The AI boom continues unabated, with hardware and cloud service providers leading the charge in performance.

Previously, Inspur Information, a frontrunner in China's server market, released its performance forecast for the first half of 2026. The announcement highlighted that the company anticipates its net profit for the first half to range between 2.6 billion and 3.1 billion yuan, marking a year-on-year surge of 226% to 288%. In the same period last year, the net profit stood at 799 million yuan. Excluding non-recurring items, the non-GAAP net profit is projected to be between 2.055 billion and 2.555 billion yuan, up 206% to 280% from 672 million yuan in the same period last year.

Remarkably, Inspur Information's net profit for the first half of this year has already outstripped its entire profit for last year.

A closer look reveals that the company's performance surge is particularly pronounced in the second quarter. Based on the forecast, the net profit attributable to shareholders for the quarter is estimated to range from 1.995 billion to 2.495 billion yuan, representing a substantial sequential increase of 229% to 312% from the 605 million yuan recorded in the first quarter. This marks a new quarterly profit high since the company's listing.

Regarding the drivers behind this significant performance growth, Inspur Information stated that in the first half of 2026, the company capitalized on industry upturns, adhered to a market-oriented strategy, continuously advanced product technological innovation, enhanced customer service capabilities, refined its product line layout, increased product value-added, strengthened supply assurance, and consequently, drove substantial growth in operating performance.

Amidst this record-breaking performance, Inspur Information's stock price also peaked at one point during the year. Data indicates that despite a notable correction since July, the year-to-date increase remains above 15%. As of the latest close, the company's total market capitalization has reached 112.7 billion yuan.

AI Boom Fuels a Billion-Dollar Server Giant

Founded in 1998, Inspur Information went public on the Shenzhen Stock Exchange two years later.

In its early years, the company primarily focused on servers. From government IT infrastructure projects and operator data center expansions to financial data centers and server cluster deployments for internet firms, Inspur Information has long been involved in a singular business model: procuring CPUs from Intel, memory modules from Samsung and SK Hynix, and delivering fully assembled servers to clients.

This business model inherently offers limited profit margins. Inspur's earnings stem not from technological innovation premiums but from economies of scale in procurement and assembly efficiency. Meanwhile, its client base, comprising governments, operators, and internet giants, all wield significant bargaining power, with price comparisons being a common industry practice.

On the supplier side, CPUs are sourced from Intel and GPUs from NVIDIA, creating an oligopoly where downstream companies have minimal bargaining power. For years, Inspur Information has been squeezed between both ends: lacking pricing power upstream and facing strong client bargaining power downstream, continuously compressing its profit margins.

Despite operating in a manufacturing business with limited room for imagination, Inspur Information has persevered for over two decades, solidifying its position as the domestic industry leader.

By 2025, Inspur Information's domestic AI server market share surpassed 50%, while its global server shipments consistently ranked third, trailing only Dell and HPE. In the domestic market, the combined share of Huawei, New H3C, ZTE, and Lenovo still falls short of Inspur's. Simply put, for every two AI servers sold domestically, at least one comes from Inspur Information.

Maintaining such a high market share, Inspur Information relies on three core capabilities honed over the long term.

Firstly, mature supply chain management capabilities. Server production involves dozens of core components, including CPUs, GPUs, memory, hard drives, network cards, and power supplies. These highly standardized products experience sharp price fluctuations driven by semiconductor cycles, while delivery timelines are vulnerable to geopolitical tensions and capacity allocations. Any shortage of critical components can disrupt server delivery schedules. Over more than two decades of operation, Inspur Information has refined its supply chain management into a systematic capability, enabling bulk procurement and inventory allocation of key components at optimal timings and prices.

Financial data reveals that as of the end of 2025, Inspur Information's inventory balance reached 47.852 billion yuan, accounting for 55% of total assets. In other industries, high inventory levels might be seen as operational risks, but in the server industry, sufficient stockpiles are crucial for ensuring stable delivery. When internet giants procure servers, delivery reliability often takes precedence over price—a fact supported by Inspur's hundreds of billions in inventory reserves, which underpin its over 50% market share.

Secondly, a diversified customer base. Government and operator clients prioritize compliance, stability, and long-term service capabilities, while internet giants pursue ultimate cost-effectiveness and flexible delivery capabilities, creating significant demand differences. Among domestic IT hardware manufacturers, few can simultaneously serve both sectors. Inspur Information has decades of channel and project accumulation in government and operator markets while also long serving internet leaders like Alibaba, ByteDance, and Tencent. In 2025, the company's server-related revenue reached 154.6 billion yuan, with clients distributed across industries, partially offsetting risks from cyclical fluctuations in any single sector.

Thirdly, continuous breakthroughs in cutting-edge technologies. In recent years, Inspur Information's liquid cooling and super-node technologies have gradually achieved large-scale deployment.

As AI servers evolve from standalone units to 10,000-card clusters, cooling solutions shift from air to liquid cooling, and delivery models upgrade from standardized standalone products to integrated rack-level solutions, industry entry barriers continue to rise. While numerous contract manufacturers can produce ordinary servers, few domestic suppliers possess end-to-end capabilities in designing, deploying, and delivering entire liquid-cooled AI computing clusters. Since 2021, Inspur Information has ranked first in domestic liquid-cooled server shipments for four consecutive years.

Leveraging these three advantages, Inspur Information fully capitalized on the 2024–2026 AI computing infrastructure development boom.

Training AI large models requires massive computing power. The hardware backbone of 10,000-card training clusters consists of supercomputing systems built with numerous GPUs interconnected via high-speed networks and cooled by liquid cooling—precisely the super-node rack products Inspur currently promotes. Computing demand continues to expand exponentially. According to the "2025 China AI Computing Power Development Assessment Report" co-released by Inspur Information and IDC, China's intelligent computing power reached 1037.3 EFLOPS in 2025, a 43% year-on-year increase, with projections climbing to 1460.3 EFLOPS in 2026, nearly doubling in two years.

Inspur Information has emerged as a core beneficiary of this industrial wave. Upstream computing chips come from NVIDIA, Ascend, and Hygon, while downstream 10,000-card cluster orders flow to internet giants and regional AI computing centers. Among midstream manufacturers capable of handling large-scale, highly complex integrated system deliveries, choices remain limited. In 2025, Inspur Information achieved revenue of 164.782 billion yuan, a 43.25% year-on-year increase, with profits continuing to surge in the first half of 2026. The 2024–2026 period represents the strongest growth cycle in Inspur's two-decade listing history.

The Gift of the Cycle

While performance shines, the market must also calmly consider: How long can this AI upcycle last?

From an industry perspective, AI's long-term development trend is clear, but companies still face multiple challenges.

The primary issue: Inspur Information's core business attributes remain fundamentally unchanged.

Whether traditional 8-card AI servers or the current higher-margin YuanNao SD200 super-node racks, Inspur still plays an integrated manufacturing and assembly delivery role in the supply chain. High-end GPUs, memory modules, and high-speed interconnect solutions all rely on external suppliers, with the company lacking self-developed capabilities for core components. Downstream purchasers—primarily internet giants and regional AI computing centers—exhibit high procurement concentration, giving buyers significant bargaining advantages.

This business model means that even for high-margin liquid-cooled cluster products, profit ceilings are not solely determined by Inspur but depend on the spread between upstream chip procurement costs and downstream procurement budgets. Although the company's profit margins have improved—with Q1 gross margin rising to 6.64% from 3.45% a year earlier—the overall level remains low, placing Inspur in the mid-to-lower tier of profit distribution within the supply chain.

More alarmingly, the sustainability of margin improvements faces challenges from competitors.

Inspur's advantages in liquid cooling and super-node technologies represent more of a first-mover advantage and phased maturity gap rather than absolute barriers built on exclusive patents. Huawei boasts a full-stack self-developed system from Ascend AI chips to rack servers, with ample commercial experience in liquid cooling technologies. ZTE and Unisplendour Corporation are also rapidly advancing their super-node product lines. If competitors continue closing the gap, Inspur's current leadership could quickly diminish.

Beyond intensifying industry competition, the issue of major client dependency cannot be ignored.

In 2025, Inspur Information's largest client—a leading internet enterprise—accounted for 39.37% of sales, while the top five clients combined represented 71.27%, highlighting significant customer concentration risks. Currently, internet giants like Alibaba, ByteDance, and Tencent are all advancing self-developed server initiatives: T-Head continues researching AI chips while deploying self-developed server projects; ByteDance keeps expanding its proprietary computing infrastructure; Tencent has long focused on customized server solutions.

In the long run, domestic large internet companies' procurement models continue evolving—from purchasing branded servers to custom ODM solutions and gradually toward self-developed technologies—constantly impacting traditional server manufacturers' order books.

To date, Inspur Information remains firmly entrenched as the domestic server leader, attesting to its solid competitive strength. With AI industry advancement and continuous performance delivery, the company has room for further growth. However, breaking free from its integrator positioning, addressing peer competition, and navigating the wave of client self-development will pose long-term challenges for the company.