Earlier research and evidence history (2)
Demand aggregation is not a delivery update
New regional computing marketplaces may create demand for AI-ready capacity, but no new GDS customer allocation, delivered MW, utilisation or cash return follows from the platform announcements.
Wholesale demand does not eliminate pricing risk
New sector modelling distinguishes AI-suitable wholesale campuses from underutilised legacy retail facilities. Stronger AI order volumes may coexist with lower market rental rates and delayed cash returns. GDS analysis should separately track contracted, energised and revenue-generating capacity; rental rates; customer commencement; refinancing; and asset-recycling proceeds. No new named customer relationship or revised company forecast is established by the sector estimates alone.
GDS faces the same system-wide questions, not the same disclosed project exposure.
The Horinger Huawei–China Telecom project and China’s computing–power policy demonstrate a wider infrastructure direction, not a new GDS contract. The US credit sell-off increases the importance of separating GDS’s wholesale data-centre financing from GPU ownership, while its offshore operations require their own capital and technology-transfer analysis.
Where GDS stands today
What large AI customers, delivered capacity and pricing mean for GDS.
How does this operator compare with the other listed China AIDC platform?
A fact-led side-by-side study now tracks demand, delivery, utilisation, economics, financing and emerging infrastructure integration without assigning a winner.
GDS has a new connection to provincial compute coordination.
Its participation in Hebei's integrated computing-power platform connects private AIDC infrastructure to provincial monitoring/scheduling. Dynamic workload routing and economics remain unproven.
Reviewed snapshot: 21 September 2026. This profile organises the existing systems thesis; it is not a new valuation or investment rating.
Position in the value chain
Power → substations → liquid cooling → AIDC campuses → geography / connectivity
The question for investors
Does AI-ready MW scarcity translate into durable returns on large physical capacity?
Ecosystem context
Hyperscaler and network ecosystems
Relationship status must be checked in the linked research; adjacency is not evidence of a commercial agreement.
Confirmation and challenge
Look for company-specific deployment, paid usage and cash-return evidence. Technical progress alone does not establish value capture; delays, weak utilisation or rising costs may offset it.
Offshore exposure and the quality of reported earnings
GDS’s domestic operating estate and its minority investment in DayOne are different economic exposures. GDS describes DayOne as an independent Singapore-headquartered hyperscale platform. Offshore demand may affect the investment’s value, but DayOne capacity and revenue should not be counted as wholly owned GDS operating capacity or revenue.
In Q2 2026, GDS reported net income of RMB837.6m. Its RMB959.9m share of equity-method investee results arose mainly from a dilution gain following DayOne’s Series C preferred-share issue. That gain is distinct from recurring data-centre operating profit and cash generation; subtracting it is not, by itself, a complete adjusted-profit calculation.
What to track: domestic move-ins, recurring margins and cash returns separately from DayOne ownership dilution, investment valuation and any distributions or realisation proceeds. The reported Tencent–Oracle lease does not establish a DayOne or GDS customer contract.
National compute-network integration
New evidence · September 2026. GDS participated as a data-centre representative in the launch of Hebei's integrated computing-power platform, described by GDS as China's first formally operational provincial compute monitoring and scheduling platform. Two GDS Langfang data-centre projects were launched at the same conference.
This creates an evidenced edge between a major private AIDC operator and a provincial orchestration layer. It does not yet show that workloads are dynamically routed into GDS capacity, disclose commercial economics, or establish GDS as the platform operator. The next test is evidence of connected GDS resources, scheduled workloads, utilisation/revenue effects or a defined commercial interface.
GDS — 2026 China Computing Power Conference / Hebei platform
25 SEPTEMBER 2026 · EVIDENCE EXTENSION
GDS appears in a real compute–electricity coordination event; Horinger also gains independent location-level corroboration.
State Grid Shanghai names GDS / 万国数据 among the companies whose 16 data centres participated in July’s coordinated event. This connects a major AIDC operator to a live grid/compute coordination process. The source does not establish that GDS supplied the specific cross-region workload transfer, disclose incremental revenue, or prove a new commercial scheduler role.
Separately, SINNET / 光环新网 (300383.SZ) has increased planned investment in its own Horinger intelligent-compute project to about RMB2.5bn and 90–120MW. That is a different project from GDS’s Horinger programme, where GDS announced an approximately RMB20bn five-year multi-campus plan and previously disclosed a 45MW Horinger hyperscale order. Multiple operators scaling the same node is location-level demand/infrastructure evidence, not evidence of a shared customer or commercial relationship.
State Grid Shanghai · coordinated event · SINNET filing · 300383.SZ · GDS Horinger plan · GDS 1Q26 · 45MW Horinger order
Research and relationships
22 SEPTEMBER 2026 · RESEARCH UPDATE
A larger potential market, allocations still unknown
Alibaba’s ambition matters to the demand outlook, but GDS-specific contracted capacity and project returns must be evidenced separately.