Shanghai AI Events Spotlight Embodied Robots and Data-Center Power Constraints

The developer conference at Zhangjiang Science Hall drew nearly 60,000 offline attendees, according to QbitAI, which republished organizer-provided material. It brought together more than 300 embodied-intelligence hardware companies, more than 200 upstream and downstream supply-chain companies, nearly 100 open-source organizations and developer communities, and more than 500 young developers and industry guests. The event generated more than 6,500 published reports and more than 6 billion online exposures, the report said.

Organizers released a “4+X” set of outcomes. They included ten real-world tracks for the skills competition, an embodied-intelligence evaluation system and tools, a Zhangjiang embodied-intelligence supply-chain platform promising one-day matching, three-day prototyping and one-week trial production, and a landmark building for Zhangjiang Robot Valley, according to the report.

The skills competition covered industrial, service and special-operation scenarios, with nine tracks and 23 events. More than 800 teams and over 1,000 robots competed, with more than 70 experts serving as judges and more than 500 industry experts joining exchanges. Awards included 16 first prizes, 21 second prizes and 31 third prizes. The closing also announced Forbes China robot awards covering humanoid robot future, multimodal interaction design, commercial value, human-machine symbiosis creativity and digital AI character image.

Guests from more than 20 countries attended, including former Google developer relations president Jeanine Banks, OpenJS Foundation executive director Robin Bender Ginn and Stanford professor Oussama Khatib, chair of the International Foundation of Robotics Research. A supply-chain exhibition gathered more than 200 companies, and a 2,000-square-meter permanent training space is intended to bring together users, robot makers, model and algorithm companies, component suppliers and research institutes. The report tied the conference to a central policy document on new quality productive forces that names embodied intelligence and humanoid robots as priorities.

At CPE 2026, Damo Technology displayed an AI energy model and agent cluster, a compute-energy coordination 2.0 platform, a forecast-dispatch-trading engine and an integrated operations platform. CEO Jian Yumin said in a keynote that green-power direct connection is stuck because generators and compute users bargain over net electricity value; hundreds of projects have been approved but few have begun construction, according to the report.

Damo proposes optimizing across energy, regulation, capacity and carbon markets. In a western-province simulation with real data, it tested 12,000 scenarios and 20 open boundaries, equivalent to 2 billion static calculations, and found more than 0.03 yuan per kWh of combined value. That filled a 0.02 yuan gap and left room, raising investor IRR by about 20% and lowering the user’s delivered electricity price by 10%, the report said.

Damo also proposed E-AIDC, or energy-driven AI data center, arguing the next token factory is an energy project before it is a data-center project. Its framework covers lifecycle energy operations, low-carbon efficiency through power-computing and source-grid-load-storage coordination, and green-power direct connection. The report cited Gartner’s 2026 forecast that agentic AI can use 5 to 30 times the tokens per task of standard GenAI, pushing demand toward GW-scale data centers, while 80% green-power consumption is a compliance threshold for new data centers at national hub nodes.