Executive Summary / Key Takeaways
-
GDS has engineered a structural transformation through its C-REIT and ABS platforms, creating a permanent capital recycling advantage that delivers levered IRRs "well into the 20s" while competitors remain trapped in traditional build-and-hold models.
-
The company sits at the epicenter of China's AI infrastructure buildout, with a 152-megawatt mega-deal in Q1 2025 validating its ~900 megawatts of powered land in Tier 1 markets that is effectively impossible to replicate under current regulatory constraints.
-
DayOne International represents a hidden equity stub worth approximately $7 per ADR (18% of current market cap) that is growing revenue at 244% year-over-year, with an IPO targeted within 18 months to crystallize this value.
-
Despite accelerating AI-driven growth, GDS maintains financial discipline: net debt/LQA EBITDA has compressed from 6.8x to 6.0x, the effective interest rate has dropped to 3.3%, and the China business is approaching self-funding status after asset monetization proceeds.
-
The central risk is not demand—AI bookings are running at 65% of new commitments—but chip supply uncertainty in China, which management prudently monitors with a "wait and see" approach for deployments beyond the next few quarters.
Setting the Scene: From Capital Trap to Capital Light
GDS Holdings Limited, founded in 2001 and headquartered in Shanghai, has spent two decades building what is now China's premier hyperscale data center platform in Tier 1 markets. The business model appears straightforward: develop high-density data centers, secure long-term contracts with hyperscale customers, and generate stable cash flows. But this simplicity masks a profound strategic evolution that has transformed GDS from a capital-intensive developer into a capital-efficient operator with a structural cost-of-capital advantage that no domestic competitor can currently replicate.
The data center industry in China operates under unique constraints. Power quotas in Tier 1 markets have become increasingly restrictive, with window guidance favoring state-owned enterprises and carbon allocation policies tightening around Beijing, Shanghai, and Shenzhen. Simultaneously, the market is experiencing an unprecedented demand inflection as domestic tech giants commit hundreds of billions of yuan to AI infrastructure. This creates a bifurcated market: those with pre-secured power and land resources are positioned to capture the AI wave, while those without face existential challenges.
GDS's competitive moat rests on two pillars that reinforce each other. First, the company has secured approximately 900 megawatts of powered land in and around Tier 1 markets, with carbon quotas already obtained. This is not merely a real estate portfolio; it is a strategic asset that is effectively non-replicable under current regulatory regimes. Second, GDS has established the first and only data center C-REIT in China, which trades at 24.6x projected 2026 EBITDA—nearly double the valuation multiple at which GDS itself trades. This valuation arbitrage creates a permanent capital recycling engine that fundamentally alters the unit economics of data center development.
Technology, Products, and Strategic Differentiation: The Capital Recycling Flywheel
The core technological advantage of GDS is not a patent or a software platform, but a financial innovation married to irreplaceable physical assets. The C-REIT and ABS transactions completed in 2025 represent more than financing events; they are the activation of a capital flywheel that delivers compelling returns while de-risking the balance sheet.
Consider the unit economics: GDS invests in new data centers at development costs that are "near historic lows," ramps them over 12-18 months, and then monetizes stabilized assets at 13-17x EBITDA through the ABS and C-REIT platforms. Management calculates the unlevered IRR at "low to mid-teens" and the levered IRR on equity at "well into the 20s" over a 5-6 year cycle. The significance lies in the fact that every dollar of equity invested in new development generates returns that are 2-3x higher than traditional hold-and-operate models, while simultaneously removing debt from the balance sheet through deconsolidation.
The C-REIT IPO achieved a 16.9x EBITDA multiple at the RMB 3 per unit offering price, and units have since traded up to RMB 4.04, representing a 35% gain and an implied 22.8x 2026 EBITDA multiple. The dividend yield is projected at 3.6%, providing income-oriented domestic investors with an attractive vehicle. For GDS shareholders, the company can develop assets at 8-10x EBITDA and sell them at 16-17x EBITDA, capturing the spread while recycling capital into the next development cycle. This is a repeatable platform that management intends to feed with RMB 4-6 billion of additional assets in 2026.
This capital advantage directly enables GDS's AI positioning. The 152-megawatt deal signed in Q1 2025—the largest single order in China's data center history—requires delivery within six months and full ramp within twelve. Such rapid deployment is only possible because GDS had pre-secured powered land and could commit capital with certainty. The deal is for AI inferencing , which is latency-sensitive and must be located within established cloud regions in Tier 1 markets. GDS's 900 megawatts of powered land is precisely the inventory needed for this demand wave, while competitors scramble for power quotas and suitable sites.
Financial Performance: Evidence of Strategy Working
GDS's financial results in 2025 demonstrate that the capital recycling strategy is delivering tangible improvements in leverage, margins, and cash generation. Revenue grew 10.8% year-over-year to RMB 11.29-11.59 billion, while adjusted EBITDA grew 10.8% to RMB 5.19-5.39 billion.
The adjusted EBITDA margin expanded to 47.3% for the full year, up from 47.2% in 2024, despite a 2.6% decline in monthly service revenue per square meter in Q1 and a 1.7% decline in Q2. This margin stability proves that GDS is successfully offsetting price resets on legacy contracts with higher utilization, faster move-ins, and stable pricing on new builds. Management expects MSR to decline 3-4% in 2026, but this is balanced by volume growth from the 79,000 square meters of gross move-ins achieved in 2024 and the accelerating booking pace in 2025.
The balance sheet transformation is equally compelling. Net debt to LQA adjusted EBITDA fell from 6.8x at year-end 2024 to 6.0x at Q3 2025, and pro forma for the C-REIT transaction it stands at 5.9x. The effective interest rate has dropped to 3.3%, reflecting improved credit metrics and access to diverse funding sources. In Q2 2025, GDS raised USD 676 million through convertible bonds and equity, strengthening the holdco balance sheet without adding restrictive covenants. This financial flexibility is the prerequisite for capturing AI demand, as customers increasingly evaluate suppliers based on financial capability to deliver multi-megawatt campuses on short timelines.
Capital expenditure discipline reinforces the thesis. Organic CapEx for the first nine months of 2025 was RMB 3.8 billion, with full-year guidance of RMB 4.8 billion. However, net of asset monetization proceeds, CapEx is expected to be only RMB 2.7 billion. Operating cash flow is projected at RMB 2.5 billion for 2025, making the China business nearly self-funding. This is a dramatic shift from the historical model of relying on external debt and equity to fund growth.
DayOne: The Hidden Value Driver
DayOne, GDS's international equity investee, represents a material component of value. Following the Series B equity raise on December 31, 2024, GDS's ownership was diluted from 52.7% to 35.6%, with DayOne recognized as an equity investee rather than a consolidated subsidiary. Based on the Series B benchmark, GDS's equity interest is valued at approximately $1.3 billion, or $7 per GDS ADR—roughly 18% of GDS's current market capitalization.
The growth trajectory is extraordinary. DayOne's total committed power reached 467 megawatts by end-2024, surged to over 530 megawatts in Q1 2025, and exceeded 780 megawatts by Q3 2025. The company is ahead of schedule to achieve its 1 gigawatt target within three years. In Q2 2025 alone, DayOne added 246 megawatts of new commitments, including an anchor customer for its Thailand campus—the largest in the country—and secured a second site in Finland, where access to renewable energy and competitive power tariffs create an attractive operating environment.
Financial performance validates the strategic rationale. DayOne's Q2 2025 revenue grew 244% year-over-year, with adjusted EBITDA up 265%. The EBITDA margin reached 31% in Q1 2025, and management expects it to reach industry benchmark levels within a few years as rapid ramp-up drives operating leverage. The development yield in DayOne's markets is in the "low-teens," higher than GDS currently achieves in China on a comparable investment cost basis.
The IPO plan is more visible, with a target to list within 18 months from Q4 2024. A Series C round could value DayOne between $4-5 billion pre-money, suggesting GDS's 35.6% stake could be worth $1.4-1.8 billion at IPO—potentially higher than the current $1.3 billion carrying value. For GDS shareholders, this represents a free option on international data center demand, which remains robust across Southeast Asia and Europe, with no exposure to China chip supply constraints.
Outlook and Execution: The AI Inflection
Management's guidance and commentary reveal a company at an inflection point. GDS is on track to achieve the midpoint of its revenue guidance and at or above the top end of its EBITDA guidance for 2025. Internal projections foresee higher bookings next year, leading to gross acceleration thereafter. This is grounded in tangible market dynamics.
The domestic tech industry has reached a critical juncture with major players making unprecedented financial commitment to AI infrastructure, marking a definitive end to the previous downturn. All major customers have announced CapEx plans in the hundreds of billions of yuan. The coming waves of AI demand are expected to be multiples of gigawatts over the next few years, primarily for inferencing workloads that require low-latency access to Tier 1 markets.
GDS's strategic response is calibrated aggression. The company is taking a more aggressive approach to new business while maintaining financial discipline, prioritizing deals that match its inventory of powered land with fast move-in schedules and long contract tenors. The 152-megawatt deal exemplifies this approach: a high-quality, AI-driven commitment from an existing hyperscale customer with no chip supply risk over the next 12 months. The contract requires delivery within six months and full utilization within twelve, a timeline that only GDS can meet at this scale due to its pre-positioned land and capital.
The unit economics of new investment are compelling. Selling prices for new builds have been stable for two years, while development costs are almost at their historical bottom. This combination produces an 11-12% cash-on-cash yield on new investment and levered IRRs "well into the 20s" when monetized through the C-REIT. Management believes this is the right time to invest because the risk/reward is asymmetrically favorable.
Risks and Asymmetries: What Could Break the Thesis
The most material risk to the investment thesis is chip supply uncertainty in China. While management states they do not see any significant risk for deployments over the next few quarters and are willing to commit to new business with short lead times, they adopt a "wait and see" approach for deployments further into the future. CEO William Huang notes that chip supply is not very clear, and the company is watching this situation closely.
The significance of this lies in the fact that the AI demand narrative assumes that domestic GPU development will catch up within 12 months, enabling the next wave of deployments. If chip supply constraints persist beyond 2026, the timing of GDS's bookings and move-ins could be delayed, impacting revenue growth and the velocity of the capital recycling flywheel. The risk is moderate in probability but high in impact, as it could slow the AI wave that justifies the current investment cycle.
Power quota and government regulation present a second risk. While leadership indicates that local governments are willing to work with the company, new guidance favoring SOE investment in AI data centers could theoretically disadvantage private players. However, GDS's 900 megawatts of powered land already has obtained the power energy quota, making it largely immune to new policy shifts. The risk is more acute for competitors trying to secure new sites, which inadvertently strengthens GDS's relative position.
Customer concentration remains a structural vulnerability. The top five customers represent approximately 70% of revenue, creating exposure to hyperscaler capex cycles. However, this concentration also reflects GDS's strategic focus on the highest-quality customers with the strongest balance sheets and most aggressive AI investment plans. The risk is mitigated by long contract tenors and the fact that these same customers are driving the AI demand wave.
On the upside, the asymmetry lies in the pace of AI adoption. If domestic chip supply resolves faster than expected, or if Chinese customers accelerate their AI investments to match U.S. hyperscaler spending patterns, GDS's bookings could exceed the baseline case. The company's 900 megawatts of powered land provides immediate capacity to capture this upside, while competitors would face 12-24 month lead times to secure comparable resources.
Valuation Context: Discounted Flywheel with a Free Option
At $39.91 per share, GDS trades at an enterprise value of $12.82 billion, representing 17.8x TTM EBITDA. This multiple appears reasonable for a data center business growing EBITDA at 10-12% with 47% margins. However, this valuation ignores the structural transformation that has occurred.
The C-REIT, which GDS controls and feeds with assets, trades at 24.6x projected 2026 EBITDA—nearly a 40% premium to the parent multiple. This implies that GDS's stabilized asset portfolio is worth substantially more than the market currently recognizes. If GDS were valued on the same multiple as the assets it creates, the stock would trade at a significant premium to current levels. The discount reflects market skepticism about execution and chip supply risks, but it also creates upside as the capital recycling model proves itself through repeated transactions.
DayOne represents a free option worth $7 per ADR, or 18% of the current stock price. With 780 megawatts committed and a path to 1 gigawatt, DayOne's equity value could appreciate significantly ahead of its IPO. The international business diversifies geopolitical risk and captures AI demand outside China, where chip constraints do not apply.
Free cash flow is negative at -$176 million TTM due to growth capex, but operating cash flow is positive at $476 million, and the China business is approaching self-funding status after asset monetization. The net debt/EBITDA ratio of 6.0x is elevated but trending down, and the effective interest rate of 3.3% is manageable. For a company capturing a generational AI demand wave with a proven capital recycling model, the valuation appears to discount execution risk while ignoring the structural advantages that have been built.
Conclusion: The Flywheel Is Turning
GDS Holdings has engineered a rare combination: a capital recycling platform that generates 20%+ levered IRRs, a strategic land bank that is impossible to replicate, and a front-row seat to China's AI infrastructure buildout. The 152-megawatt mega-deal in Q1 2025 is not an isolated event but a harbinger of the "multiples of gigawatts" of inferencing demand that will require precisely the assets GDS has assembled.
The C-REIT and ABS transactions are not financial engineering but a fundamental transformation of the business model. By monetizing stabilized assets at 16-17x EBITDA and redeploying capital into new developments at 8-10x EBITDA, GDS has created a self-reinforcing flywheel that competitors cannot match. This advantage is amplified by the 900 megawatts of powered land with pre-secured carbon quotas, which insulates GDS from the regulatory constraints that will limit new supply.
DayOne provides a separate path to value creation through its 244% revenue growth and imminent IPO, while the core China business approaches self-funding status. The primary risk—chip supply uncertainty—is being managed prudently with short-lead-time commitments and a "wait and see" stance on longer-term deployments.
For investors, the thesis hinges on two variables: the velocity of the capital recycling flywheel and the resolution of chip supply constraints. If GDS can execute two to three asset injections into the C-REIT annually while domestic GPU capacity catches up to demand, the company will generate returns that justify a significant re-rating. The current valuation appears to discount execution risk while ignoring the structural moat that has been built. As the flywheel turns and AI demand accelerates, that discount should narrow, creating meaningful upside for shareholders who recognize that GDS is no longer a capital-intensive developer but a capital-efficient platform for China's AI infrastructure.
