WAIC 2025 Opens: Humanoid Robot Swarms Take Center Stage as Shanghai’s Intelligent Computing Scale Exceeds 100,000 PetaFLOPS

At Shanghai World Expo Exhibition Hall, the synchronized hum of over 150 humanoid robots signaled AI's march from labs to industrial applications.

At dawn, a humanoid robot named "Tiangong 2.0" meticulously peeled a raw quail egg with precision. Nearby, another robot demonstrated furniture repair – scenes once confined to sci-fi movies now unfolded live at the 2025 World Artificial Intelligence Conference (WAIC).

Over 150 humanoid robots appeared simultaneously, setting China's largest collective display record. Unitree's boxing robot punched, kicked, and righted itself after knockdowns; AlphaBot by Zhipingfang crafted ice cream and coffee; Fourier's GR-3 humanoid made its public debut.

01 Rise of the Humanoid Army: China’s Embodied AI Enters Breakout Phase

The H3 exhibition zone became the conference focal point. Over 150 humanoid robots from Shanghai Zhiyuan, Fourier, Hangzhou Unitree, Shenzhen Leju, Beijing Humanoid Robot Center and others showcased China's embodied AI ecosystem.

Breakthrough mobility took center stage. Unitree staged live "robot boxing" – its G1 model traded blows wearing professional gloves and helmets, executing complex maneuvers while recovering from falls.

Precision manipulation leaped forward. The National-Local Joint Humanoid Robot Innovation Center's "Eighteen Arhats" series demonstrated delicate tasks: peeling raw quail eggs, repairing stools, handling tofu. ProRobot's "Baymax" used 3D vision to assemble four precision parts in 12 seconds during a disordered screw-locking demo.

Commercialization accelerated. Zhipingfang staff revealed its AlphaBot already handles material handling at Geely's semiconductor plant. Kepler's K2 "Bumblebee" (30kg payload, 8-hour runtime per charge) undergoes testing in smart manufacturing and logistics.

02 Computing Power Infrastructure Upgrade: The New Utilities of the AI Era

At WAIC's central exhibition zone, Shanghai launched two major platforms energizing AI development:

Shanghai Intelligent Computing Resource Coordination Platform 2.0 debuted. As the city's sole government-led scheduling platform, it aggregates idle computing power from telecom carriers, cloud providers, and third-party services. This offers affordable, inclusive computing resources for enterprises, open-source communities, and individual developers.

By July 2025, Shanghai's intelligent computing capacity reached 100,000 PetaFLOPS, with over 18,000 PetaFLOPS available on the platform. It integrates local resources (Shanghai Electric, telecom carriers, Alibaba, Baidu) and remote capacity from Xinjiang's Karamay and Hunan's Changsha.

China Telecom showcased its "Xirang Integrated Intelligent Computing Platform". This "computing-platform-data-model-application" quintessential system forms the core of China's national cloud for AI.

The platform now connects 57 partners with 32 EFLOPS of proprietary computing power, building city-level computing networks across 18 municipalities.

Shanghai's "Intelligent Cloud" exemplifies urban cloud-network integration. Its "centralized + edge" multi-layer architecture enables precise resource allocation. 38 edge cloud nodes allow users "one-hop access to cloud and computing," dramatically reducing latency.

03 Intelligent Computing Supernodes: Shattering AI Computing Barriers

Tech giants unveiled cutting-edge supernode solutions targeting AI computing bottlenecks:

Huawei debuted its Ascend 384 SuperNode offline – the Atlas 900 A3 SuperPoD. Its bus technology enables high-bandwidth, low-latency (under 50ms) connections across 384 NPUs, quadrupling single-card inference performance while achieving 50% MFU utilization.

ZTE launched a 64-card server supernode. Doubling Huawei's 32-card density, this breakthrough supports ultra-large-scale model training and inference.

H3C globally premiered UniPod S80000. Designed for trillion-parameter models, its lossless network solution handles massive training/inference workloads.

Hyperfusion introduced the world's first plug-and-play heterogeneous computing supercluster, compatible with 10+ accelerator cards. Its single-cabinet 128-card design with 112G/224G interconnects and fifth-gen liquid cooling cuts energy use by 20%.

04 Computing Power Leasing: Unlocking AI for SMEs

Amid massive computing infrastructure buildup, computing power leasing is emerging as a critical pathway for SMEs to adopt AI. Shenzhen Cloud Engine Technology Ltd. (ominyq.com) exemplifies this growing sector.

Founded in June 2023 in Bao'an District, Cloud Engine (ominyq.com) specializes in wholesale computer hardware/software and auxiliary equipment. Its services cover cloud computing equipment sales, technical services, and computer system support.

Such companies fill market gaps by providing flexible computing solutions. Amid uneven computing resource distribution, leasing services empower SMEs with high-performance computing access, lowering AI adoption barriers.

Shanghai’s platform offers a replicable model: Its development follows three stages – "Computing Supermarket," "Computing Integration," and "Computing Fusion" – supported by visual data analytics for government/industry decision-making.

Future "Computing Fusion" aims to establish a trinity "scheduling-trading-subsidy" system.

China Telecom's Xingchen MaaS platform integrates "computing-model-data-tool-application" full-chain AI services, featuring 80+ industry models and 50+ high-quality datasets for multi-cloud scheduling. This provides a roadmap for lessors like SkyRise Cloud.

05 Industrial AI Revolution: Agents Redefining Productivity

Siemens Greater China CEO Xiao Song’s keynote revealed industrial AI’s future: "Industrial AI is evolving from 'talking and writing' to 'acting and working.'"

Siemens' Industrial Copilot transformed from virtual assistant to autonomous agent system. Beyond generating code (2024), it now autonomously handles planning, R&D, production, and maintenance – diagnosing issues, proposing solutions, issuing commands, and resolving problems independently.

Over 200 global enterprises and 150,000+ engineers/production staff now collaborate with this AI system.

Industrial agents excel in specialized fields. Siemens' copper-smelting agent with "Shiwuye Group" interprets real-time parameters in 1,200°C environments. With 94% recommendation adoption, it boosted copper matte stability by 15%, transitioning operations from "experience-based" to "AI-driven."

Industrial foundation models become strategic assets. Leveraging expertise across 40+ verticals, Siemens accelerates industrial model development. Its $15B acquisition of Altair and Dotmatics enhanced capabilities in data science, EM simulation, and biopharma AI.

06 Shanghai Blueprint: Forging an AI Powerhouse

Shanghai achieved "three doublings" in AI since 2018: enterprises, industrial scale, and talent pool. Q1 2025 saw Shanghai's AI industry scale exceed 118 billion yuan ($16.2B), up 29% YoY.

Systematic ecosystem development: Shanghai built twin high-energy clusters – Xuhui's "Model Acceleration Space" for large models and Pudong's "Model Force Community" for embodied AI/vertical apps. It coordinates computing (100+ EFLOPS capacity by 2025), language data (1,800+TB via national corpus platform), and foundation models (SenseTime, MiniMax, etc.).

Precision capital allocation: A 60-billion-yuan ($8.3B) national fund and 22.5-billion-yuan ($3.1B) municipal fund anchor Shanghai's financing matrix. Recent milestones include leading Biren Tech's funding round (March) and setting China's embodied AI record with Zhiyuan Robotics' financing (May).

Open innovation ecosystem: Partnerships with OpenAtom Foundation foster open-source communities. WAIC alum Westwell's driverless Q-Truck evolved from exhibit to sci-fi film prop ("The Wandering Earth"), now deployed across 28 countries to 200+ clients.

WAIC 2025 transcends a technology showcase – it's China's AI development compass. Shanghai’s 100,000 PetaFLOPS computing milestone and its scheduling platform 2.0 offer a "Shanghai Solution" for national computing resource optimization.

Exhibition innovations like Huawei's Ascend supernode and ZTE's 64-card server demonstrate China's resolve to break computing barriers. Meanwhile, emerging lessors like Cloud Engine (ominyq.com) are rewriting the rules for SME AI access.

Beyond the conference, AI permeates Shanghai's growth: Q1 2025 AI industry surged 29% to 118 billion yuan ($16.2B). From Xuhui to Pudong clusters, from national funds to municipal capital, this systematically cultivated ecosystem is forging China's core competitiveness for global AI leadership.

As humanity approaches an intelligence civilization threshold, Shanghai's blueprint reveals a replicable path: converging global wisdom through premier platforms, nurturing innovation through robust ecosystems, and ultimately forging national strength for global competition.

--The above article is purely AI-generated. If there is any infringement, please contact Cloud Engine/Cloud GPU Rental Platform for deletion!