SHANGHAI, July 22 (Xinhua Silk Road) -- After years of dazzling demonstrations and headline-grabbing breakthroughs, the focus of artificial intelligence (AI) is shifting decisively toward commercial deployment and measurable productivity. This transition, investors believe, is fundamentally reshaping the AI investment landscape.
During the 2026 World AI Conference (WAIC) and High-Level Meeting on Global AI Governance held in Shanghai from July 17 to 20, fund managers from some of China's largest asset management firms spent days touring exhibition halls, engaging with companies and testing products firsthand alongside Xinhua Finance. Their conclusion was strikingly consistent: AI has moved beyond proving what it can do to demonstrating what it can deliver.
This paradigm shift is creating a new investment playbook -- one centered less on technological promise and more on commercial execution. Domestic computing infrastructure, embodied AI, and enterprise AI agents have emerged as the sectors with the strongest long-term growth prospects and increasingly visible earnings potential.
-- Robots move from "showtime" to workplace
For Zhang Lu, a fund manager at Maxwealth Fund Management, WAIC has become a barometer of China's AI progress. Having attended the conference for four consecutive years, he has watched the evolution of robotics unfold in real time.
"In the first year, robots sat behind glass as static exhibits," Zhang recalled. "The following year, they could dance and perform eye-catching demonstrations. Last year, companies focused on application scenarios, such as factory assembly lines and household tasks like folding clothes."
This year, Zhang's biggest takeaway was that robots have truly integrated into the exhibition itself, acting as "busy staff members".
"Across the exhibition halls, many humanoid robots welcomed visitors, provided directions and assisted with public services," he said. "I personally experienced robot-assisted water purchasing and intelligent pharmaceutical sorting. The application scenarios are much richer and more refined."
His observations echoed a broader consensus among investors: humanoid robots are no longer being judged by how impressive their demonstrations are, but by how effectively they perform real-world work.
-- Productivity emerges as new industry benchmark
Ma Lei, a fund manager at China Universal Asset Management, identified three defining shifts that stood out at this year's conference.
First, the industry has moved beyond "showcasing large language model capabilities" toward "demonstrating complete AI-powered productivity". Previous conferences revolved around model parameters, benchmark rankings and content generation quality. This year, exhibitors focused instead on how AI agents interact with software, access enterprise data and complete real-world tasks.
Second, innovation is increasingly driven by the integration of hardware and software rather than isolated technological breakthroughs. Domestic computing exhibits have evolved from showcasing individual AI chips to presenting complete computing ecosystems, including chips, servers, supernodes, high-speed interconnects, compilers and model adapters.
Third, the focus has pivoted from technical proof-of-concept to large-scale commercial deployment. Companies are increasingly focused on answering three practical business questions: "Will customers pay?" "Is the product sustainable?" and "Can AI generate a measurable return on investment?"
"The conversation has shifted from large models to AI agents, world models, AI coding, token economics and open agent ecosystems," Ma said. "Competition is no longer about whether a model can answer questions. It's about whether it can reliably complete work."
-- Domestic foundation models come of age
Xiao Wanyuan, a fund manager in the index research department at E Fund Management, pointed to another notable change: the rapid maturation of China's domestic large language models.
"In previous years, relatively few domestic foundation models appeared at WAIC," Xiao said. "Most companies showcased cloud services or computing platforms, and only a handful could realistically compare themselves with leading international models."
"This year tells a very different story," he added. "A large number of mature domestic models have emerged. When cost efficiency is taken into account, many have already entered the global top tier."
Just ahead of WAIC's opening, Beijing-based startup Moonshot AI released Kimi K3, a new open-source model with 2.8 trillion parameters, drawing broad attention and highlighting China's rising AI capabilities. Independent evaluator Artificial Analysis gave K3 an Intelligence Index score of 57, placing it third globally, only behind Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol.
For Xiao, the value of attending WAIC lies less in collecting information than in experiencing industry progress firsthand.
"Walking on the exhibition floor allows you to see how deeply AI is becoming embedded in real business workflows," he said. "It also provides a much clearer understanding of how computing infrastructure is evolving to support this transformation."
-- Domestic computing infrastructure tops investor watchlists
Based on discussions at WAIC and their assessment of industry fundamentals, fund managers broadly agreed that AI investment opportunities are expanding across the entire value chain. Among the highest-conviction themes, domestic AI computing infrastructure stands out as the most mature.
"This remains the segment with the strongest industrial momentum and the greatest visibility on future orders," Ma said.
As AI models evolve toward multimodal reasoning and autonomous agents, demand for computing is increasingly driven by inference rather than training. More AI interactions mean more tokens generated, longer reasoning chains and greater demand for high-performance inference infrastructure.
At the same time, China's push for greater technological self-reliance is accelerating investment across the domestic computing ecosystem, from AI chips and servers to supernodes and data-center systems.
Zhou Jingxiang, a fund manager at Lion Fund Management, noted that the sector is entering a phase where financial performance is beginning to catch up with technological progress.
"Order visibility has improved significantly across the supply chain," Zhou said, citing liquid cooling systems, high-speed interconnects and AI servers as key beneficiaries. "Capital is increasingly rotating away from pure AI concepts toward companies with verified orders and commercial deliveries."
Enterprise AI agents represent another area attracting growing investor interest. Unlike consumer-facing AI applications, enterprise AI solutions deliver more tangible economic value through lower development costs, higher productivity and measurable efficiency gains.
Companies deploying AI coding assistants or intelligent customer-service agents can directly quantify improvements in development speed and operating costs, making returns on investment easier to validate.
-- Market focus shifts to commercial execution
Perhaps the biggest takeaway from this year's WAIC was not the technologies themselves, but the changing mindset of investors. AI investing in China's secondary market is increasingly moving away from speculative themes and lofty expectations toward tangible business fundamentals, including customer orders, revenue growth and cash-flow generation.
The robotics sector serves as a prime illustration. While investors remain bullish on the long-term potential of embodied AI, they believe the next phase of value creation will hinge on manufacturers' ability to achieve large-scale commercial production, rather than merely demonstrating technical feasibility.
Zhang Xiangyu, a researcher at Galaxy Asset Management, expected upstream component suppliers, including screw drives, joint modules and sensors, to be among the earliest beneficiaries once production volumes accelerated.
"China already has strong manufacturing capabilities across many of these components," Zhang said. "If 2026 proves to be the year of large-scale production validation, upstream suppliers are likely to be the first to see meaningful order growth."
Looking further ahead, Zhou believed that China's AI ecosystem is approaching a critical inflection point. Domestic computing infrastructure is increasingly optimized for domestic large models, creating a self-reinforcing AI ecosystem spanning hardware, software and applications.
As supernode systems enter commercial deployment and large-scale computing clusters continue to expand, improvements in computing stability, energy efficiency and resource scheduling are providing an increasingly solid foundation for next-generation AI models.
"In the medium to long term," Zhou said, "continued breakthroughs in computing infrastructure and foundation models will unlock productivity gains across virtually every downstream industry."
For investors, this marks a profound transition. The era of investing in AI's potential is giving way to one of investing in its execution and, increasingly, its earnings. (Contributed by Wei Yutian and Gao Pan)
A single purchase
