Vnet Group (VNET) Q2 2026 Earnings Call Transcript

VNET’s Q2 2026 total net revenues rose 14.2% YoY to RMB 2.78B, driven by a 29.3% YoY jump in wholesale revenues to RMB 1.10B (39.8% of total). Wholesale capacity in service increased 49.4% YoY to 1,007MW (utilized 744MW; 73.9% utilization) and the company booked 347MW of new orders in Q2 with 862MW of YTD wholesale orders; adjusted EBITDA grew 25.4% YoY to RMB 918.3M with a 33% margin and adjusted net income turned to RMB 7.4M from a RMB 53.6M loss. Guidance reiterated for FY2026: revenues of RMB 11.5B–11.8B (+15.6% to +18.6%) and adjusted EBITDA of RMB 3.55B–3.75B (+19.2% to +25.9%) while cash totaled RMB 7.21B as of June 30, 2026. Management also highlighted a structural AI-HPC demand backdrop and a CATL strategic deal to build an integrated compute-energy/zero-carbon token ecosystem, though gross margin was pressured by higher utility pass-through in Q2.

Analysis

VNET is one of the few China infra names where AI demand is translating into contracted revenue rather than just narrative, so the equity should trade more like a capacity-ramp story than a cyclical IDC name. The market mechanism is that committed wholesale megawatts plus seven-year average lease terms reduce downside to revenue, while the real upside comes from utilization turning into incremental EBITDA with minimal SG&A growth. That said, the stock should not be valued as if every reserved megawatt is immediately monetizable; power access, chip availability, and customer deployment timing are the gating variables.

The biggest second-order winner is the power-and-facility stack around VNET: grid, cooling, electrical, and land pipeline assets become more valuable as scarce effective compute capacity, not raw capacity, becomes the bottleneck. The losers are weaker IDC operators with fragmented power access and higher leverage, because they will need to spend more to win less differentiated demand. CATL is strategically interesting but should be treated as a call option on lower-energy-intensity expansion; it is unlikely to move near-term earnings, but it could lower perceived execution risk for new builds and overseas projects.

The main risk is that gross margin keeps leaking from utility pass-through and capex intensity before conversion catches up. If Q3/Q4 move-ins do not accelerate with domestic chip supply, the market will likely re-rate the order book as long-dated rather than near-dated cash flow. The contrarian view is that consensus may be underestimating backlog quality and over-focusing on reported margin noise; however, with leverage still elevated, any miss in delivery timing or guidance would hit equity disproportionately.

Key Decisions for Investors

• Go long VNET on a pullback rather than chasing a gap-up; use a 1-3 month horizon and size for elevated leverage risk. Thesis works if Q3 utilization and wholesale revenue step up as management expects; falsify if move-ins stall or guidance is cut.
• For higher-conviction expression, buy a 3-6 month VNET call spread instead of common stock. This captures backlog-to-revenue conversion while limiting downside if utility pass-through keeps suppressing gross margin.
• Watch for a relative-value long VNET / short a China internet basket or lower-quality IDC proxy if available; the edge is that VNET has the clearest contracted AI capacity runway, while weaker peers face the same power constraints without the same backlog visibility.
• Use a stop/alert if net debt to EBITDA drifts above current levels without matching contracted delivery growth. That would signal capex is outrunning monetization and would break the bullish thesis.