Alibaba integrates Nvidia’s robotics stack into its cloud AI offering

Alibaba Group will add Nvidia’s suite of tools for building AI that interacts with the physical world to its cloud software platform, enabling customers to develop applications ranging from humanoid robots to self-driving cars.

The company announced the move at its annual Apsara developer conference in Hangzhou, saying Alibaba Cloud’s Platform for AI will include the full Nvidia Physical AI software stack among its menu of options for developers. The integration gives users of Alibaba’s Cloud Intelligence unit access to Nvidia’s “physical” artificial intelligence (AI) development tools, the company said.

What the integration enables

Alibaba’s statement frames the change as a practical expansion of options for developers on its platform: with the Nvidia stack available, customers can prototype and build AI models tuned for perception, control and simulation tasks required by robotics and autonomous vehicles. The announcement describes use cases explicitly as “humanoid robots to self-driving cars”.

At a product level, the arrangement places Nvidia’s tooling alongside other development options in Alibaba Cloud’s Platform for AI, allowing developers to choose the stack that best fits their hardware, software and deployment constraints.

The announcement followed a company statement that Chief Executive Eddie Wu plans to raise Alibaba’s AI infrastructure investment beyond a previously announced 380 billion yuan ($53 billion). Alibaba said the commitment supported a rally in its share price, noting the move helped drive the company’s shares to their highest point in nearly four years.

Separately, Alibaba has introduced its Qwen-3 Max large language model; the company has positioned the model as designed to reduce hallucinations — “making stuff up” — and to deliver higher-quality responses to open-ended queries. The Qwen-3 Max launch is cited by Alibaba as part of a broader push to expand AI capabilities across its cloud and product portfolio.

Geopolitical and regulatory tensions noted in the company material

It is mentioned in the text that, a few years ago, Nvidia CEO Jensen Huang stated that he is disillusioned by the strict prohibition on the sale of advanced chips of the company to China. The context of that highlights that the nature of operational and supply limitations may make the practical deployment of hardware-dependent software stacks more difficult.

The text notes that “A while back, Nvidia CEO Jensen Huang said he is disappointed by the tight restrictions on selling the company's advanced chips to China.”

Risks, trade-offs and what to watch:

Hardware access and deployment Hardware accessibility and deployment The Nvidia stack can be integrated with special accelerators; any restrictions on chip sales or export restrictions can place a barrier on the feasibility of a complete development and deployment cycle in China.

Developer integration work: Making the software stack available on Alibaba Cloud reduces a software barrier, yet system integration, robotics, and autonomous platform safety and verification remain among the key barriers to developers developing physically grounded AI.

Competitive positioning: Alibaba frames the move as being part of a broader AI investment initiative; the selection customers would make between in-house models, Alibaba toolchain and third-party stacks will be determined by performance, cost and regulatory limitations.

Safety and standards of operation: Uses in robotics and vehicles demand more safety and regulatory assurance than a standard cloud AI workload, which will require safety controls and regulatory ways to be proven.

The addition of the Nvidia Physical AI software stack to the Alibaba Cloud platform increases the possibilities available to developers on the platform and plans with an identified rise in AI infrastructure expenditure. The relocation highlights an attempt to render Alibaba Cloud useful in the development of physical world AI, such as robots andautonomous vehicles, yet the feasibility of actual use will be determined by the hardware accessibility, developer adoption, and regulatory conditions in a complicated US-China technology environment.