Shanghai's Mega AI Summit Focuses on Hardware and Global Governance
The 2026 World Artificial Intelligence Conference (WAIC) in Shanghai has emerged as a showcase not only of cutting-edge algorithms and chips, but also of China’s drive to reshape global AI governance amid intensifying competition with the United States.[3][9][1] From record-breaking exhibition halls to hard-tech breakthroughs unveiled under sweltering summer temperatures, the event highlights how AI hardware, computing power and international rule‑making are becoming central to China’s technological strategy.[3][6][1]
Record-Breaking Scale and a Focus on Computing Power
Held from July 17 to 20, the ninth edition of WAIC spans more than 100,000 square meters of exhibition space across four halls and three venues in Shanghai, the largest footprint in the conference’s history.[3][9] Organizers say more than 1,100 enterprises are exhibiting over 3,000 technologies, with more than 300 AI products making their global debut, underscoring the event’s role as a launchpad for new hardware and software.[3][4][9]
The conference is structured around two core tracks: intelligent computing, which focuses on AI infrastructure and chips, and embodied intelligence, which centers on robots and AI-enabled devices.[3][4][10] Each track has attracted more than 200 companies, marking a shift from pure algorithmic innovation toward large-scale, reliable computing platforms and real-world deployments.[3][4][2]
Domestic Hardware Under the Spotlight
A central theme at WAIC 2026 is the race to build powerful, homegrown computing infrastructure in the face of U.S. export controls on advanced chips.[2][6][10] Industry analysts at the conference describe Chinese domestic chips as crossing an inflection point from merely “usable” to “highly usable,” reflecting rapid progress in performance, scheduling efficiency and ecosystem maturity.[2]
Among the most prominent hardware exhibits is Huawei’s Atlas 950 Super Node, presented as one of the industry’s largest AI computing platforms and a symbol of China’s efforts to reduce reliance on foreign components.[6][7] The system is designed to support massive training workloads for large models, highlighting how computing clusters, rather than individual chips, are becoming the benchmark of national AI capacity.[6]
Other debuts include near‑memory computing 3D chips and new accelerator designs aimed at improving energy efficiency and throughput for AI tasks.[6] These innovations speak to a broader “heat” problem that looms over the sector: as models grow larger and data centers denser, thermal management and power consumption become critical constraints on AI scaling, especially during China’s hot summer months.[2][6] Exhibitors at WAIC showcase cooling solutions and architecture-level optimizations as part of this push to manage the physical limits of AI hardware.[2][6]
Gigantic Models, Agents and the Hardware Behind Them
On the software side, Chinese firms are using WAIC to demonstrate frontier models that depend heavily on domestic computing power. MiniMax’s M3 multimodal large model, revealed as an open-source foundation model, reportedly carries around 2.8 trillion parameters and supports a context window of up to 1,000,000 tokens across text, images, audio and video.[6][7] Such specifications rank it among the largest open-source models in China and underscore the compute‑intensive nature of current AI research.[6]
The conference also spotlights the rapid evolution of AI agents—systems capable of executing tasks and transactions rather than simply conversing. A new “Agent Operating System” called Jieyue is billed as the industry’s first full agent OS, aimed at moving agents from proof‑of‑concept demos to everyday productivity tools.[6][2] In tandem, exhibitors introduce what they describe as the world’s first AI agent smartphone, integrating agent capabilities at the device level and hinting at a future where terminals are rebuilt around AI-native operating systems.[6][2][7]
These software advances reinforce the strategic importance of reliable computing infrastructure. Conference commentary emphasizes that competition is shifting from raw scale—such as total FLOPS or number of GPUs—to end‑to‑end performance, scheduling efficiency and integration with industrial workflows.[2][8] WAIC sessions repeatedly highlight how improvements in hardware utilization and orchestration can rival algorithmic breakthroughs in impact.[2]
Embodied Intelligence: Robots Leave the Lab
The second major track, embodied intelligence, showcases humanoid robots, dexterous hands and other physical systems designed to take AI out of the data center and into factories, homes and public spaces.[3][4][6] Exhibits include multiple humanoid platforms and robotic manipulators intended for logistics, manufacturing and service tasks, reflecting China’s ambition to integrate AI deeply into its industrial base.[6][7][8]
Conference organizers say WAIC has already facilitated 57 major application scenarios, with intended cooperation agreements totaling 16.2 billion yuan (around US$2.2 billion), illustrating the push to convert research prototypes into commercial deployments.[4][6] Those scenarios span sectors from smart manufacturing and autonomous logistics to urban management and healthcare, where embodied intelligence is expected to complement, rather than replace, human workers.[4][8]
Xi Jinping’s Bid to Shape Global AI Governance
Beyond the technical showcases, WAIC 2026 is being used by Beijing as a platform to advance a new global AI governance architecture.[1][4][7] In his address to the opening ceremony, President Xi Jinping urged countries to seize what he called the “historic opportunity” of open‑source AI, warning that unequal access to the technology could create “new historical injustices.”[1] He framed China as a champion of greater AI access for developing nations and argued that rules for AI should not be dominated by a single country.[1][7]
Xi’s speech coincided with the formal launch of the World Organization for Cooperation in Artificial Intelligence (WAICO), an intergovernmental body headquartered in Shanghai.[4][5][6] According to official accounts, 29 founding countries—including Brazil, Russia, Pakistan and Indonesia—plus 10 African and 12 Asian nations signed on to establish WAICO as a mechanism intended to ensure “fair and impartial” global AI governance.[4][6]
China also pledged 5,000 AI training slots for developing countries through WAICO and related initiatives, positioning the new organization as a vehicle for capacity building in the Global South.[6][7] Chinese media and officials describe the move as a “fresh signal” of openness and cooperation, even as it implicitly challenges existing Western-led forums and standard‑setting bodies in the AI domain.[7][1]
Geopolitics, High Stakes and Competing Visions
The geopolitics surrounding WAIC are unmistakable. By coupling massive hardware showcases with a new multilateral governance body, Beijing is signaling that control over AI infrastructure and standards are strategic priorities comparable to those in telecommunications and semiconductor policy.[1][4][6] Xi’s emphasis on open‑source models and accessible computing power is likely to be read in Washington and other capitals as both a genuine bid for broader participation and a challenge to U.S. influence in setting norms.[1]
At the same time, the conference’s commercial figures underscore the economic stakes. Intended cooperation deals totaling 16.2 billion yuan and dozens of large‑scale application projects suggest that WAIC is evolving from a showcase event into a major marketplace and policy stage for China’s AI ambitions.[4][6] For international investors, partners and competitors, the convergence of heat‑intensive data centers, rapidly advancing domestic hardware and newly proposed global institutions makes Shanghai’s World AI Conference a focal point for watching how the next phase of the AI race will be contested.[1][2][6]
