Goldman Sachs·2025-07-04 03:04Investment Ratings - The report maintains a "Buy" rating for GDS and VNET, with target prices set at US$37/HK$36 for GDS (20% upside) and US$12 for VNET (69% upside) [1][21] Core Insights - The China data center sector is expected to benefit from deleveraging and improved order visibility, which will support share price performance in the second half of 2025 [1] - GDS is positioned to capture demand from Generative AI-driven cloud growth, while VNET aims for significant capacity expansion under its Hyperscale 2.0 strategy [18][23] Summary by Sections GDS Holdings - GDS's C-REIT offering implies a 16.9x 2026 EV/EBITDA, scheduled for listing on July 14, which supports deleveraging with approximately Rmb1.6 billion proceeds [4][8] - Revenue forecasts for GDS remain largely unchanged, with EBITDA revised down by 7% for 2025E [9] - GDS's financials show a projected revenue of Rmb11,545 million for 2025E, with an EBITDA of Rmb5,335 million [16] VNET Group - VNET has raised its guidance for 2025 revenue and EBITDA, aiming for a total data center capacity of 10GW by 2036 [21][25] - The company has added 555MW in data center capacity from 2019 to 2024 and plans to reach 1GW by 2025 [23][27] - VNET's revenue for 2025E is projected at Rmb9,316 million, with an adjusted EBITDA of Rmb2,810 million [41] Market Dynamics - Improved visibility of chip availability is expected to lead to stronger order volume and faster revenue growth for both GDS and VNET in 2026E [4] - The report anticipates that Alibaba and Tencent's capital expenditures will recover sequentially in the second half of 2025 [5] Financial Metrics - GDS's net debt/adj EBITDA ratio is projected to decline to 5.1x by 2028E due to improving EBITDA and reduced capex [13] - VNET's EBITDA margin is expected to improve to 34.0% by 2027E, reflecting operational efficiencies [41] Strategic Outlook - GDS is focusing on backlog delivery to enhance revenue growth visibility, while VNET's Hyperscale 2.0 strategy emphasizes a platform-based approach to data center operations [18][23]