VNET Group beats revenue estimates on AI demand strength; Shares edge higher

VNET reported Q4 revenue of RMB2.69 billion, up 19.6% YoY and beating the RMB2.62 billion consensus, while adjusted EPS was RMB0.17, missing estimates by RMB0.01. Wholesale revenue surged 47.1% YoY to RMB978.1 million and adjusted EBITDA rose 11.6% to RMB805.1 million, though EBITDA margin declined to 30.0% from 32.1% a year earlier. For fiscal 2026 the company guided revenue of RMB11.5–11.8 billion (midpoint RMB11.65 billion vs analyst RMB11.58 billion) and adjusted EBITDA of RMB3.55–3.75 billion; capacity in service reached 889MW (vs 486MW a year earlier) with 135MW of new wholesale orders in the quarter.

Analysis

VNET sits at the center of a bifurcated data‑center market where scale and execution on wholesale contracts drive disproportionate cash flow improvements while rapid capacity additions create near‑term margin stress. The second‑order winners from this cycle are not just hyperscalers but power and site‑construction vendors (medium‑term) and GPU suppliers (near‑term), because the bottleneck for useful capacity increasingly becomes compute allocation and power availability rather than shell space.

Key risks are concentrated and time‑phased: in the next 3–6 months, booking cadence and GPU allocation signals will determine utilization and pricing power; in 12–36 months, macro and supply overhangs (aggressive buildouts across peers) can force spot leasing and compress margins. Regulatory moves affecting high‑end GPU exports or tighter Chinese demand would be binary catalysts that could flip valuation quickly, while proof of sticky multiyear wholesale contracts would materially derisk the story.

The consensus leans optimistic on AI demand but underweights two asymmetries — customer concentration/contract mix and the capital intensity needed to keep power and PUE competitive. If market underestimates the durability of fixed‑price wholesale contracts VNET outperforms peers; if it underestimates short‑term oversupply, downside is sharper than multiples suggest. Watch wholesale utilization, new order conversion rates, and GPU allocation notices as the highest‑value datapoints over the next four quarters.

Key Decisions for Investors

• Long VNET (VNET) — buy post‑earnings weakness into a 6–12 month horizon. Position size 2–4% portfolio, target +30% if utilization ticks up and gross margins stabilize, stop at -15% if sequential wholesale bookings fall or GPU allocation warnings emerge. Rationale: execution on wholesale orders and scale benefits should reprice cash flow if utilization stays >~80%.
• Relative trade: long VNET / short GDS — 12‑month pair sized dollar‑neutral. Thesis: VNET’s rapid wholesale execution and land‑cheap capacity expansions can outperform GDS if customer mix and contract tenors favor committed wholesale; pair mitigates China macro beta. Take profits if spread widens >20% or compress if both report synchronized demand weakness.
• Hedge/option idea: buy NVDA 3–6 month call spread (e.g., buy 1x shorter‑dated ATM call, sell 1x higher strike) to express continued tightness in GPU supply without naked long exposure. This acts as a positive demand hedge for data‑center capacity upside; cap premium to defined risk with asymmetric payoff if GPU scarcity persists.
• Downside protection: buy VNET 9–12 month put spread sized to 50% of equity exposure (protect half the position). Use this instead of outright short to limit carry cost; unwind if quarterly wholesale utilization prints show sustained improvement.