The 2026 World Artificial Intelligence Conference will be held in Shanghai in July, featuring a 100,000-square-meter exhibition area with over 1,100 companies and more than 3,000 exhibits. The conference highlights three clear pathways of China’s AI development: At the system competition level, China is no longer following but establishing its own priorities, shifting from isolated breakthroughs to integrated capabilities spanning chips, connectivity, thermal management, and software; at the industrial application level, AI is evolving from chat tools into intelligent agents capable of operating tools, understanding environments, and breaking down tasks, while humanoid robots are transitioning from performances to practical production tools; at the global governance level, 29 countries have signed to establish the World Artificial Intelligence Cooperation Organization, as China moves from building a domestic ecosystem to fostering a global cooperation framework. China’s AI is advancing from technology to products to industry on a global scale, contributing its own solutions and strength to global development.

Author and source: WeChat Official Account Huashang Taoli

China is now playing its own cards.

July 17–20, 2026, Shanghai, World Artificial Intelligence Conference.

A 100,000-square-meter exhibition area, over 1,100 companies, more than 3,000 exhibits, and over 300 global debut products. From the Kimi K3 large model to ultra-nodes from Huawei, Sugon, Alibaba, and ZTE; from humanoid robots to smart cars, from industrial agents to AI glasses... nearly all major forces in China’s AI industry are showcased here. But to view this event merely as a product launch would be to greatly underestimate its significance.

What it truly reveals is not just a company or a product, but China’s increasingly clear AI development roadmap—some might even say: China’s AI hand has been fully laid out at this conference.

01 System competition has been revealed.

The competition in AI is ultimately a competition between China and the U.S. In recent years, this rivalry has primarily centered on models and chips, with China largely following the U.S. lead: How can we get closer to GPT? How do we overcome the chip supply bottleneck?

But this year, Shanghai is different. China is playing its own cards.

Models and chips remain a priority, but China has gone further by establishing its own priorities beyond those of the United States.

After large models like DeepSeek, Kimi, Qwen, Zhipu, MiniMax, and StepAI are racing ahead. Performance continues to surpass the U.S., but the focus of competition has shifted:

Previously, we compared parameters, rankings, and how clever a single response was; now, we compare reasoning efficiency, task execution, and cost of use—whether it can truly integrate into enterprise business processes, and how to transform from a “chatting software” into an intelligent agent that can invoke tools, understand environments, break down tasks, and get work done.

Regarding chips and computing power, the focus is no longer on competing with NVIDIA, but on forging our own path.

The computing power system includes not only chips but also high-speed interconnects, storage, networking, liquid cooling, scheduling software, and system-level engineering. At this conference, Chinese companies are no longer highlighting isolated components but instead offering comprehensive solutions for super nodes and computing power systems.

Huawei's Ascend 950 super node made its first physical debut; the "compute wall," composed of cabinets with thousands of accelerators, is one of the most eye-catching exhibits in the hall. Sugon unveiled China's first fully domestic 100,000-accelerator cluster, the Sugon 8000 (Dengfeng), with everything from chips to cooling自主研发. Alibaba's Panjiu AL128 super node features 128 self-developed AI accelerator cards per cabinet, assembled into a larger "computer." H3C, ZTE, and Kunlun芯 also showcased their own solutions.

China's computing power is shifting from simply stacking servers to designing chips, interconnects, power supply, cooling, and software together.

Previously, the adaptation of DeepSeek V4 with Huawei Ascend and Cambricon followed the same logic: full autonomy and optimization from the underlying chips to the algorithmic level.

Behind this is a profound transformation: every link in China’s AI industry chain is connecting at an unprecedented pace, combining collaborative innovation to generate systemic strength and competitive advantage.

Individually, Chinese companies may not be strong in every single area, but now all these points are connected, leveraging strengths and compensating for weaknesses to create a comprehensive advantage through an integrated system.

When chips aren’t powerful enough, compensate with interconnect efficiency and cluster scale; when there are gaps at individual points, make up for them with systems engineering; when proprietary solutions are too expensive, adopt open source and low-cost inference; when overseas supply is unstable, ensure mutual compatibility among models, chips, servers, and cloud platforms.

This system isn’t perfect, but it means China’s AI has achieved maximum autonomous closure and is now building a positive feedback loop: large models are optimized for domestic chips, domestic chips enable domestic computing power, and this computing power and these algorithms are integrated into cars, phones, robots, and production lines—working together to get AI running, then making it run faster.

Once this feedback loop starts running, the flywheel will spin faster.

02 Industry acceleration, clearly revealed

The most noticeable feeling at this year’s conference is that there are many more tangible, actionable solutions.

Robots are handling material transport, assembly, and sorting; intelligent agents are operating software, processing documents, and running enterprise tasks; AI has been integrated into cars, glasses, headphones, medical devices, and industrial machinery.

Enabling AI to move from screens into the real world to get things done is both the biggest difference between the U.S. and Chinese AI approaches and the area where China has the strongest foundation to build a leading advantage—while the U.S. may be better at creating more powerful chips and smarter models, China excels at quickly integrating models into products, connecting them to factories, and embedding them into industries.

China has the world’s most comprehensive and largest manufacturing system, along with massive markets in consumer electronics, smart vehicles, e-commerce, logistics, mobile payments, drones, and industrial internet—these are not just customers for AI, but also the fertile ground for its continuous training and evolution.

In the U.S., a technology might first become a cloud service; in China, the same capability will quickly be integrated into smartphone operating systems, vehicle infotainment systems, robot bodies, and factory production lines.

Thus, we see an especially high concentration of humanoid robots at this conference, collectively transitioning from "performances" to "production tools."

In the past, robots were judged on whether they could walk or dance; today, companies are seriously asking: Can they work continuously for eight hours? Can they handle loading, assembly, and quality inspection? Can they operate in workshops, warehouses, or hazardous environments? And can their costs be reduced to an affordable level?

In these specific industry application areas, China is also waging a systemic battle.

Companies like Unitree Robotics and AgiBot are gaining attention not only because their movements are more agile, but also because China has a complete supply chain to support the rapid industrialization of robots—encompassing motors, gearboxes, sensors, batteries, controllers, and structural components, along with large-scale manufacturing capabilities that enable rapid iteration and cost reduction.

AI provides the "brain," China's manufacturing provides the "body"—only when combined do they form a true artificial intelligence industry.

The same applies to agents.

Previously, large models required users to constantly ask follow-up questions; now, intelligent agents can autonomously understand goals, break down tasks, invoke software, organize data, and generate solutions, continuously answering with facts. Are they reliable? Controllable? Worth it?

At this conference, an increasing number of companies are no longer focusing on how powerful their models are, but rather on token costs, invocation auditing, data security, and business闭环.

H3C's Turing pilot platform supports over 90 large models; China Telecom showcases a token-secure router that enables enterprises to see who is accessing models, how much they are using, and how much they are spending; Kingsoft Office integrates AI into enterprise data and workflows to meticulously manage costs.

These products may not sound flashy, but they signal that AI is truly becoming a business—companies won’t pay for imagination; they’ll pay for efficiency, cost savings, and results.

Over the past one to two years, the United States has primarily focused on raising the upper limits of AI capabilities, while China has worked to lower the barriers to AI adoption. Ultimately, industrial progress depends on the breadth and accessibility of its applications.

Chinese companies may not always be the first to propose a concept, but they are often the fastest to turn that concept into a product, the product into an industry, and then scale up the industry while reducing costs—ultimately enabling widespread adoption and accessibility.

China has already gone through this with photovoltaics, power batteries, new energy vehicles, and drones—AI is likely to follow the same path.

03 Global cooperation, revealed

The development of artificial intelligence should not be a solo performance by one country, but a symphony of global cooperation.

The most significant event at this conference, more impactful than any product, was the signing on July 16 of the Agreement on the Establishment of the World Artificial Intelligence Cooperation Organization by 29 countries in Shanghai.

This intergovernmental international organization, headquartered in Shanghai, will collaborate on AI technologies, industries, talent, standards, security, and governance to help more countries enhance their AI capabilities.

This means that Chinese AI is transitioning from building its own domestic ecosystem to establishing a global collaborative network, moving from catching up to actively participating in global development.

For many years, the rules and platforms of global technology have been dominated—or所谓 “tech hegemony”—by the United States and its corporations. In the AI era, the U.S. seeks to continue this path: tightening its grip on the most advanced chips, models, and cloud platforms through companies like NVIDIA, Microsoft, Google, Amazon, and OpenAI, while using export controls, investment restrictions, and supply chain alliances to convert technological advantages into industrial dominance and build a closed circle of influence.

At the end of June this year, the United States hosted the second "Silicon Peace" summit, where 35 economies signed the Joint Statement on the Partnership for AI Opportunities; at the same time, 10 new members, including Argentina and Chile, joined, increasing the total number of members in the Silicon Peace initiative to 24.

This circle excludes China and other developing countries.

On July 6, ahead of the Shanghai Artificial Intelligence Conference, the United Nations held its first intergovernmental global AI governance dialogue in Geneva, with representatives from both China and the United States in attendance, though their governance approaches showed clear differences:

China advocates for the United Nations to serve as the primary channel for global AI governance and to bridge the "intelligence divide"; the United States favors establishing a regulatory framework led by itself, with standards set by small cliques such as the G7.

Under these circumstances, the establishment of the World Organization for Artificial Intelligence Cooperation carries particular significance. China’s approach—open source, inclusive, capacity-building, and industry collaboration—is especially welcomed by many developing countries.

For many developing countries, and even numerous developed nations, what they truly lack is not a powerful model, but rather computing infrastructure, AI talent, local data, industry-specific solutions, and opportunities to participate in shaping international rules—in short, accessible, affordable, and customizable tools.

This is precisely what China can offer: communication networks, data centers, cloud platforms, smart terminals, and industrial equipment—first helping to build digital infrastructure, then connecting AI to agriculture, healthcare, education, and more.

China’s AI development encompasses more than just chips, computing power, and large models. As Chinese AI expands globally, what is being sold is not merely a model or a chip, but an entire suite of industrial capabilities—from infrastructure to models, from equipment to applications, and from talent training to operations and maintenance. These are areas where China has already accumulated mature experience in communications, power, and transportation, making them easy to implement and deliver tangible results.

In some respects, global AI will also be led by the U.S. and China: the U.S. may have an advantage in exporting GPUs, models, cloud services, and software, while China excels in exporting industry, engineering, applications, ecosystems, and collaboration.

Neither path is inherently superior; they represent two distinct logics of AI globalization: one focuses on maintaining technological superiority, while the other expands industry reach; one relies on export controls and supply chain alliances to solidify advantages, while the other grows its network through open source, applications, and international organizations.

The security, risks, and governance of AI cannot and will not be solved by any single country alone, but there will also not be many viable approaches or pathways—ultimately, it will largely be dominated by cooperation and competition between China and the United States.

In the past, how Chinese AI participated in global development was relatively unanswered and lacked action. The establishment of the World Organization for AI Cooperation is both a clear answer and a firm action.

From then on, China began building platforms, connecting partners, and participating in shaping global development rules for the AI era—Chinese AI also began to expand globally, from technology to products to industries.

And this is precisely what makes this conference truly stand out beyond just an industry exhibition.

The 2026 World Artificial Intelligence Conference demonstrates to the world through facts and actions that global AI competition is evolving from model-versus-model and company-versus-company rivalry to system-versus-system competition. China will offer its own solutions and contribute its strength through its independent system for global development.

Models, chips, computing power, agents, robots, terminals, global governance and cooperation... At this conference, China’s AI was presented comprehensively on the table.

How can Chinese AI develop itself, build advantages, and participate in and contribute to globalization—

All the cards are on the table.