Analysis of GDS reveals that AI-driven data center orders are surging ahead of recognized revenue, highlighting a lag between contracted capacity and actual monetization.
The answer points to data center operators. They do not manufacture compute power; instead, they build facilities, supply power, manage operations, and lease racks to cloud providers and internet companies. On August 13, GDS Holdings released its unaudited financial results for the second quarter of 2026. The market responded swiftly and positively, with shares surging as much as 13.3% intraday to touch HK$35.16 the following day.
At first glance, GDS delivered a remarkably strong set of figures. First-half net profit reached approximately RMB 3.49 billion, marking a turnaround to profitability year-over-year. The second quarter alone posted a net profit of RMB 838 million, successfully reversing a RMB 71 million loss from the same period last year, yielding a healthy net margin of 27.1%.
Operationally, the performance was even more striking. New signed capacity totaled 470 megawatts in the first half, surpassing the entire 2025 full-year total in just six months. Management has since doubled the full-year sales target to 1 gigawatt. Amid the AI compute demand surge, GDS, acting as the quintessential “shovel seller,” is seizing unprecedented historical opportunities and undergoing a market re-rating.
Yet the numbers require closer scrutiny. How much of the profit surge stems from core operations? When will the frenzied order book actually convert into realized revenue? As China’s leading third-party data center operator, what is GDS’s true standing in the AI era? Fundamentally, GDS operates as a digital-era “landlord,” leasing completed facilities, stable power, cooling, and network bandwidth by rack or wattage. Clients are typically compute-heavy enterprises: cloud service providers, major internet firms, and financial institutions requiring disaster recovery and core systems. Contracts span five to ten years, ensuring high client stickiness and highly predictable revenue.
This heavy-asset, long-cycle, high-leverage model was established when GDS was founded in 2000. However, the advent of AI has dramatically altered the dynamics of this traditional business. While overall revenue has stabilized, the cash flow trajectory offers the most robust positive signal in the report. The sharp increase in cash flow indicates that previously accumulated contracted orders are finally converting into tangible cash inflows. The logic is straightforward: data center contracts are predominantly “pay-or-play.” Regardless of actual utilization, payments are mandatory. Once clients complete deployment, these payments transform into stable recurring revenue. The Q2 cash flow jump signals that large deals signed in previous years are now entering steady billing cycles—a hallmark of the IDC business where front-loaded investments yield exceptional cash flow certainty and longevity once ramped.
Overall, this revenue growth rate may not seem spectacular for an “AI compute foundation” company, but it reflects steady, rather than aggressive, expansion. After 26 years in operation, GDS has transitioned from rapid scaling to a mature ramp-up phase, consistent with industry norms. This indicates that the core rental business no longer boasts the early-year annual growth rates exceeding 30%. Instead, it relies on a heavier, accumulation-driven model where orders lead and revenue follows gradually. Meanwhile, gross margins faced pressure, declining to 21.5% in Q2 from 23.8% a year earlier. GDS attributed this to rising electricity costs and the front-loading of depreciation and operational expenses as new facilities came online—costs incurred upfront while revenue trails behind.
This dynamic underscores the profound shift AI has brought to IDC cost structures. Data centers have always been energy-intensive, but AI workloads have multiplied power densities several times over. A single rack now consumes as much power as multiple legacy units, pushing electricity further up the operating cost curve. According to the National Energy Administration, national computing center electricity consumption reached 170 terawatt-hours in 2025, accounting for 1.6% of total social consumption—a ratio poised to climb. As GDS expands its operational footprint, utility bills will inevitably grow. Furthermore, government initiatives promoting “computing-power and power synergy” mandate higher green power consumption ratios for new facilities. Since renewable energy carries a premium over coal-fired power, this introduces additional cost pressures.
Beyond cost dynamics, AI infrastructure demand has generated substantial new orders, expanding GDS’s client base beyond traditional internet and cloud giants into emerging sectors. From a workload perspective, roughly half of GDS’s current capacity runs CPU-based tasks and half handles GPU workloads. Management anticipates the GPU share will rise further next year—demand profiles that barely existed before 2023. In H1 2026, GDS secured 471 MW in new signings, eclipsing the 2025 full-year total. Q2 alone contributed 260 MW, prompting management to raise the annual target from 500 MW to 1 GW. Currently, GDS has locked in 600 MW in resource reservation agreements and holds constrained backlogs of 757 MW, bringing total committed capacity above 2 GW. Management projects reserved capacity will exceed 1 GW by year-end 2026.
Notably, approximately half of H1’s new contracts originated from emerging markets such as Ulanqab, Horinger, and Shaoguan, balancing mature and emerging hubs evenly. In June 2026, GDS signed a strategic partnership with the Ulanqab municipal government to invest over RMB 30 billion across five years, developing a gigawatt-scale green power direct-connected data center cluster. Through direct grid connections and green power trading, the project aims to achieve over 80% renewable energy coverage. This strategic push targets scarce large-capacity power quotas and green energy access in western China, aligning with procurement standards increasingly mandated by top-tier cloud providers.
Translating signed contracts into recognized revenue involves a 12-to-24-month pipeline encompassing construction, delivery, client deployment, commissioning, and billing. Consequently, Q2’s RMB 3.088 billion in revenue, representing a modest 6.5% year-over-year increase, primarily reflects projects signed in 2024 or earlier. The explosive H1 order book will drive its true revenue realization peak in the second half of 2027. Management explicitly stated during the earnings call: “Delivery volume in 2027 will more than double that of 2026.” Thus, the current financial report represents the “eve,” with H2 2027 serving as the “main stage.” Today’s market valuation is pricing in expectations for 2027’s revenue explosion, not current profitability.
GDS is effectively rewriting its playbook, transitioning from “selling racks” to “providing AI compute infrastructure.” The traditional IDC logic was straightforward: secure land, power, and energy quotas in core metropolitan rings, construct facilities, and lease them out, profiting from rent and electricity spreads. Location and resource access were the primary moats. AI compute demands have fundamentally altered this paradigm. Clients no longer seek basic server rooms; they require high-power-density clusters capable of running large language models. Rack power ratings have surged from 4–6 kW to over 20 kW, making liquid cooling a standard requirement and green power supply a non-negotiable metric. Facilities failing to meet these specifications are simply unmarketable, regardless of price.
GDS’s geographic strategy mirrors this split: peripheral nodes in the Beijing-Tianjin-Hebei, Yangtze River Delta, and Greater Bay Area regions cater to low-latency inference and real-time computing, while emerging hubs like Ulanqab, Horinger, Shaoguan, and Zhongwei leverage abundant green power and lower land costs to host large-scale AI training workloads. However, capital efficiency remains the central question. It is not merely about funding availability, but whether each unit of capital deployed generates adequate returns. The IDC sector inherently struggles with ROE and ROIC metrics during expansion phases, as depreciation hits the income statement before assets reach full utilization. GDS currently navigates this exact inflection point. Previously funded projects are steadily delivering and ramping, yet their full profit contribution remains unrealized, while new contract wins necessitate continued capital expenditure.
Ultimately, GDS is betting on a singular thesis: AI compute infrastructure will shift from “adequate” to “structurally insufficient.” When that tipping point arrives, operators possessing ready-made high-power facilities, secured green power corridors, and entrenched client relationships will capture pricing power premiums. The industry currently faces a surplus of generic capacity but a critical shortage of AI-ready, high-quality infrastructure. Inefficient supply is being cleared out, allowing tier-one players to monopolize incremental demand. GDS’s industry-leading market share constitutes a formidable long-term moat. Yet moats require constant reinforcement. Liquid cooling architectures evolve rapidly, power grids undergo continuous upgrades, and client PUE requirements grow ever stricter. Sustaining a first-mover advantage hinges on precisely timing technological shifts. Falling behind by even a year or two could quickly erode competitive positioning.