Inspur Information

Inspur Information (浪潮信息), headquartered in Jinan, Shandong, is not a chip designer or a foundry — it is China's dominant AI-server systems integrator, the company that takes GPUs, accelerators, HBM memory, networking, and power/cooling hardware and assembles them into deployable rack-scale infrastructure for Chinese hyperscalers and enterprises. That positioning has made it the single most important physical link between chip supply and deployed AI compute inside China: Inspur held approximately 52.3% of China's AI server market in Q1 2026, a share that has remained stable above 50% for several consecutive years, and it ranks among the top five server makers globally alongside Dell and HPE. The business model is fundamentally an assembly and integration one, not a design one, and the economics reflect that: Inspur's gross margins run in the range of 6–8%, thin even by hardware-integrator standards, because the bulk of the value in an AI server sits in the GPU or accelerator silicon Inspur does not itself design. Full-year revenue for the most recent fiscal year reached approximately RMB 114.8 billion (roughly $16 billion), a figure that reflects enormous volume moving through low-margin assembly rather than high-margin intellectual property. Inspur is publicly listed on the Shenzhen Stock Exchange (000977.SZ). What makes Inspur's supply chain distinctive is that it straddles both sides of China's chip self-sufficiency transition simultaneously. Its server chassis carry export-compliant NVIDIA silicon — chips like the H20-class accelerators that NVIDIA has designed specifically to clear US export-control thresholds for China sale — alongside a growing share of domestic accelerators from Huawei's Ascend line and newer fabless entrants such as Moore Threads and MetaX. This dual-sourcing posture makes Inspur a useful barometer for China's actual AI-hardware self-sufficiency progress: the ratio of NVIDIA-compliant chips to domestic accelerators moving through Inspur's assembly lines shifts with both export-control tightening and the maturing performance of Chinese-designed alternatives. Inspur's export-control exposure is direct and significant. Inspur Group was added to the US Entity List on March 2, 2023, with the Commerce Department citing the company for acquiring or attempting to acquire US-origin items in support of China's military modernization efforts, specifically tied to supercomputer development for Chinese government and military end use — a designation that traces back to a 2020 Department of Defense action that had already halted direct Intel chip sales to Inspur. The restriction deepened in March 2025, when six Inspur Group subsidiaries — five based in China and one in Taiwan — were added to the Entity List for their role in Inspur's supercomputer projects. In practice, the listing means Inspur (and its listed subsidiaries) require a license before receiving many US-origin components, forcing greater reliance on both export-compliant NVIDIA products purpose-built to clear licensing thresholds and on the domestic accelerator ecosystem it is increasingly integrating. Inspur's customer base is essentially the roster of Chinese cloud providers and AI consumers documented elsewhere in this dataset: Alibaba Cloud, Tencent Cloud, and Baidu all rely on Inspur-built infrastructure to deploy the compute that trains and serves models across China's AI industry. Its power and cooling supply chain draws on both domestic vendors and global players like Vertiv, reflecting the same universal-supplier pattern seen among other server integrators. Inspur's trajectory illustrates a structural reality of China's AI hardware buildout: even a company under direct US Entity List restriction remains commercially essential to the domestic AI ecosystem, because someone has to physically build the servers regardless of whose silicon sits inside them. Its near-term outlook depends less on whether it can escape Entity List status — unlikely in the current environment — and more on how quickly the domestic accelerator share of its server shipments rises as Huawei Ascend, Moore Threads, MetaX, and other Chinese chip designers mature, gradually reducing Inspur's dependence on export-compliant NVIDIA silicon altogether.