VNET Group Inc (NASDAQ:VNET) reported a significant increase in net revenue, up 18.3% year over year to RMB2.25 billion for the fourth quarter.
The wholesale business achieved remarkable growth, with revenue reaching RMB665 million, marking a 125.4% year-over-year increase.
Adjusted EBITDA for the fourth quarter increased by 63.8% year over year to RMB721.3 million, with a margin improvement to 32.1%.
VNET Group Inc (NASDAQ:VNET) secured a total of 252.5 megawatts in orders from customers across various industries, indicating strong demand.
The company achieved a full-year net profit of RMB248 million, marking a turnaround from a net loss in 2023.
Negative Points
The wholesale business utilization rate decreased slightly to 72.6% due to concentrated delivery of 127 megawatts in the fourth quarter.
Capital expenditure for 2025 is expected to be significantly high, ranging from RMB10 billion to RMB12 billion, representing a 101% to 141% increase year over year.
The retail IDC business utilization rate was relatively low at 63.5% as of the end of December 2024.
Despite strong revenue growth, the company faces potential supply shortages in key regions like the Great Beijing area by 2025.
VNET Group Inc (NASDAQ:VNET) has a high net debt to trailing 12 months adjusted EBITDA ratio of 4.9, indicating significant leverage.
Q & A Highlights
Q: Could management provide details on the arrangement of CapEx for 2025 and how much of this spending will be covered through the funds from our projects? Also, could management provide insights into the potential impact of AI-related orders on the retail segment? A: In 2025, over 90% of CapEx will be directed into the wholesale IDC business, with 400 megawatts to be delivered. 83% of these are determined orders. The remaining CapEx will go into retail IDCs for high-powered density and value-added business. AI applications, like autonomous driving, are driving demand, and we expect higher pricing for AI server deployment-friendly cabinets.
Q: Could you walk us through the framework partnership with the Changzhou partner and how that is reflected in our revenue or guidance? Also, how should we think about the supply and demand dynamics over the next one or two years? A: The JV with Changzhou SOE, where we hold 13% equity, is not included in our CapEx or balance sheet. We profit through management and capital operation fees. Third-party data suggests a supply shortage in the Great Beijing area by 2025 and the Yangtze River Delta by 2026. Most peers are focusing on overseas markets, reducing competition domestically.
Q: What's your expectation of the per kilowatt rental in the next one or two years? Is it likely to see some price hike? Also, will there be any pricing pressure on unit CapEx due to supply chain constraints? A: The demand for wholesale IDC centers is high, and supply is slightly short, leading to stable pricing driven by market demand. While traditional equipment brands face shortages, Made-in-China equipment is increasingly used, slightly lowering costs. Overall, costs are slightly lower, and revenue is expected to increase.
Q: Do you think your retail revenue this year can get back to mid-single-digit growth, and will it further accelerate in 2026? How will you fund the RMB10 billion CapEx in 2025? A: The threshold for SMEs and individual developers has decreased, leading to increased demand for retail IDCs. We expect mid-single-digit growth in retail revenue. CapEx will be funded through operating cash flow, project financing, and low-interest loans, with a significant portion already covered by existing cash flow and financing arrangements.
Q: Could you share your observations on customer preferences between in-house data centers and third-party operators? How do your wholesale campuses cater to inference versus training demand? A: Most customers prefer renting over building in-house due to resource efficiency. Wholesale campuses are designed for high efficiency, while inference models are typically deployed in retail IDCs for convenience. Latency is not a major issue for these applications, making retail IDCs suitable for inference needs.
