GDS (GDS) Q2 2026 Earnings Call Transcript

GDS Holdings reported net revenue of RMB 3,088.0 million (+6.5% y/y) and swung to net income of RMB 837.6 million (vs. net loss of RMB 70.6 million), driven by a dilution gain tied to DayOne Data Centers. Adjusted EBITDA rose to RMB 1,406.0 million (+2.5% y/y) but margin declined 180 bps to 45.5% due to higher utility costs; gross profit margin fell to 21.5% (from 23.8%). The key upside is operational momentum: Q2 bookings of 260 MW and 1H 2026 bookings of a record 470 MW, lifting the full-year sales target to 1 GW and raising full-year revenue guidance to RMB 12.7–13.0B and adjusted EBITDA guidance to RMB 5.9–6.1B, alongside CapEx raised to RMB 10B (from RMB 9B). Management expects 2027 move-ins to more than double 2026, providing visibility into accelerating future growth.

Analysis

The key setup is not this quarter's earnings power; it is the conversion of contracted capacity into future billings. When bookings are running ahead of delivery capacity, the market usually rewards the visible backlog more than current-period margins, but it also starts discounting execution risk in power access, permitting, and capex discipline. That makes the winners the upstream enablers of buildout—power gear, cooling, construction, and local land/power holders—while weaker China IDC peers with thinner financing access or less scarce capacity become relative losers.

Near term, the stock likely trades on whether investors trust the 2027 move-in ramp, not on 2026 revenue. The biggest bear case is that utility inflation and legacy pricing resets keep gross margin suppressed long enough to mute the operating leverage story; if same-quarter service revenue keeps deteriorating into late 2026, the backlog will be treated as lower quality than management wants. The C-REIT review matters mainly as a balance-sheet catalyst: approval would lower funding risk and support multiple expansion, while delay would keep dilution fears alive despite the strong demand narrative.

Contrarian read: consensus is probably overweighting GPU scarcity and underweighting the breadth of demand from CPU-driven cloud work and reservation conversion. That said, the move can be overdone if the market extrapolates bookings straight into cash flow; this is still a capital-intensive, permit-sensitive business with a lagging P&L. The thesis breaks if 2H26 move-ins slip, year-end backlog fails to clear the targeted threshold, or project financing gets meaningfully tighter than the current 60/40 structure.

Key Decisions for Investors

• Tactically long GDS on any post-earnings pullback; 3-6 month horizon. Risk/reward is best if the market fades the quarter because the 2027 move-in inflection is the real earnings catalyst.
• Pair trade: long GDS / short VNET over the next 1-2 quarters. GDS has the cleaner backlog, better financing access, and stronger visibility into future billings; VNET is the more fragile balance-sheet expression of the same theme.
• Use the call as a watch item on China AI infrastructure rather than a direct long on NVDA. Domestic GPU catch-up reduces the clean read-through to U.S. chip upside; the better trade is the infrastructure bottleneck, not the silicon headline.