forming competitiveness in intelligent computing cloud services.

The way transactions are conducted is also changing. In the past, computing power procurement mainly revolved around GPU type, quantity, and unit price; now, companies care more about whether they can invoke by task, meter by actual usage, scale up during R&D peaks, and release resources during the validation stage. Computing power is shifting from equipment procurement to service delivery.

This capability has already entered the physical AI R&D process. In joint validation with a leading world model company, Jiuzhang Intelligent Computing Cloud supported the parallel scheduling of thousands of virtual robot policy samples in an ecosystem partner's simulation environment, accelerating reinforcement learning and safety testing; in joint validation in an automotive assembly scenario, elastic multi-stage computing scheduling supported the improvement of policy iteration efficiency. Facing computing demand that changes with simulation scale, model complexity, and iteration rounds, elastic scheduling and multi-stage computing invocation help reduce idle resources and improve R&D efficiency.

Third-party research also confirms this trend. Citi Research pointed out in "Embodied Intelligence: The Rise of Physical AI" that digital twins, real-world data collection, simulation, and the computing software ecosystem for coordinating hardware are all important pillars of physical AI development. Computing infrastructure has therefore become an important growth opportunity in the industry chain.

Taken together, the direction is clear: 英伟达 connects training, simulation, and edge computing, CoreWeave integrates cloud platforms into robot R&D, and Jiuzhang Cloud uses heterogeneous computing scheduling and application engineering systems to integrate computing resources and provide elastic computing services. The paths differ, but the goal is the same: to make computing infrastructure adapt to the continuously changing R&D needs of physical AI.

The heat of the physical AI industry is extending from robot bodies to the computing systems that support their continuous evolution. As R&D tasks become increasingly complex, computing infrastructure must not only provide computing power but also efficiently organize and schedule it, turning it into sustained R&D capability. The robot body determines the capability boundary of physical AI, while computing infrastructure affects its iteration efficiency. This may be the next wave worth watching more closely in the physical AI boom, and Jiuzhang Cloud has already entered early.

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by 雷锋网
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗