China pushes toward task-oriented AI Firms leveraging industrial heft to develop dynamic applications, system integration 2026-11-16    LI JING

Visitors watch and take photos of bionic robots during the 2026 World Internet Conference Wuzhen Summit in Wuzhen, Zhejiang province, on Nov 2.

Visitors are at the exhibition area of Alibaba's Qwen during the 2026 WIC Wuzhen Summit on Nov 3.

A new generation of intelligent bionic hand from BrainCo is showcased during the 2026 WIC Wuzhen Summit on Nov 2.

China is accelerating efforts to turn advances in artificial intelligence into a vital driver of economic growth as technology companies and manufacturers use AI models to develop new products and improve productivity.

Discussions at the 2026 World Internet Conference Wuzhen Summit highlighted three areas shaping this transition: the growing role of open-source models in expanding access to AI, the integration of AI into industrial operations, and the development of robots capable of performing physical tasks. Together, they not only point to opportunities for Chinese companies in software, manufacturing and services, but also underline the challenges of controlling costs, ensuring reliability and making the technology commercially viable.

Held from Nov 2 to 5 in Wuzhen, Zhejiang province, the WIC Wuzhen Summit brought together technology executives, researchers and policymakers to discuss the development and governance of AI.

Zhang Yaqin, founding dean of Tsinghua University's Institute for AI Industry Research, said AI is moving toward a new stage in which intelligent agents could perform tasks across industries, scientific research and biology. He also envisaged an internet in which AI agents interact with one another.

China has a substantial industrial base to support this transition. The country's core AI industry was valued at more than 1.2 trillion yuan ($180 billion) in 2025, while the number of AI companies exceeded 6,600 as of June 2026, according to the China Academy of Information and Communications Technology, a research institute under the Ministry of Industry and Information Technology.

Zhu Keli, founding director of the China Institute of New Economy, said competition in AI is expanding beyond model development to include industry-specific applications, data and business ecosystems.

Companies are increasingly focused on computing efficiency and sustainable revenue, making access to customers, industry expertise and the ability to integrate AI into existing operations critical to commercial success.

International cooperation is one way to bring these capabilities together. In a keynote speech at the opening ceremony of the summit on Nov 3, Premier Li Qiang highlighted the nation's commitment to working with other countries to build an open, inclusive and dynamic digital innovation ecosystem.

Li said China will continue to embrace open innovation and promote technology transfer and knowledge sharing. He underlined the need to adhere to inclusive and balanced development of the internet to allow digital dividends to be shared.

DeepSeek research engineer Gao Xi also highlighted the practical value of openness, saying that sharing improvements in model design and training methods allows other developers to test and build on them more quickly.

Alibaba Group Chairman Joe Tsai offered an example of how open-source models can support cross-border development. A nonprofit organization in Uganda had used Alibaba's Qwen models to develop AI models covering 68 African languages, he told the summit. The initiative illustrates how developers can build on existing foundation models while contributing local language expertise and adapting the technology to their own markets.

Tsai said open-source models can lower research and development costs for small and medium-sized enterprises and help developers in emerging economies build AI applications for local needs.

Zhu said no single country can easily provide the full range of data, computing resources and industry expertise required for AI development. However, making a model publicly available does not automatically remove the barriers to international cooperation.

"Open source can open up access to code, but it is much harder to overcome barriers involving data, compliance and trust," he said.

For businesses, the next step is turning access to AI technology into practical applications that customers are willing to pay for.

Siemens' Eigen Engineering Agent, showcased at Wuzhen, is one example. The industrial AI tool can generate programmable logic controller software, configure equipment and carry out other engineering tasks, according to the German industrial technology company.

In an announcement on June 17, Siemens said the agent was being used by more than 100 companies in 19 countries. It reported that the tool could complete certain engineering tasks two to five times faster than manual workflows. The new capabilities announced at the time were included in its standard subscription at no additional cost.

The example shows how industrial AI can be integrated into existing engineering software and business processes rather than offered as a standalone technology. For manufacturers, the potential benefits include shorter development cycles and less time spent on repetitive engineering work.

Service providers are pursuing similar opportunities. DAS-Security, a leading cybersecurity company in Hangzhou, Zhejiang province, showcased its AI-powered digital security employees during the summit, with functions including vulnerability testing and code review.

The company's chairman Fan Yuan said AI can make specialist cybersecurity services available around the clock, allowing providers to extend their expertise to more tasks and customers. This can also create opportunities for service companies to move beyond traditional project-based work and offer software-based services.

China's manufacturing base offers another potential advantage, particularly in the development of physical AI.

"To develop physical AI, companies need manufacturing lines, sensors and the data generated by those production lines and sensors to train their models," said David Siroky, Cisco's AI architecture lead for Asia-Pacific, Japan and Greater China. "China has spent more than two decades building that infrastructure."

The economic logic is straightforward on the factory floor, he noted, pointing to automotive assembly. A paintwork blemish caught and buffed out on a domestic assembly line costs a fraction of what it would take to rectify once the vehicle has been shipped overseas to an export market such as Australia.

However, moving from identifying problems to performing physical tasks brings additional engineering challenges.

Unitree Robotics founder Wang Xingxing told a summit forum that robots can achieve success rates approaching 100 percent on certain fixed tasks after sufficient training, although their overall efficiency still needs to improve.

More complex operations, such as assembling mobile phones and threading needles, remain challenging. Wang said 2027 could be a turning point for the robotics industry.

Zhang, however, cautioned that developing general-purpose robots would take considerably longer time.

Wang also pointed to opportunities in overseas markets, where Chinese robotics companies can supply hardware platforms that international customers adapt for local applications.

He said overseas customers had accounted for a substantial share of Unitree's business in recent years, with international partners adapting its robots for applications including construction and commercial services.

For businesses considering wider AI adoption, however, the investment calculation extends beyond the initial cost of acquiring a model or machine.

Siroky said companies moving AI projects beyond the pilot stage sometimes found that their existing networks or electricity supplies were insufficient to support the additional computing demand. Smaller models running on a company's own equipment could be more suitable for certain applications and help avoid recurring cloud-computing charges for every request.

"You don't always need the largest, most expensive system," Siroky said. "Sometimes you only require the model that can do the job reliably, without compromising on the value or business outcome, at a lower cost."

Reliable deployment also requires cooperation on standards, testing and safeguards. John Higgins, chair of the International AI Governance Association, said applications of AI in healthcare and transportation require trust supported by evidence. He called for greater international cooperation in testing, evaluation and research.

At a Wuzhen forum, standards organizations and companies also discussed agent identities, communication protocols and safety requirements, with the aim of enabling AI agents developed by different providers to work together.

These requirements are creating business opportunities in cybersecurity and AI governance as well. Fan said risks including malicious instructions, data leakage and AI agents exceeding their authorized permissions are generating demand for model evaluation and protection services.

"The more open source thrives, the more important security becomes," he said.

The economic benefits of AI will depend not just on the growth of model developers, computing infrastructure providers and specialist service companies, but also on whether businesses using the technology achieve lasting improvements in productivity and develop products and services that generate demand.

Zhu said the key indicators to watch include lasting improvements in productivity, the development of new products, stronger domestic suppliers and job creation.

"If only a few leading companies benefit, AI's impact on overall domestic demand and employment will be greatly reduced," he said.