West-to-east energy, east-to-west computing: The Uyghur costs of China’s digital order

China is positioning Xinjiang as an AI and digital-infrastructure gateway to Central Asia, supported by 19 ports of entry, 28 cross-border fiber-optic cables and planned computing clusters scaling to hundreds of thousands of accelerator cards. The region produced 510 million metric tons of standard-coal-equivalent primary energy in 2025 and has more than 170 million kilowatts of installed new-energy capacity, underpinning Beijing's strategy to export power, computing capacity and AI services. The SCO's 2026-27 digital-cooperation agenda expands Chinese technology, data systems and standards across Eurasia, but the article highlights substantial surveillance, labor-rights, environmental and data-governance risks that could complicate investor and policy engagement.

Analysis

The investable implication is less a pure AI-demand signal than a state-directed reallocation of compute supply toward low-cost western power regions. This favors Chinese telecom incumbents with regulated network access and balance sheets capable of funding backbone, cloud and edge buildout—China Mobile (0941 HK), China Telecom (0728 HK), and China Unicom (0762 HK)—but raises the risk of excess regional data-center capacity and lower utilization economics for merchant operators such as GDS (GDS). Training workloads can migrate to cheaper-power locations, while inference and latency-sensitive enterprise workloads remain concentrated in coastal hubs; the result could be a widening spread between wholesale AI capacity pricing and premium metro colocation pricing over 6-18 months.

The more material second-order risk is sanctions and ESG contagion. Hardware, networking, solar, and cloud vendors linked to projects in the region could face customer diligence, index-exclusion, or export-control friction even where no direct legal violation is established. That risk is asymmetric for firms dependent on foreign components or overseas customers—ZTE (0763 HK) and Hikvision (002415 SZ) are more vulnerable than domestic telecom operators—while Chinese-made accelerator, server, power-equipment and fiber supply chains may gain protected domestic demand but remain constrained by access to leading-edge chips.

Near term, this is unlikely to move broad China AI equities without disclosed capex budgets, project awards, power tariffs, or utilization commitments. Over 1-3 months, monitor procurement announcements from the three telecom operators and evidence that western compute capacity is being contracted rather than merely announced. The contrarian view is that investors may overvalue headline compute-card targets: constrained high-end GPU supply, transmission bottlenecks, water/cooling requirements, and weak third-party cloud monetization can turn nominal capacity additions into low-return infrastructure.

A reversal of the thesis would be signaled by slowing telecom cloud capex, utilization below management targets, a renewed US restriction on networking/AI-server components, or evidence that coastal enterprises will not shift training workloads inland despite lower power costs.

Key Decisions for Investors

• Maintain a 6-12 month relative-value bias: long China Mobile (0941 HK) versus short GDS (GDS) only after confirmation of operator-led compute/network procurement. Thesis is that regulated connectivity and state-backed utilization capture value while merchant colocation faces incremental supply; exit if GDS reports sustained utilization improvement and pricing growth above China Mobile cloud revenue growth.
• Avoid adding to ZTE (0763 HK) or Hikvision (002415 SZ) on domestic digital-infrastructure headlines alone. Treat any regional project disclosure as an ESG/export-control diligence trigger; downside can be discontinuous if new US, EU, or institutional restrictions emerge, making risk/reward unfavorable absent verifiable overseas-demand resilience.
• Create an alert for China Mobile, China Telecom (0728 HK), and China Unicom (0762 HK) quarterly capex guidance: a meaningful increase in cloud, backbone, or AI-compute spend without corresponding cloud revenue/bookings is a signal to avoid the telecom complex rather than chase it, given historically low incremental returns on policy capex.
• Watch GDS quarterly mainland utilization, signed AI capacity, and EBITDA-per-MW over the next 2-3 earnings cycles. A decline in utilization or new-contract pricing would support the relative short leg; do not initiate on policy rhetoric without these operating data.