Where AI compute lands is a power-grid question, not a jobs question

AI compute is landing where clean power is being thrown away, not where the workers are. Rystad Energy analyst Simeng Deng reads the pattern this way: the boxes follow surplus electrons, and the wages follow the owners.

The renewable-curtailment siting logic is the hidden driver. China's 80% clean mandate for new data centers makes the cheapest compliance path a place where wind and solar already overproduce. Rystad Energy's Simeng Deng reads the rural western push as operators chasing that surplus. The build is grid-shaped.

The wage data is the falsifier. TechRadar, citing Taiwan-based researchers at DSET, reports Guizhou posted 7.4% average annual GDP growth over the past decade, even as wage growth ranked second to last among Chinese provinces. Singapore Management University's Andrew Stokols warns data centers do not generate meaningful local employment ripples. The leading read is the capex is solving a curtailment problem, not a regional jobs problem. The alternative is that wage convergence is just slow and the gains arrive later; a decade is long enough to test that, and Guizhou's rank does not flinch.

In China's case under the Eastern Data Western Computing policy, compute routes to provinces with binding clean-power mandates and overproduced renewables, while wages stay concentrated with the owners of the servers, the models, and the dispatch rights. The ladder never arrives with the data center.