The AI narrative of cloud vendors is shifting from capex to the income statement. UCloud (SSE STAR: 688158) reported for the first three quarters of 2026: revenue of RMB 1.591 billion, up 29.67% year-on-year; net profit of RMB 24.83 million, returning to profitability; AI-related revenue exceeding 40% of the total; and rack utilization above 95% at its Ulanqab intelligent computing center.

Blue-gold tech illustration of wind turbines and data-center campus
The real test of a computing center: not how much was built, but how much is used

1. AI is the engine behind the turnaround

For the vendor once listed as "China's first cloud computing stock," profitability took years. AI-related services now make up more than 40% of the RMB 1.591 billion revenue — GPU computing and intelligent-computing services have replaced traditional public cloud as the primary growth engine, driving most of the 29.67% revenue increase.

2. What 95% rack utilization means

Beyond the income figures, "rack utilization above 95% at Ulanqab" deserves attention. After two years of booming intelligent-computing construction, doubts about "built but unused" capacity have persisted; rack utilization is the key indicator separating a construction boom from operating reality. Ulanqab leverages abundant green power to serve compute demand from eastern China — reports of Shanghai AI firms expanding northward have already surfaced. High utilization suggests inference demand is real and sustained.

3. Multiple confirmations of compute demand

UCloud's report is not an isolated case: optical-module orders booked through end-2027, liquid cooling becoming mandatory, and continued funding for domestic AI chips — upstream, midstream and cloud vendors' revenues are confirming the same thing: AI compute demand has moved from expectation to reported fact.

UCloud Q1-Q3 2026 key figures

· Revenue: RMB 1.591B, up 29.67% YoY

· Net profit: RMB 24.83M, turnaround to profit

· AI revenue share: >40%; Ulanqab rack utilization >95%

The compute industry is completing a narrative switch: markets no longer pay a premium for planned PFLOPS but price what is actually consumed. For compute users, improving utilization of existing resources and scheduling across multiple sources has become the new cost-reduction theme — the very premise of NineZenith's Tianxuan compute scheduling platform.

(Information synthesized from Securitimes, Sina Finance, Eastmoney and other public reports)