UBS: IDC Sector in Pre-Dawn Quietness, Bullish on H2 Demand Acceleration

Overview

This report previews the Q1 2026 performance of China's Internet Data Center (IDC) industry and outlines future trends. UBS believes that although Q1 results may appear muted due to limited visibility on chip supply, the sector is currently in a 'calm before the dawn' phase. Multiple incremental data points indicate that industry demand will accelerate starting in the second half of 2026, supported by rolling releases of hyperscaler orders, ramp-up of domestic GPU capacity (e.g., Ascend 950), and robust downstream AI demand. The report maintains Buy ratings for GDS and VNET but adjusts valuation methodologies to DCF due to near-term delivery delays and financing uncertainties, resulting in downward revisions to both companies’ target prices.

Core views

The industry’s supply-demand dynamics are improving, with demand acceleration imminent. On the demand side, following large-scale tenders in January, hyperscalers continue to roll out orders; DeepSeek indicated its V4 Pro model pricing will decline in the second half of 2026 as Ascend 950 Supernode capacity ramps up, confirming the scaling trajectory of domestic GPUs; rising token and cloud service prices also signal robust downstream AI demand. On the supply side, the central government remains cautious in approving power quotas, helping prevent overcapacity. Leading operators are positioned to gain market share at this cyclical inflection point. GDS: Steady western expansion, but longer delivery timelines. Since Q3 2025, GDS has been aggressively expanding in China and has accumulated approximately 3GW of power quotas in the west. Given that new campus deliveries typically take over a year, most of these new western projects are expected to be delivered in the second half of 2027. Although EBITDA acceleration will take time, the company has already signed numerous letters of intent and orders, with new orders in 2026 expected to exceed 500MW—making new order intake a more critical driver of share price performance in the upcoming upcycle. VNET: Inner Mongolia campus shows strong appeal; monitor financing progress. UBS views VNET’s Inner Mongolia campus favorably, considering it one of the most attractive campuses in the industry, having already secured orders from multiple hyperscalers. However, if 2026 capital expenditures exceed the announced RMB 10–12 billion guidance (vs. GDS China’s RMB 9 billion), VNET may require additional financing. While the medium- to long-term outlook remains positive, share prices could face near-term pressure until financing plans become clearer. Valuation Adjustment: Shift to DCF to smooth short-term volatility. Given uncertainties around the timing of domestic GPU ramp-ups, UBS has switched its valuation methodology from 2027 EV/EBITDA multiples to a DCF model to better capture the medium- to long-term demand outlook. For GDS, due to extended delivery timelines for western projects, 2027–28 EBITDA forecasts were reduced by 2–6%, and DayOne net profit ownership assumption was adjusted from 100% to 20% (without impacting SOTP valuation), leading to a minor target price reduction from US$62 to US$60. For VNET, 2026–27 EBITDA forecasts remain largely unchanged, but post-2028 forecasts were lowered to reflect financing uncertainty, reducing the target price from US$17.0 to US$15.5.

Analysis framework

UBS employs a combined top-down and bottom-up analytical approach. First, industry checks validate hyperscaler order cadence and domestic GPU capacity ramp-up to identify the sector’s inflection point. Second, company-specific analysis focuses on core assets (e.g., GDS’s western power quotas, VNET’s Inner Mongolia campus), assessing competitive advantages and delivery/financing constraints. Finally, given that short-term chip supply uncertainty may distort traditional multiple-based valuations, UBS adopts a DCF model, anchoring intrinsic value through WACC and terminal growth rate assumptions to objectively evaluate each company’s potential during the industry upcycle.

Methodology notes

• Valuation MethodologyDCF Discounted Cash Flow

  DCF Discounted Cash Flow

  The report switches from EV/EBITDA multiples to DCF because, under technological transitions (e.g., GPU ramp-up) with uncertain timing, multiples can be overly sensitive to short-term sentiment, whereas DCF more stably reflects the sector’s true long-term demand growth potential by discounting future cash flows.

• Valuation MethodologySOTP Sum-of-the-Parts

  SOTP Sum-of-the-Parts

  SOTP is applied to GDS by separately valuing its China operations and DayOne business. This approach better captures the distinct growth drivers and risk profiles of each segment, avoiding valuation distortion from aggregating dissimilar businesses.

• Industry/Sector Analysis FrameworkSupply-demand framework

  Supply-Demand Framework

  The report assesses the industry inflection point by analyzing the alignment between demand-side factors (AI orders, GPU capacity) and supply-side constraints (power quota approvals). In the capital-intensive IDC sector, policy-driven supply constraints—particularly power quotas—are often key determinants of competitive dynamics and profitability.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• GDS (GDS.O)

  Beneficiary: Holds substantial western power quotas and new orders, poised to gain market share in the upcycle.

  Strengths

  Accumulated approx. 3GW of power quotas in the west; 2026 new order expectation exceeds 500MW; efficient supply chain.

  Weaknesses

  Long delivery cycles for new western projects (over one year); most deliveries scheduled for H2 2027, limiting near-term EBITDA acceleration.

  Comparison

  More aggressive than peers in securing western resources, but slower in delivery pace.

  Risks

  Slower-than-expected cloud/AI industry growth; reputational damage from power outages or rising electricity costs; elevated interest rates; intensified competition from oversupply in Southeast Asian data centers.

• VNET (VNET.O)

  Beneficiary: Inner Mongolia campus is highly attractive, with orders already secured from multiple hyperscalers.

  Strengths

  Inner Mongolia campus regarded as one of the industry’s most attractive assets; high-quality customer base.

  Weaknesses

  Potential financing pressure if capex exceeds guidance; short-term share price weighed down by financing uncertainty.

  Comparison

  Superior asset quality, but slightly less financial flexibility than GDS; financing progress warrants close monitoring.

  Risks

  Inability to refinance at reasonable cost; hyperscaler cloud customer growth below expectations; difficulties in securing land and utility access in Tier-1 cities; high churn in retail business.

Key data

• GDS Western Power QuotasApprox. 3GWAccumulated since Q3 2025
• GDS 2026 New Order Expectation500MW+Based on signed MOUs and orders
• GDS Target Price AdjustmentUS$62 → US$60Due to delayed western project deliveries leading to lower EBITDA forecasts
• VNET Target Price AdjustmentUS$17.0 → US$15.5Due to financing uncertainty leading to lower long-term growth expectations
• GDS DCF AssumptionsWACC 7%, Terminal Growth Rate 2.5%Used for China operations valuation
• VNET DCF AssumptionsWACC 7.2%, Terminal Growth Rate 2.5%Incorporates higher net debt to reflect potential financing needs

Impact & implications

The report views the current IDC sector as being on the cusp of a breakout. Investors should focus on leading operators with high-quality asset reserves and strong order acquisition capabilities. Although short-term share prices may fluctuate due to limited chip supply visibility, the medium- to long-term outlook is positive as domestic computing power fills gaps and AI applications deepen, driving tangible industry growth. For GDS, new order acquisition is more important than near-term EBITDA; for VNET, clarity on financing plans will be the key catalyst for share price stabilization.

Risks

• AI demand below expectations
• Data center outages damaging reputation
• Interest rates higher than expected
• Unfavorable regulatory environment
• GDS/VNET-specific risks: financing difficulties, delivery delays, intensified competition

What to watch

• Ramp-up progress of domestic GPUs (e.g., Ascend 950)
• Order release cadence from hyperscalers
• Clarity on VNET’s financing plans
• Delivery progress and utilization ramp-up of GDS’s western projects