Skip to content
voltsdaily

China Southern Power Grid Shows Full-Stack Grid AI Platform at Shanghai WAIC

Share
AI-generated illustration for: China Southern Power Grid Shows Full-Stack Grid AI Platform at Shanghai WAIC
AI-generated image

China Southern Power Grid (CSG) presented an AI-based power platform combining electricity, carbon and computing at the 2026 World AI Conference (WAIC) and the High-Level Meeting on Global AI Governance in Shanghai, according to a China Southern Power Grid release distributed by PR Newswire. The company describes the result as China's first industry-level electricity-carbon-computing coordination platform.

The claim rests on merging domains that utilities elsewhere still run in separate stacks. "We have integrated three systems that traditionally operated independently - electricity, carbon and computing - to establish China's first industry-level electricity-carbon-computing coordination platform," the company said.

At the centre sits MegaWatt Multimodal 2.0, a foundation model purpose-built for the power sector. That is the layer most Western grid operators currently buy rather than own, and CSG says the model is proprietary.

The operational demonstration was narrow but specific. In a live session, the MegaWatt Yunrui system automatically produced a full distribution network planning report, covering risk assessment and grid topology recommendations, in 10 minutes.

That exhibit was named one of the conference's Top 10 Signature Exhibits, and CSG says it was the only project from a central state-owned enterprise to receive the distinction.

Below the model layer, CSG showed PowerAtomOS, a proprietary unified operating system for power-sector connected devices. It is designed to provide centralized connectivity and management for hundreds of millions of devices, and supports critical power equipment across virtually every operational scenario inside CSG's service area.

The stack is described as end to end. CSG says it has built a technology architecture on its digital power grid spanning computing infrastructure, power-sector foundation models, AI agents and field devices, with the company framing the effort as a fully proprietary technology stack covering foundation models, intelligent robots and digital infrastructure.

The policy backdrop matters for how quickly that architecture scales. Coordination between computing infrastructure and power systems was included in China's Government Work Report for the first time this year. For a state-owned grid operator, that inclusion moves data-centre load coordination from an engineering preference to a stated national objective.

The distinction that will interest utility buyers and vendors outside China is ownership of the layers rather than the demonstration itself. Grid operators in liberalised markets typically assemble planning tools, device management and analytics from multiple commercial suppliers, each with its own contract, data boundary and upgrade cycle. A single operator controlling the foundation model, the device operating system and the field hardware collapses those interfaces into one internal roadmap, which is the integration benchmark CSG is now asserting.

Carbon accounting is the second unusual element. Emissions data at most utilities is handled downstream of dispatch, for reporting rather than for real-time decisions. Folding carbon into the same coordination layer as electricity and computing, as CSG describes, puts all three in the same optimisation problem instead of in sequential systems.

How much of the platform is deployed rather than exhibited is not established by the material. The release is a paid placement issued from Shanghai on July 22, 2026, with further inquiries directed to PR Newswire, and the operational evidence made public so far is the 10-minute planning report generated on stage.

Sources

Related