Goldman Sachs Initiates Coverage on Runjian Co., Ltd. with Buy Rating, Bullish on AI Compute Leasing and Expansion in Western Hubs

Overview

This report, published by Goldman Sachs, focuses on structural shifts within China’s data center industry amid expanding AI infrastructure demand. It initiates coverage on two A-share data center operators: Runjian Co., Ltd. (300442.SZ) and Shanghai Data Harbor (603881.SS). The core thesis is that rising AI investment is driving a shift toward high-density infrastructure and frontier markets (e.g., western hubs). The GPU-as-a-Service (GPUaaS) market has experienced sharp price increases due to supply-demand imbalances, resulting in highly attractive unit economics. Runjian Co., Ltd. — with its massive capacity reserve, full-stack AIDC capabilities, and diversified low-cost financing channels — receives a Buy rating. In contrast, Shanghai Data Harbor, though operationally stable, receives a Neutral rating due to its conservative expansion strategy and high customer concentration. Additionally, Buy ratings are maintained on Hong Kong- and U.S.-listed peers including GDS and 21Vianet.

Core views

The GPUaaS market is experiencing simultaneous volume and price growth. Driven by AI inference demand (e.g., multimodal inference, AI agents) and constrained by memory shortages and extended semiconductor manufacturing ramp-up timelines, GPUaaS rental prices in both the U.S. and China have risen ~30% year-to-date. Goldman Sachs estimates that, based on current Nvidia H100 cluster rental rates, the business can generate double-digit five-year internal rates of return (IRRs) and achieve payback in 3–4 years if utilization remains above 80%. Securing long-term demand is thus critical to sustaining healthy profitability. Runjian Co., Ltd. (Buy): Growth potential remains underpriced. As one of China’s largest data center operators, Runjian Co., Ltd. currently operates 750 MW of capacity and holds a 6 GW reserve (5.4 GW in mainland China, most of which has already passed Neng Ping review), the largest among covered peers. The company is transitioning from traditional IDC to an integrated intelligent computing operator, with projected 2025–2028 compound annual growth rates (CAGRs) of 40% for revenue and 47% for EBITDA. Key strengths include: 1) abundant capacity reserve enabling ~290 MW of annual delivery; 2) deep penetration into leading internet giants and AI firms, with top-five customers’ revenue contribution declining to 49%; 3) full-stack AIDC capabilities, including liquid cooling and deep network optimization; and 4) diversified low-cost funding (interbank market rates as low as 1.65%, and planned C-REIT issuance). The Rmb117 target price is derived from discounted 2030E EV/EBITDA. Shanghai Data Harbor (Neutral): Solid but lacking breakout potential. As a state-owned enterprise, Shanghai Data Harbor operates 401 MW of capacity, primarily serving Alibaba (>98% of 2025 revenue). Its expansion strategy is relatively conservative, relying mainly on delivery of its Langfang project (43 MW) and renewal of existing contracts. Although exploring AIDC upgrades and land acquisition in western regions, near-term capacity expansion visibility remains limited, alongside elevated customer concentration risk. Valuation-wise, the Rmb34 target price reflects a 20x 2027E EV/EBITDA multiple, indicating current valuation already incorporates its growth outlook. Industry Trend: Western hubs emerge as the new expansion frontier. Under the ‘East-to-West Data Transfer’ and ‘Computing-Electricity Synergy’ policies, western hubs such as Inner Mongolia, Ningxia, and Gansu — endowed with abundant green power resources and low costs (per-unit capex ~20% lower than Tier-1 cities; operating costs ~1/3 lower) — are becoming the primary source of new compute supply. National Development and Reform Commission (NDRC) window guidance on new data centers makes pre-approved Neng Ping capacity critically valuable, potentially accelerating industry consolidation.

Analysis framework

Goldman Sachs employs a hybrid top-down and bottom-up analytical framework. First, it analyzes structural AI compute demand growth at macro and industry levels, comparing cloud vendors’ capex in China and the U.S. to highlight China’s catch-up potential. Second, it dissects the unit economics of the emerging GPUaaS business model, with sensitivity analysis showing high utilization and high rental rates as key profit drivers. Third, at the stock level, it emphasizes ‘capacity reserve’ and ‘Neng Ping approval’ as critical barriers to entry, treating them as core indicators of mid-term expansion potential. Finally, it applies a sum-of-the-parts (SOTP) approach combined with EV/EBITDA multiples, assigning differentiated valuations to distinct business segments (e.g., traditional IDC vs. high-growth GPUaaS) to reflect divergent growth profiles.

Methodology notes

• Valuation MethodEV/EBITDA valuation

  Applying enterprise value multiples (EV/EBITDA) to value the capital-intensive data center sector, using differentiated multiples across business segments (e.g., traditional IDC vs. high-growth GPUaaS) via weighted or segmented valuation.

  Data centers are capital-intensive with substantial depreciation and amortization. EV/EBITDA better reflects operating cash flow generation capability. In this report, Runjian’s traditional IDC business is assigned a 25x multiple, while its higher-risk, higher-growth GPUaaS business receives a 10x multiple, reflecting differentiated risk-return profiles.

• Industry/Vertical Analysis FrameworkSupply-demand framework

  Analyzing pricing pressure in the GPUaaS market arising from supply constraints (chip shortages, production ramp-up lags) and surging demand (AI inference, large-model training).

  The report notes that tight supply of high-end GPUs and lengthy manufacturing cycles, coupled with explosive AI demand, have created a seller’s market, driving sharp rental price increases. This supply-demand mismatch is the principal driver of short-term outsized returns.

• Corporate Fundamentals & Financial FrameworkFree cash flow analysis

  Assessing the impact of capital expenditures (capex) on free cash flow, particularly large-scale GPU procurement and data center construction.

  The report details projected capex plans for Runjian and Shanghai Data Harbor, evaluating whether their funding channels (e.g., C-REITs, bank loans, convertible bonds) are sufficient to support expansion, underscoring financing capability as a key growth constraint.

• Industry/Vertical Analysis FrameworkUpstream-Midstream-Downstream Value Chain Transmission

  Analyzing price and cost transmission mechanisms across the value chain — from upstream chips/hardware, through midstream IDC/GPUaaS providers, to downstream cloud vendors/AI application layers.

  The report observes rising upstream hardware costs and robust demand transmitting upward to midstream GPUaaS rental rates; meanwhile, downstream cloud vendors are adjusting pricing models (e.g., token-based billing) to pass on costs, illustrating strategic bargaining across chain segments.

Asset mapping & comparison

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

• Runjian Co., Ltd. (300442.SZ)

  Beneficiary: Possesses industry-leading 6 GW capacity reserve and strong GPUaaS capabilities, directly benefiting from AI compute demand surge and rental rate increases.

  Strengths

  Largest capacity reserve, full-stack AIDC technology, diversified low-cost financing, optimized customer mix.

  Weaknesses

  Massive capex driving higher leverage, execution risk in overseas expansion.

  Comparison

  Fastest growth among peers (revenue CAGR 40%), with reasonable valuation (22x 2027E EV/EBITDA).

  Risks

  Lower-than-expected order intake, slower-than-anticipated utilization ramp-up, GPU supply constraints, financing difficulties.

• Shanghai Data Harbor (603881.SS)

  Neutral: Operationally sound but slow-growing, highly dependent on single customer Alibaba, lacking new growth engines.

  Strengths

  State-owned background, strong operational track record, low debt-to-equity ratio.

  Weaknesses

  Excessive customer concentration (>98% from Alibaba), conservative expansion strategy, lagging AIDC transformation.

  Comparison

  Lower near-term growth visibility versus Runjian; valuation at historical average.

  Risks

  Langfang project delivery delays, fluctuations in Alibaba’s demand, contract renewal uncertainty, subpar AIDC execution.

• GDS Holdings (VNET)

  Beneficiary: Transitioning toward wholesale IDC, benefiting from accelerated AI investment.

  Strengths

  Rising wholesale IDC revenue share, accelerating growth.

  Weaknesses

  High leverage, pricing pressure.

  Comparison

  Faster transformation pace than traditional retail IDC providers.

  Risks

  Financing capacity, geopolitical risks.

Key data

• Runjian Co., Ltd. Target PriceRmb117Based on discounted 2030E EV/EBITDA, implying 31% upside
• Shanghai Data Harbor Target PriceRmb34Based on 20x 2027E EV/EBITDA, implying 8% downside
• GPUaaS Rental Rate Increase~30%Average YTD increase across U.S. and Chinese markets
• Runjian Co., Ltd. Capacity Reserve6 GWOf which 5.4 GW in mainland China, majority approved under Neng Ping
• Western IDC Unit Capex Advantage-20%Relative to Tier-1 city markets
• Runjian Co., Ltd. 2025–28E Revenue CAGR40%Significantly above industry average

Impact & implications

For Runjian Co., Ltd., the report concludes that its massive capacity reserve and leadership in GPUaaS position it as a core beneficiary of AI compute expansion and rental rate hikes, with its share price not yet fully reflecting its high-growth potential. For Shanghai Data Harbor, operational stability is acknowledged, but the lack of clear expansion catalysts and high customer concentration constrain its valuation ceiling. For the industry, operators with pre-approved Neng Ping capacity and low-cost financing will further consolidate their leadership positions, accelerating industry consolidation. Investors should focus on western hub players possessing dual advantages of ‘compute + power’.

Risks

• Intensifying competition leading to lower-than-expected order intake
• Slower-than-expected customer utilization ramp-up
• Greater downward pressure on GPUaaS rental rates
• Execution risk and geopolitical uncertainty related to overseas expansion (Hong Kong, Indonesia)
• Regulatory changes or domestic capacity delays causing high-end chip supply shortages
• Tighter financing conditions or delayed C-REIT issuance affecting capex execution

What to watch

• Delivery progress and customer rack-up rates for Runjian’s Langfang B1 and other liquid-cooled clusters
• Sustainability of GPUaaS rental rates and long-term contract signings
• Delivery timeline for Shanghai Data Harbor’s Langfang project and evolving demand from Alibaba
• Neng Ping approval status for western hubs (e.g., Ningxia, Inner Mongolia)
• Progress on regulatory approvals for C-REIT asset injections