Quick ReadPlain-language overview · read this first
GDS Holdings (GDS.US) is China’s leading carrier-neutral third-party data center provider (IDC, meaning data center infrastructure that bundles power, buildings, cooling, and operations into leased capacity; carrier-neutral facilities allow customers to connect to major carrier networks and stay close to major public clouds). The report rates it “Hold.” The core view is that AI orders are accelerating and the capital recycling channel is now in place, but leverage and the pace of execution still leave the margin of safety too thin.
GDS earns revenue by leasing high-performance data center capacity and providing long-term service contracts, typically with terms of 7 to 10 years. The company now has a two-layer structure: a consolidated core IDC business in mainland China, plus DayOne, the overseas platform that has been carved out and is accounted for under the equity method (only the profit or loss attributable to the stake is recognized; it is not consolidated). The key fundamental tension lies here: reported revenue rose 23.6% year on year in the first quarter of 2026, and adjusted EBITDA (earnings before interest, taxes, depreciation, and amortization) rose 47.2%, which looks strong. After stripping out one-off items, however, normalized revenue increased only 7.9% and normalized EBITDA only 8.0%. Of the quarter’s net profit of 2.652 billion yuan, 2.137 billion came from DayOne equity-method income rather than core leasing profit, so earnings quality needs to be unpacked. Orders are indeed accelerating: net new bookings in the first quarter were about 200MW, the highest single-quarter level in the company’s history. The moat rests on scarce data center resources (land, power quotas, and rack deployment schedules are hard to replicate quickly), deeply embedded top-tier customers, and established domestic asset securitization channels through REITs/ABS. Customer switching costs are extremely high.
On valuation, the current price of 32.56 US dollars and market capitalization of about 6.52 billion US dollars place the stock in the lower half of the report’s neutral range of 32 to 40 US dollars. The report’s ideal buy zone is 19 to 22 US dollars, corresponding to a low-multiple valuation for core China, a discount on the DayOne stake, and roughly a 20% margin of safety. 50 to 58 US dollars is the clear overvaluation line. The look-through EV/EBITDA after excluding DayOne is about 11 to 12 times, close to peer VNET but well below the 22 to 24 times of global REITs, with the gap reflecting the China discount, the VIE structure, and high leverage.
There are three main risks: high leverage, as the company has clearly accepted net debt/EBITDA rising from about 4.7 times to 5 to 6 times; customer concentration, with the largest customer alone accounting for 37.9% of committed area; and price pressure, with monthly service revenue per square meter expected to fall another 3% to 4% by the end of 2026. The report is therefore cautious and neutral: GDS is worth tracking and worth holding for existing positions, but at the current price it looks more like an asset to hold while waiting for execution than a stock that should be bought heavily right now.
The above is a summary of the report’s views and does not constitute investment advice. The stock market involves risk; invest with caution.
LeadGDS Holdings is China's leading carrier-neutral third-party data center operator, packaging power, cooling, and operations into long-term capacity leases across domestic facilities and its overseas DayOne interest. Normalized Q1 2026 revenue grew only 7.9%, while RMB 2.137 billion of RMB 2.652 billion in net income came from DayOne gains and net leverage was about 4.7x. Research rating Hold: AI orders are accelerating and the capital recycling loop is opening, but leverage and execution leave the margin of safety too thin.
Prices in the article are as of publication; see the valuation band above for the live price.
Metadata
Ticker: GDS.US
Full company name: GDS Holdings Limited
Current price and market cap: 32.56 USD / about 6.52 billion USD (as of the Nasdaq close on 2026-06-12; market cap estimated based on the 1,603,020,903 ordinary shares disclosed by the company on 2026-04-29 and the ratio of 1 ADS = 8 ordinary shares)
Currency: USD (all valuation and price figures use USD throughout; the company's original reporting currency is RMB, and conversions use the exchange rate disclosed in the company's Q1 2026 report, 1 USD = 6.8980 RMB, unless otherwise noted)
Report date: 2026-06-14
Industry classification: Data centers
One-sentence positioning: China's leading carrier-neutral third-party IDC operator, monetizing long-term capacity leases and operations services.
Research Summary
This report is part of zh.app's "AI value chain" coverage, with a research baseline date of 2026-06-14. The main analytical lens is the U.S.-listed ADR, while Hong Kong-listed 9698 is used as a secondary listing reference. The investment horizon covers both the next 12 months and the next 3 to 5 years. The perspective is heavy-asset infrastructure, not asset-light internet. That premise defines the focus: whether AI compute demand can keep converting into billable contracts, whether capital expenditure will push the balance sheet into a dangerous zone, and whether the REIT/ABS asset securitization channel can turn part of the heavy asset base back into liquidity.
GDS is, at its core, an infrastructure company that sells a bundled package of "power, buildings, cooling, network redundancy, and compliant operations." It earns money from high-performance data center capacity leases, long-term service contracts, and a small amount of value-added service fees, rather than traffic monetization in the style of an internet platform. In its early days, it started with disaster recovery and business continuity for financial institutions, then moved into self-developed high-performance data centers. The current story has evolved into a two-layer structure: the consolidated core IDC business in mainland China, and the overseas platform DayOne, which has been carved out and is accounted for under the equity method. By the Q1 2026 report, the market's main question had moved away from the old "is the company an IDC leader" debate toward "is AI demand real orders or only MoUs," "is improving domestic chip supply enough to generate sustained move-in," and "does the company dare, and can it execute, a three-year RMB 30 billion to 50 billion investment plan that turns demand into revenue."
The main narrative being traded by the market is clear: AI is steepening the demand curve for China's IDC industry again, and after a downcycle GDS happens to hold three cards that support a new story. The first card is the order book. Q1 reporting showed about 200MW of net new bookings in the first quarter, and management disclosed on the Q4 2025 earnings call that since the beginning of the year it had secured 200MW of new orders plus more than 500MW of MoUs, with total demand of 700MW mainly from the top three customers. The second card is the balance sheet. The DayOne equity sale and the USD 300 million private placement of convertible preferred shares provided additional ammunition. The third card is capital recycling. The company first completed a private REIT/ABS transaction in 2025 and then a C-REIT IPO, proving that an onshore Chinese asset securitization channel has already been opened.
The key point about this company is precisely that upside and risk are tied together. On the surface, Q1 2026 looked very strong: revenue rose 23.6% year on year and adjusted EBITDA rose 47.2%. But once one-off items and DayOne-related investment gains are stripped out, the true trend of the core China business is roughly 7.9% revenue growth and roughly 8.0% adjusted EBITDA growth. This is not bad news. It is a reminder that the company is still in a transition phase where orders accelerate first and revenue is recognized later. In the same quarter, operating cash flow was only RMB 448 million, while payments for property, plant, equipment, and land-use rights were RMB 770 million. The cash flow profile did not suddenly become easy just because of the AI narrative.
The share-price history maps to the same narrative. GDS reached an all-time closing high of 115.71 USD in February 2021, then was dragged down by cooling China internet capex, the regulatory discount on Chinese ADRs, valuation compression on long-duration assets under high interest rates, and the company's own high leverage and free cash flow pressure. In 2025-2026, the stock recovered as AI demand improved, DayOne's valuation became more visible, REIT asset monetization progressed, and orders recovered. Yet as of the 2026-06-12 close it was still only 32.56 USD, far below the 2021 peak. This is a company undergoing valuation reset, not one returning to the old track. The old story was the traditional Chinese IDC leader. The new story is a China AI infrastructure platform plus an asset securitization machine.
The most important bull-bear debate now centers on four points. Bulls focus on domestic chip supply becoming less ambiguous than in 2024, leading customers resuming large-scale orders, and the company's ability to use REIT/ABS channels to shift capital on and off the balance sheet, producing simultaneous upgrades to revenue, EBITDA, and valuation in 2026-2028. Bears focus on the other side: customer concentration is high, with the top five customers representing a very large share of committed area; management itself acknowledges that monthly service revenue per square meter is still falling and expects another 3% to 4% decline by the end of 2026; leverage may rise from about 4.7x now to 5x to 6x; and a large portion of the "strong" Q1 net income came from equity-method gains related to DayOne dilution and share sales, not profit from core cabinet leasing.
If looking only at the company profile, my label for GDS is "in valuation reset." Its capital intensity is too high and cash flow is still far from loose enough to compound like a snowball, so it does not qualify as high-quality compound growth. There is indeed a new AI logic behind the demand recovery, which also makes it different from pure distressed recovery in a cyclical turn. In 2026-2028, it will still be in a construction phase of high investment, high financing, and high expansion, far from a mature cash cow. The easiest mistake with this type of company is to mistake "demand improvement" for "the cash flow problem has been solved." GDS is worth studying and even worth continuing to hold, but it has not yet entered the category of infrastructure asset that one can hold blindly for the long term.
Company History
From Disaster Recovery Outsourcing to a Self-Built IDC Platform
GDS was not originally the "compute real estate company" it is today. When the company began operating in 2001, its core identity was that of an IT services provider focused on business continuity and disaster recovery solutions, with financial institutions as its early customer base. Because those customers demanded very high reliability and compliance, GDS first leased data center capacity from third parties and then bundled service capabilities for customers. At that stage, the company was solving the problem that large Chinese institutions did not want to bear the complexity of building and operating high-standard server rooms themselves. It was not yet about a compute boom.
The turning point came in 2009. Management wrote it plainly in the annual report: at the time, they recognized a shortage of qualified data center capacity in core markets, and began developing self-owned data center assets. In 2010-2011, the company launched three facilities in Shanghai and Chengdu. In 2013-2014, it further entered Beijing and Shenzhen, and by 2014 it had covered China's main established markets. From that point onward, GDS's business model changed from "leasing others' capacity and reselling services" to "owning or controlling facilities and turning power and floor area directly into long-term contracted revenue." This step determined the company's current form, and also embedded its heavy-asset, high-depreciation, high-leverage nature.
STT Capital, Nasdaq Listing, and Hong Kong Secondary Listing
In 2014, STT GDC injected critical capital into GDS, helping the company enter a true scale expansion phase. The annual report described the financing as a "major infusion of capital," after which the two sides formed a long-term strategic partnership. For GDS, this meant more than money. More importantly, it gained global data center industry experience, customer resources, and a governance framework. Many Chinese IDC companies during expansion tend to fall into a pattern of many projects, expensive funding, and loose governance. STT's involvement at least made GDS look more like an infrastructure company that global capital could understand.
In 2016, GDS completed its Nasdaq IPO, and in 2020 it completed a secondary listing in Hong Kong. The listing path itself was not especially complicated, but the market backdrop differed sharply between the two stages. In 2016, U.S. investors were willing to assign a high growth premium to Chinese cloud computing and internet infrastructure. In 2020, Hong Kong was more of a capital-market backup after the regulatory environment for Chinese ADRs changed. For today's investors, the important point is the reminder that GDS has always been an asset shaped by the intersection of U.S.-China capital-market sentiment, regulatory risk, and China's digital infrastructure investment cycle.
Four Phases Show How the Company Grew
The first phase was the service validation period from 2001 to 2008. The company built credibility through outsourced disaster recovery for financial institutions and developed operational know-how around high availability, low failure rates, and strong compliance. This phase did not have today's large-scale capex, but it explains why GDS later won leading customers.
The second phase was the self-build expansion period from 2009 to 2015. The company decided to own and develop high-performance data centers, quickly entering tier-one cities and core economic belts. The constraint was funding. Choosing this path rather than continuing with asset-light services was essentially management's bet that China's internet and cloud computing industries would keep high-standard facility supply tight for a long time. That judgment later proved correct, but the cost was also obvious: the balance sheet never became light again.
The third phase was the capital-market highlight period from 2016 to 2021. After listing, GDS benefited simultaneously from rising Chinese internet capex, higher cloud penetration, and valuation expansion for long-duration growth assets. The stock reached its historical peak of 115.71 USD in February 2021. At this stage, the market believed in the "scarce neutral IDC leader in China" story, worried little about leverage, cared little about cash flow, and focused on long-term capacity scarcity and binding relationships with leading customers.
The fourth phase has been the revaluation period since 2022. First demand and valuation cooled together. Then the company placed its overseas business into DayOne. In late 2024, after external financing at DayOne, GDS stopped consolidating the business and switched to equity-method accounting. In 2025, it completed a private REIT/ABS transaction and a C-REIT IPO in China. In 2026, it began shifting the narrative back to a new AI-driven expansion cycle. The biggest change in this phase is that the company finally moved from a single model of self-building data centers with a debt-heavy balance sheet to a three-part structure: core China platform, overseas equity investment, and onshore asset securitization.
DayOne Separation Is an Accounting Event and a Capital Structure Event
DayOne deserves separate discussion. In May 2022, GDS established DigitalLand Holdings, later renamed DayOne, as the holding platform for overseas data center assets and operations. It is headquartered in Singapore and has an independent management team. In 2024, DayOne completed a total of USD 672 million in Series A financing and USD 1.2 billion in Series B convertible preferred share financing. GDS chose not to participate, so by the end of 2024 its stake in DayOne was diluted to 35.6%, and from then on GDS stopped consolidating DayOne and accounted for it under the equity method. By December 2025, DayOne signed the first tranche of a Series C financing of more than USD 2 billion, of which USD 1.3 billion closed by the end of 2025. In January 2026, DayOne repurchased USD 385 million of shares from GDS. By April 2026, GDS's remaining stake in DayOne had fallen further to about 19.9%, and management said its value had exceeded USD 2.2 billion.
This has a large impact on the financial statements. On one hand, after the overseas business was deconsolidated, GDS's revenue, capital expenditure, and debt looked more domestic. On the other hand, equity-method investment income and dilution gains can lift net income in certain quarters, making the income statement look better than the core leasing business. Q1 2026 is a typical example: of RMB 2.652 billion in net income, a large portion came from RMB 2.137 billion of equity-method income from DayOne, not from the core China IDC business itself. Understanding this is necessary to avoid using surface-level P/E to conclude that the company has suddenly become asset-light.
Financial Review and Share-Price History
Over the past three years, GDS's operating cash flow has been acceptable: operating cash flow in 2023, 2024, and 2025 was about RMB 2.07 billion, RMB 1.94 billion, and RMB 3.37 billion, respectively. But investing cash flow remained deeply negative, especially in 2023 and 2024. Only in 2025, because of DayOne deconsolidation and asset monetization, did cash flow pressure temporarily ease. In other words, GDS's problem has never been "it cannot make money." It is that "the cash it earns is still not enough to naturally cover expansion." This is a typical infrastructure company trait: the income statement and cash-flow statement are directionally consistent, but free cash flow can easily turn negative again during expansion.
The 2025 annual report is easy to misread. Full-year revenue grew 10.8% year on year and adjusted EBITDA grew 10.8%, which looked stable. But net income turning positive to RMB 959 million did not mean the company had crossed the capital cycle. It came more from deconsolidation gains and equity-method income caused by DayOne financing. The company itself also noted in the Q4 report that if the revenue and EBITDA removed through deconsolidation were added back, 2025 pro forma growth would be 13.2% and 14.2%, respectively. That is closer to the true growth slope of the core business.
The share-price path is also clear: 2016-2021 was growth-stock expansion; 2021-2023 was valuation compression plus China risk discount; 2024-2026 was a rebound driven by AI order recovery, visible overseas equity value, and asset securitization revaluation. The issue is that today's 32.56 USD, while far below the 2021 peak, is no longer the pure "leverage panic price" of 2022-2023. The market has begun paying for the new story, just not at the high multiples of global REITs.
Business Model and Industry Cycle
GDS's Business Machine Is About Contracts and Power, Not Photos of Cabinets
GDS's revenue mainly comes from data center services. Customers include hyperscale cloud service providers, large internet companies, financial institutions, telecommunications carriers, and large enterprises. The company is carrier-neutral and cloud-neutral. This neutrality is central to the business model, not marketing language: customers can access major carrier networks in its facilities and stay close to major public cloud nodes. GDS is therefore more like a "connection field" that lets multiple parties coordinate on the same physical infrastructure than a server room attached to one cloud vendor.
The first layer of moat is geography and power location. A truly high-value IDC must balance low-latency established markets with new growth markets that have policy, power, land, and network conditions. It is not a matter of building wherever land is cheap. GDS's existing platform covers core regions such as Shanghai, Beijing, Shenzhen, Guangzhou, Chengdu, and Chongqing, while it is also increasing investment in national hub nodes such as Inner Mongolia and Ningxia. On the Q4 2025 earnings call, management mentioned that it was advancing a 3GW pipeline and already held 700MW of powered land as future development reserves. This is hard to replicate quickly because it involves land, power quotas, move-in timing, customer pre-commitments, local approvals, and engineering organization.
The second layer of moat is leading customer relationships and delivery track record. The company disclosed that as of the end of 2025, the largest customer accounted for 37.9% of committed area, the second-largest 11.8%, and the third-largest 8.8%. This shows high concentration, but it also shows deep binding with leading customers. For large-model companies and cloud vendors, the scarce asset is not an empty building. It is a data center campus that can deliver on schedule, support high-density loads, provide stable power, and solve network latency through cross-region coordination. Customers do not migrate easily just because the price is a little lower, because migration itself carries very high engineering, network, business interruption, and recertification costs.
The third layer of moat is capital access. For data center companies, project returns matter, and financing channels matter just as much. In 2025, GDS completed a private REIT/ABS transaction with an implied EV/EBITDA of about 13x. It also completed a C-REIT IPO in the same year, with an issue price corresponding to an implied EV/EBITDA of 16.9x and a unit distribution yield of 5.2%. These two transactions proved that stable mature assets can be priced in China as infrastructure assets rather than ordinary manufacturing assets. Once this channel is open, the company's capex flexibility becomes greater than if it relied purely on banks and USD bonds.
Cost Structure and Operating Leverage
This business has heavy fixed costs. Buildings, power distribution, cooling systems, land-use rights, network redundancy, and depreciation all require upfront spending before customers move in. The main costs that can vary quickly with revenue are electricity, part of operations, and sales expenses. As a result, the company naturally carries heavy depreciation in accounting terms. Management repeatedly emphasizes that Gross Profit does not represent the core operating trend well and prefers to use Adjusted Gross Profit and Adjusted EBITDA to assess ramp-up. Financially, this means new facilities can look very unattractive on margins at launch because of depreciation, but mature assets gradually improve. Conversely, if orders weaken, depreciation does not disappear immediately and the income statement comes under pressure quickly.
This is why GDS's "appears profitable" and "true owner earnings available to common shareholders" are not the same thing. Operating cash flow in 2025 was RMB 3.365 billion, well above net income of RMB 959 million. But payments for property, plant, equipment, and land-use rights in the same year were RMB 4.611 billion. Excluding proceeds from asset monetization, equity free cash flow was still tight. Q1 2026 was similar: operating cash flow was RMB 448 million, while payments for property, plant, and equipment were RMB 770 million. For an IDC company, P/E should not be the main axis. EV/EBITDA and the capital payback period matter more.
Management, Governance, and Industry Cycle
Chairman and CEO William Huang is the founder and has served as CEO since 2002. The company's path closely resembles that of a typical founder-led infrastructure company: betting on direction early, betting on financing in the middle, and betting on capital allocation later. Management's most important delivery over the past two years lies in three completed events: DayOne deconsolidation and equity monetization, the private REIT/ABS transaction, and the successful C-REIT IPO. Together, they show that management is at least not merely talking when it comes to finding capital exits for assets.
The governance discount cannot be ignored. GDS still relies on a VIE structure to conduct its core business in China, and the annual report shows that 97.5% of 2025 revenue came from the VIE and its subsidiaries. The company explicitly acknowledges that it has no equity ownership in the VIE and can only control operations through a series of contractual arrangements. This is not a new risk, but it always exists and can be amplified by changes in the regulatory environment for Chinese ADRs. For an infrastructure asset, this means GDS cannot be valued like a U.S. REIT purely on stable rent. It must continuously bear a China regulatory and structural discount.
From a cyclical perspective, GDS is exposed simultaneously to the capex cycle, interest-rate cycle, policy cycle, and technology iteration cycle. In an upcycle, the most beneficial variables are leading customers' AI capex and the availability of domestic high-performance chips. In a downcycle, the most fragile variables are financing costs, project delivery schedules, and move-in delays. This industry is not defensive. It only looks smoother because contract terms are long.
Peer Analysis
Competitive Landscape: Two Reference Systems
If looking only for Chinese peers, GDS's most direct public-market comparable is VNET. If the research is meant to support an investment judgment, mature global REIT references also need to be included, namely Equinix and Digital Realty. In A-shares, Runze Technology represents another Chinese compute infrastructure valuation logic that has been significantly lifted by the AI narrative and REIT expectations. In other words, GDS carries two comparison frameworks at the same time: a local comparison as a "Chinese third-party IDC platform," and a quality comparison as part of "global data center infrastructure." Precisely because these two reference systems imply very different valuations, GDS stays for a long time in the gray zone of being both cheap and not cheap.
VNET has become another version of a Chinese IDC. It is also a carrier-neutral and cloud-neutral platform, but its recent growth has leaned more clearly toward wholesale IDC and AI-related demand. In Q1 2026, VNET revenue grew 19.8% year on year to RMB 2.69 billion. IDC business revenue grew 27.0%, wholesale IDC revenue grew 58.1%, and adjusted EBITDA grew 30.6% year on year. This is a more "wholesale, volume-driven" path with a higher growth slope, but its margins and asset quality do not clearly lead GDS. The market values it more like a high-beta China AI infrastructure stock than stable infrastructure.
Equinix and Digital Realty are completely different species. Equinix's strengths lie in dense interconnection and a high-quality retail/interconnection ecosystem, with a mature platform, regional diversification, and extremely strong capital-market credibility. After Q1 2026, the company raised full-year guidance, with full-year adjusted EBITDA guided to USD 5.165 billion to USD 5.245 billion. The market assigns it a market cap of about USD 104.1 billion and enterprise value of about USD 124.4 billion. Digital Realty leans more toward wholesale and platformized campuses. Its Q1 2026 adjusted EBITDA was USD 920 million, up 16% year on year, and the market assigns it a market cap of about USD 64.7 billion and enterprise value of about USD 82.2 billion. Both earn far higher valuations than Chinese peers because their dividend predictability, legal structures, financing costs, and asset securitization maturity are all stronger.
Runze Technology provides a third mirror. Its 2025 revenue was RMB 5.674 billion, up 29.99% year on year, and net income attributable to shareholders was RMB 5.050 billion, up 182.07% year on year. But that figure includes distortion from non-recurring gains related to a public REIT listing. The company itself also emphasizes that growth excluding non-recurring items was far below headline growth. It represents the end of the A-share market willing to pay a higher growth premium for "AI infrastructure + REIT platform + domestic policy support." The difference between GDS and Runze is not only business detail. It is listing venue and investor structure. The former bears both the Chinese ADR discount and VIE discount, while the latter benefits from A-share preference for AI infrastructure.
Key Peer Data Table
| Metric | GDS | VNET | EQIX | DLR |
|---|---|---|---|---|
| Latest closing price | 32.56 USD | 8.87 USD | 1,055.85 USD | 184.20 USD |
| Market cap | About 6.52 billion USD† | 2.52 billion USD | 104.1 billion USD | 64.7 billion USD |
| Enterprise value | 11.88 billion USD‡ | 5.66 billion USD | 124.4 billion USD | 82.2 billion USD |
| Latest core revenue growth | Q1 2026 normalized about +7.9% | Q1 2026 +19.8% | 2026 guidance about +10% to 11% | Q1 2026 EBITDA +16% |
| Latest core EBITDA growth | Q1 2026 normalized about +8.0% | Q1 2026 +30.6% | 2026 guided EBITDA margin 51% | Q1 2026 Adjusted EBITDA 920 million USD |
| Rough EV/EBITDA | About 13.7x-14.2x; about 11x-12x after DayOne | About 10.9x | About 23.9x | About 22.4x |
Market cap, EV, and growth metrics in the table are rough calculations based on each company's market data and latest disclosed financial reports/guidance as of 2026-06-12. GDS's market cap is estimated using publicly disclosed share count. VNET, EQIX, and DLR market caps and EV use public-market snapshots. GDS's EV/EBITDA would decline meaningfully if the value of management-disclosed remaining DayOne equity, exceeding USD 2.2 billion, were deducted.
This table shows how the market prices each company, not "which one is cheaper." GDS and VNET have close valuations, but GDS has higher EBITDA margins and a clearer asset securitization path, so in theory it deserves some quality premium. Global REITs are almost twice as expensive because of financing structure, legal environment, dividend discipline, and a fully different capital-market pricing system, not because their server rooms glow on their own. GDS's current niche is China's leading high-performance third-party IDC operator and one of the Chinese data center platforms with the greatest AI order elasticity, while still carrying a China discount and construction-phase discount. It most directly competes for the local third-party IDC profit pool. What it is most likely to lose is pricing power in the next round of AI wholesale orders, not existing customers.
Current Fundamentals and Valuation Analysis
The Last Four Quarters Show Strong Headline Profit, but Core Operating Growth Is Less Dramatic
GDS's fundamental state by Q1 2026 can be summarized in one sentence: orders lead, financial recognition lags. In Q4 2025, revenue grew 8.6% year on year and adjusted EBITDA grew 5.2%; full-year revenue and adjusted EBITDA both grew 10.8%. In Q1 2026, revenue grew 23.6% year on year and adjusted EBITDA grew 47.2%, but after excluding one-off items, revenue and adjusted EBITDA grew only 7.9% and 8.0%, respectively. The most obvious improvements were bookings, pre-commitment rates for projects under construction, and company-level cost discipline, not a sudden explosion in current-quarter leasing revenue.
Operating metrics are moving in the right direction. By the end of March 2026, committed and pre-committed area grew 11.7% year on year to 725,485 square meters; area in service grew 10.4% year on year to 674,269 square meters; utilization rose from 75.7% a year earlier to 77.3%; and the pre-commitment rate for projects under construction rose from 71.6% to 84.4%. These figures show that demand is continuing to convert into contracts rather than remaining only in management commentary. More importantly, William Huang said directly in the Q1 report that net new bookings in the first quarter were about 200MW, the highest single-quarter level in the company's history.
However, earnings quality still needs to be separated. Q1 net income was RMB 2.652 billion, of which equity-method investment income was RMB 2.137 billion, mainly from DayOne financing dilution gains and partial share sales. Net interest expense fell 14% year on year, but the absolute amount was still RMB 380 million. Operating cash flow was RMB 448 million, while capex payments were RMB 770 million, showing that the core business is still in a capital-consuming phase. Current liabilities also included RMB 4.285 billion of current convertible bonds, so short-term debt repayment and refinancing arrangements still need monitoring.
What the Market Is Trading Now
The market today is trading whether China's AI training and inference infrastructure will enter a new construction wave like the United States, not a broad "IDC industry recovery." GDS management was explicit on the Q4 2025 earnings call: domestic high-performance chip supply availability has improved; since the beginning of the year the company has locked in 200MW of orders and more than 500MW of MoUs; total demand of 700MW mainly comes from the top three customers; and the company has begun accelerating the acquisition of multi-gigawatt powered land, focusing on national hubs such as Inner Mongolia and Ningxia. It even lifted its annual gross bookings target for the next three years to 500-800MW. This narrative is large enough to move the stock from a "balance sheet pressure stock" back into an "AI infrastructure beneficiary."
But there is still a time lag between fundamentals and market narrative. The CFO also mentioned on the call that business won in Q1 2026 usually takes four quarters or more to form move-in, with contract terms mostly 7 to 10 years and the vast majority close to 10 years. In other words, GDS is now selling higher-certainty future revenue, not current revenue that is already fully occupied and steadily settled. If move-in is slower than the market expects, the stock can pull back first and wait for earnings to catch up.
Bull-Bear Debate
Bulls have three strongest pieces of evidence. First, chip constraints are less ambiguous than in 2024, and management has moved from "wait and see" to actively securing land and building a 3GW pipeline. Second, after the private REIT/ABS and C-REIT succeeded, the company found a healthier capital recycling method in China than simply adding leverage. Third, after DayOne deconsolidation, overseas assets no longer consume consolidated capex and have instead begun feeding value back to the parent through visible equity value.
Bears also have three strongest pieces of evidence. First, customer concentration is high: the top five customers account for nearly 70% of committed area, and the top three alone exceed half. If leading customers delay usage or substitute with self-build, the impact will be very specific. Second, management acknowledges that MSR, or monthly service revenue per square meter, is still declining. It fell 2.4% year on year in Q4 2025 and is expected to fall another 3% to 4% by the end of 2026. Third, the company explicitly accepts net debt/EBITDA rising from about 4.7x now to 5x to 6x, which means that if the financing environment does not cooperate, REIT injections are not smooth, or orders convert into move-in more slowly than expected, shareholders face capital-structure pressure rather than merely an EPS miss.
Valuation Analysis
I use three main valuation anchors for GDS: EV/EBITDA for the core China business, look-through EV/EBITDA after deducting the value of the DayOne stake, and the mature-asset multiples implied by completed asset securitization transactions. The current market snapshot shows that as of 2026-06-12, GDS had a market cap of about USD 6.64 billion and enterprise value of about USD 11.88 billion. Based on the company's 2026 EBITDA guidance of RMB 5.75 billion to RMB 6.00 billion, reported EV/EBITDA is roughly 13.7x-14.2x. If management's disclosed "remaining DayOne equity value exceeding USD 2.2 billion" is deducted, core China look-through EV/EBITDA falls to about 11x-12x.
This multiple is not unreasonable among Chinese peers and is close to VNET's rough 10.9x. It is also clearly cheap versus the global REIT system, where Equinix and Digital Realty both roughly trade around 22x-24x. The difference is higher risk, not that the market has failed to notice GDS: VIE, customer concentration, leverage, policy, electricity prices, geopolitics, and listing-venue discount all prevent it from easily enjoying the U.S. REIT valuation framework. Conversely, as long as China AI orders are fulfilled, REIT follow-on offerings continue to land, and leverage does not get out of control, GDS has no reason to be priced forever as a traditional low-growth Chinese IDC asset.
Looking through cash flow, this company should not be judged only by accounting net income. Operating cash flow in 2025 was RMB 3.365 billion, higher than net income, but expansion-phase capex still suppressed free cash flow. The company has not disclosed a precise split between maintenance capex and expansion capex, so I treat valuation conservatively: I assume most capex in 2025-2027 is still expansionary, but owner earnings cannot simply equal operating cash flow and cannot simply equal net income. For GDS, this means "cheapness" must be grounded in EV/EBITDA, asset payback multiples, and capital recycling capability, not a low P/E ratio.
Valuation Scenarios
| Dimension | Bearish | Base | Bullish |
|---|---|---|---|
| Revenue/margin assumptions | 2026-2027 AI order conversion falls short of expectations, normalized EBITDA growth stays below low double digits, pricing remains under pressure | 2026 guidance is achieved, 2027 move-in improves clearly, normalized EBITDA rises steadily | AI orders continue to be fulfilled, domestic chip supply and hub projects land smoothly, 2027 EBITDA steps onto a higher level |
| Cash-flow assumptions | Capex remains high, REIT follow-on is slow, free cash flow remains under pressure | Capex and asset monetization broadly match, on-balance-sheet cash pressure stays manageable | REIT/ABS continues rolling, while DayOne and mature-asset monetization provide extra ammunition |
| Valuation multiple assumptions | Core China about 9x EV/EBITDA, DayOne included at a discount | Core China about 11x EV/EBITDA, DayOne included near management's value reference | Core China about 13x EV/EBITDA, a midpoint between mature infrastructure and a high-growth platform |
| Key catalysts | REIT/ABS lands normally but does not beat expectations | Orders convert to move-in as planned, leverage does not worsen | Major customers keep adding orders, hub projects fill quickly |
| Key risks | Leading customers delay move-in, financing costs rise | Pricing keeps falling and erodes returns | Overexpansion pressures leverage and valuation at the same time |
| Implied return range | Corresponding trading range about 19-22 USD | Corresponding trading range about 32-40 USD | Corresponding trading range about 50-58 USD |
| Permanent loss risk | Trigger: move-in undershoots expectations for multiple consecutive quarters and leverage rises above 6x | Trigger: AI demand materializes but capital recycling fails | Trigger: the market prices in the bullish case too early and then returns to infrastructure valuation |
This is not investment advice. It is a scenario analysis that places the company back into the chain of orders, capex, leverage, and asset securitization. The core meaning of the scenario range is to show the approximate risk-reward position, not to be precise to 1 USD: GDS is no longer a deeply distressed stock, and the current price is roughly in the lower half of the base-case range.
Margin of Safety Check and Key Data Table
At the 2026-06-12 closing price of 32.56 USD, the current price carries a meaningful premium to the bearish scenario range, so the margin of safety is zero. Relative to the base-case range, it is holdable but not comfortable. The two most fragile assumptions are the conversion speed from MoUs to firm orders and then to move-in, and the pace of asset securitization. If both assumptions are haircut by 30%, the base case quickly converges toward the low 20s in USD. For a heavy-asset, highly leveraged infrastructure stock, that is enough to make the point: a good company and a good price are not the same thing here.
| Tracking item | Latest value | Watch range | Warning threshold |
|---|---|---|---|
| Normalized revenue growth | Q1 2026 about +7.9% | High single digits to low double digits | Below +5% for two consecutive quarters |
| Normalized Adjusted EBITDA growth | Q1 2026 about +8.0% | In line with or slightly above revenue | Below revenue growth for two consecutive quarters |
| Single-quarter net new bookings | Q1 2026 about 200MW | Annualized close to management target | Clearly below full-year run rate |
| Pre-commitment rate for projects under construction | Q1 2026 84.4% | Above 70% | Below 60% |
| Utilization of area in service | Q1 2026 77.3% | Gradual improvement | Falls below 75% |
| Operating cash flow / asset construction payments | Q1 2026 0.58x | Gradually moving toward 1 | Stays below 0.6x for a long time |
| Net debt/EBITDA | Management says current level is about 4.7x | 4.5x-5.5x | Rises above 6x |
| REIT/ABS injection progress | Management targets first follow-on in 2Q26 | Quarterly progress | Repeated delays |
| MSR pricing trend | 4Q25 down 2.4% YoY; management expects another 3% to 4% decline by end-2026 | Moderate decline | Decline keeps widening |
The most important indicators are bookings, move-in, capital recycling, and leverage, not the net income number the market most likes to watch. For GDS, quarterly profit has a lot of noise. But if these four indicators deteriorate at the same time, the story will fall from "AI infrastructure growth" back into "heavy-asset financing stock."
Risks, Catalysts, and Cross-Sectional/Longitudinal Summary
Risk Analysis
The first risk that could cause permanent capital loss is order fulfillment risk, with medium probability and high impact. The strongest current upside comes from orders and MoUs. But if leading customers delay move-in because of chips, model roadmaps, internal self-build, or capex adjustments, revenue and EBITDA will not collapse immediately, but the market will reprice the stock immediately because GDS's valuation is already partly based on the assumption that the next four quarters will deliver. The observable indicators are the transmission speed from bookings to move-in and whether the pre-commitment rate for projects under construction can stay high.
The second is leverage and refinancing risk, with medium probability and high impact. At the end of Q1 2026, the company had cash of RMB 14.82 billion, total short-term debt of RMB 9.37 billion, and total long-term debt of RMB 36.53 billion. Current liabilities also included RMB 4.285 billion of current convertible bonds. Management has clearly stated that as expansion proceeds, it will tolerate net debt/EBITDA rising to 5x to 6x. For an infrastructure company, this is not immediately dangerous, but it is already high enough. If external funding becomes expensive or REIT follow-on offerings are delayed, shareholders first absorb valuation compression, without necessarily waiting for a default.
The third is customer concentration risk, with medium probability and high impact. The top five customers represent a high share of committed area, and the largest customer alone accounts for 37.9%. In an upcycle, this concentration is interpreted as "deeply bound to leading customers." In a downcycle, it becomes "any one customer's actions can change full-year expectations." Observable indicators include the renewal, move-in pace, and pricing negotiations of any single large customer, and whether the company can diversify the customer mix in new growth markets.
The fourth is pricing and return erosion risk, with high probability and medium-high impact. Management acknowledged on the earnings call that MSR is falling. It declined 2.4% year on year in 4Q25 and is expected to fall another 3% to 4% by the end of 2026. Although the company also emphasizes that unit development costs are falling and new-project yields can still remain at 10% to 11%, this requires construction costs, land, power, and financing costs to cooperate. If any one side falls out of balance, the data center business can quickly turn from "high-certainty rental asset" into a capital sink with thinning returns.
The fifth is governance and structure risk, with low-to-medium probability and high impact. The VIE is not news, but almost all of GDS's core revenue comes from the VIE and its subsidiaries. This means a change in the regulatory environment does not need to alter operations to change capital-market pricing. Chinese ADR investors are familiar with this, but familiarity does not make it unimportant. It is a long-term anchor attached to the valuation discount.
Catalysts and Tracking Indicators
The strongest positive catalysts over the next 12 months fall into three categories. The first is earnings: if normalized revenue and EBITDA growth in Q2 and Q3 2026 starts rising from high single digits, the market will believe Q1 orders were not a short-term disturbance. The second is capital actions: management mentioned on the earnings call that it aims to complete the first follow-on REIT injection in Q2 2026. If that lands smoothly, it means the loop of "build, mature, securitize, redevelop" is truly turning. The third is orders: if the annual gross bookings target of more than 500MW shows signs of being exceeded around mid-year, GDS's valuation anchor will drift further from "traditional IDC" toward "AI infrastructure platform."
Negative catalysts are equally direct. If two consecutive quarters show normalized growth below expectations, REIT follow-on offerings are delayed, or leverage rises significantly toward 6x without capital recycling progressing at the same time, the market will refocus on heavy assets and debt. For GDS, the risks are visible and almost all will leave traces in financial reports.
Bull Case and Bear Case
The bull case can be summarized in four points. First, China's AI infrastructure demand is indeed improving, orders and MoUs are not empty words, and first-quarter bookings reached a record single-quarter high. Second, after DayOne deconsolidation, overseas assets have changed from a capital-consuming item into monetizable equity value. Third, REIT/ABS has moved from concept to transaction, and GDS has opened the asset recycling channel earlier than most Chinese IDC peers. Fourth, within the global data center valuation system, GDS remains far below mature REITs, while among Chinese peers it already shows better margins and a stronger capital-market toolkit.
The bear case also has four points. First, the company remains heavy-asset and highly leveraged, and the smoother the expansion, the greater the funding need. Second, the current headline net income does not represent core leasing profit, with large noise from investment income and accounting treatment. Third, customer concentration is high and unit pricing is still falling, so growth quality still needs validation. Fourth, the current share price is no longer an extremely pessimistic level, and the margin of safety is not obvious.
Pre-mortem
If this investment loses 50% three years from now, I think the most likely scripts are two.
The first script happens on the demand side. From the second half of 2026 to 2027, conversion from leading customers' MoUs to firm orders and from firm orders to move-in is clearly below market expectations. Domestic chip supply improvement fails to translate into sustained deployment. GDS is forced to sign longer contracts at lower prices in new growth markets, and MSR declines by more than the 3% to 4% currently indicated by management. Because depreciation and interest are highly rigid, core EBITDA growth falls close to zero. The market revalues the company at 8x-9x EV/EBITDA, and the stock can easily fall below 20 USD.
The second script happens on the capital structure side. The company accelerates expansion as planned and net debt/EBITDA climbs to around 6x, but REIT follow-on injections are delayed and external financing costs are higher than expected. Even if the market acknowledges that AI demand exists, it will first question whether shareholders can ultimately earn sufficient returns. Infrastructure companies in this situation usually face sustained multiple compression, not necessarily revenue collapse. Once capital recycling fails, GDS stock will look more like a developer under construction than a mature data center asset.
Final Research Conclusion
From a longitudinal view, GDS has proven three real capabilities. First, it can keep winning and delivering for demanding customers in China's core markets. Second, it can continue organizing capital during an industry trough without being forced to stop construction. Third, it can layer overseas businesses and mature assets so different asset types receive different capital-market pricing. These capabilities are real, so GDS is not as fragile as many short-cycle AI theme stocks.
But from a peer-comparison view, its weaknesses are equally real. GDS still has not escaped the three constraints of a Chinese heavy-asset infrastructure company: high leverage, regulatory discount, and customer concentration. More difficult still, AI demand has enlarged the story again while also raising market expectations for capex and revenue realization speed. Today's GDS therefore sits in a very typical middle state: the company is stronger than in 2022-2023, and the stock is also more expensive than then. The business logic has improved, but the margin of safety has not expanded in parallel.
My conclusion is therefore cautious-neutral. GDS deserves serious tracking and is worth holding for existing positions, but at the 2026-06-12 price, it is more an asset that can be held while waiting for execution than one that should be bought aggressively now. I care most about two hard conditions, not a single-quarter EPS figure: whether AI orders can smoothly convert into move-in over the next four quarters, and whether REIT/ABS can keep providing off-balance-sheet or semi-off-balance-sheet funding for expansion. If both happen together, GDS's valuation center still has room to move upward. If either fails, the market will refocus on debt and depreciation.
【Company Profile Scorecard】
Fundamental quality: Medium
Growth: High
Moat: Medium
Financial resilience: Weak
Management credibility: Medium
Valuation appeal: Medium
Risk level: High
Suitable investor type: Long-term growth / Cyclical / Not suitable for ordinary investors
【Investment Rating】
Rating: Hold
One-sentence investment thesis: AI orders are accelerating and the capital recycling channel has opened, but leverage and the pace of execution still leave the margin of safety too thin.
Three price signals: 【Ideal/Fair Buy Price】19-22 USD Rationale: This corresponds to a low-multiple valuation for core China under the bearish scenario, discounts the DayOne stake, and leaves a margin of safety of about 20%.
Holdable price: 32-40 USD
Clearly overvalued price: 50-58 USD
Current price category: Holdable
Is it worth waiting for a better price: Yes. If the price falls back below 22 USD, the REIT follow-on injection lands, and net debt/EBITDA is not meaningfully above 5.5x, I would raise the buy-point assessment significantly. The opportunity cost of waiting is that if bookings continue to beat expectations in the second half of 2026, the stock may move up first.
Target holding period: 3-5 years
Expected annualized return: Bearish around -14%, base 5%-8%, bullish 18%-23%
Maximum downside risk: If order conversion into move-in loses speed and leverage moves toward 6x, the stock could fall below 20 USD within three years, and in an extreme case could nearly halve.
Signals that trigger reassessment: First, normalized revenue growth below 5% for two consecutive quarters; Second, the pre-commitment rate for projects under construction falls below 60%; Third, net debt/EBITDA exceeds 6x; Fourth, follow-on REIT/ABS injections are repeatedly delayed; Fifth, MSR price declines meaningfully exceed management's current expectations.
【Valuation Range】
current: 32.56 (as of the 2026-06-12 close)
bear (conservative, ideal buy zone): [19, 22]
base (reasonable, acceptable holding zone): [32, 40]
bull (optimistic, above the clear overvaluation line): [50, 58]
Research Uncertainties
First, the claim that "total bookings at quarter-end were about 1.8GW" does not appear as a fully matching written statement in the company press release text I could directly verify. I did not include it in the core valuation model, and used only publicly verifiable company metrics such as 200MW of new orders, more than 500MW of MoUs, and the 500-800MW annual gross bookings target. Second, this report did not separately hand-recalculate Runze's precise A-share EV/EBITDA and full share-count basis after the 2026-06-12 close, so Runze is used only as a narrative and structural comparison, not as a core valuation anchor. Third, the company has not explicitly disclosed the split between maintenance capex and expansion capex, so I used a conservative treatment in the cash-flow look-through section. Fourth, DayOne's equity value of more than USD 2.2 billion is inferred from the latest financing price and is not an observable secondary-market price, so a holding-company discount is naturally appropriate.
Reference Sources
The core basis comes from the company's 2025 20-F, Q1 2026 report, Q4 2025 report, and two earnings-call transcripts. The asset securitization section refers to announcements for the private REIT/ABS and C-REIT IPO. Peer comparison uses the latest public results and market valuation snapshots for VNET, Equinix, and Digital Realty. Industry background supplements use data from CAICT and MIIT as cited by mainstream institutions and media.
Other Securities Mentioned in the Report
VNET.US — A direct Chinese third-party IDC comparable, used to compare AI-driven wholesale expansion speed and valuation.
300442.SHE — Runze Technology, an important A-share valuation benchmark for AI compute infrastructure and the REIT narrative.
EQIX.US — The global neutral data center leader, providing a mature infrastructure valuation and margin anchor.
DLR.US — A global wholesale data center REIT, providing a reference for platformized campuses and capital structure.
This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.
Full report
Sign in to read the full report
Sign up free to unlock the full text, the Baillie growth scorecard, and full-text search.
Log in / Sign up free