CHINA AI EVOLUTIONSearch
Choose how much time you have:2 minGive me the picture5–10 minWhat changed & whyCompany viewWho matters & economicsEvidenceSources & uncertaintyLearnTeach me the system
EXPLAINER · CHINA AI INFRASTRUCTURE

Why is China trying to coordinate AI compute with electricity?

AI workloads consume large amounts of power, while China’s compute demand and electricity resources are unevenly distributed. Compute–energy coordination asks whether suitable workloads can be placed or shifted according to power availability and price. This guide separates the mechanism from the still-unproven national-scale economics.

Compute–electricity coordination · 算电协同Virtual power plant · 虚拟电厂Green electricity · 绿电
LESSON 06 OF 07POWER & ENERGY

Can computing workloads respond to where electricity is available?

By the end of this lesson: Follow the workload-to-power mechanism; distinguish pilots from national-scale economics; understand the cost test that decides whether shifting is worthwhile.

← Learning path
YOU ARE HEREExplainer library
THIS PAGE ANSWERSWhat does “compute–energy coordination” mean?
GO DEEPERArchitecture & evidence →
In plain English: It means treating computing demand as something that can sometimes be placed or shifted in response to where computing capacity and electricity are available. Instead of power being a passive input to a fixed data centre, schedulers, networks, data centres and power systems can potentially interact.
MECHANISM · SIMPLIFIEDCompute and electricity can influence one anotherCoordination is more than attaching a data centre to the grid.
Grid & renewablesavailability / price
⇄
BESS / power systemsbuffer / stabilise
⇄
AIDCphysical load
⇄
Schedulerwhen / where suitable work runs
⇄
AI workloadlatency & compute needs

Still unproven at national scale: that workloads are sufficiently portable and that dynamic coordination produces better economics after network, storage and operational costs.

The mechanism

A portable workload can be identified by a compute scheduler, sent over the network to a suitable data centre, and executed where capacity and power conditions are more favourable. The electricity side may involve grid signals, virtual power plants, storage or renewable availability.

What is actually evidenced

Chinese policy explicitly calls for compute–electricity coordination. Operating evidence is now stronger than a single pilot: State Grid Shanghai reports a coordinated 16-data-centre event and a three-day electricity-price-aware cross-region inference test; a separate Zhejiang report describes partial workload migration from Jiaxing to Qinghai. These demonstrate that suitable workloads can respond to power conditions in bounded settings. They do not establish national scale, universal portability or durable commercial savings.

Evidence state · 25 Sep 2026: the mechanism is operationally demonstrated in bounded deployments. The harder test is repeatability, net cost after networking/software, and whether a broad enough set of workloads is actually movable.

The economic test

The relevant question is not whether a workload can move. It is whether the combined savings in power and infrastructure exceed network, latency, software, data-movement and operational costs while preserving useful performance.

Evidence, claims and limits

This explainer separates documented policy or company statements from our interpretation. Company performance claims are not treated as independently verified economics unless external evidence supports them.
CONTINUE THROUGH THE RESEARCH
CHECK YOUR UNDERSTANDING · 2 QUESTIONS

Can you apply the idea?

Choose an answer, then reveal the reasoning. This is not scored — the aim is to practise distinguishing what evidence proves from what it merely suggests.