Inspur Information Deep-Dive Research

Vertical Development and Financial Review

Inspur Information began as a fairly typical product of China's state-backed IT industrialization. Its listing prospectus shows that the company was established on 1998-10-28, with sponsors including Inspur Electronic Information Industry Group, Yantai Dongfang Electronics Information Industry Group, and Beijing Suantong Technology. It listed on the Shenzhen Stock Exchange on 2000-06-08 with 58.50 million tradable shares, stock code 0977, and an issue price of RMB 7.71. From the start, it grew inside China's institutional push for domestic information-industry capability, rather than as a start-up that began with a single technological invention and was later discovered by capital markets.

That origin shaped its later business choices. On its official website today, the company positions itself as a provider of IT infrastructure products, solutions, and services to more than 100 countries and regions, with core businesses in servers, storage, switching, liquid cooling, and diversified compute solutions. Its true underlying capability, however, has always been organizing upstream chips, boards, interconnects, power supplies, thermal systems, and full-system manufacturing into infrastructure that can be delivered at scale. In other words, Inspur earns money from complex supply-chain scheduling and customer delivery capability, not from a proprietary chip.

Financially, the past few years can be divided into three broad phases. The first was the traditional server expansion period from 2020 to 2022: revenue rose from RMB 63.038 billion in 2020 to RMB 69.525 billion in 2022, and net profit attributable to shareholders increased from RMB 1.466 billion to RMB 2.080 billion. Operating cash flow was extremely unstable, with a large outflow of RMB -8.290 billion in 2021 before returning to a positive RMB 1.800 billion in 2022. This shows that even before the AI boom fully arrived, this business was already highly sensitive to stocking, collection, and payable timing, rather than a business that looked stable just from the income statement.

The second phase was the AI server re-rating period of 2023-2024. Revenue fell back to RMB 65.867 billion in 2023, then jumped to RMB 114.767 billion in 2024 on AI server volume growth, up 74.24% year over year. In 2024, server and component revenue was RMB 114.002 billion, almost all of the main business; overseas revenue grew 256.98% year over year, and revenue from the industry-customer sales model grew 95% year over year. On the surface, the company captured one of the strongest phases of global AI server expansion. The other side of the statements was equally clear: gross margin in the electronics industry was only 6.75% in 2024, gross margin for servers and components was 6.76%, net operating cash flow was only RMB 98 million, down 81.18% from the previous year, and the annual report directly attributed this mainly to increased operating stocking. Scale surged, but cash did not follow.

The third phase since 2025 is “domestic compute substitution plus Entity List repricing.” Start with policy and geopolitical constraints. The U.S. Commerce Department placed Inspur Group on the Entity List in 2023, and in 2025 it also formally placed Inspur Electronic Information Industry Co., Ltd., the English legal-entity name of Inspur Information, on the Entity List with footnote 4. This means relevant foreign-made products can also fall under licensing constraints if they trigger the Foreign Direct Product Rule. At the same time, Reuters successively reported that H3C faced H20 supply shortages in 2025 and that NVIDIA's share of China's AI chip market had been pushed to nearly zero in 2026. For Inspur Information, this is a concrete variable that changes product mix and order structure, not an abstract backdrop.

The company's response to this phase is clear: keep compatibility and product-development capability in the international chip ecosystem while deepening domestic compute full-system platforms. In 2025, the company released YuanNao SD200, emphasizing a single machine with 64-way domestic AI chip high-speed unified interconnect and support for a 4-trillion-parameter single model. In the same year it released YuanNao HC1000, with a focus on inference cost falling below RMB 1 per million tokens for the first time. The company's website also emphasizes that the YuanNao AI system already supports more than 30 AI chips and, beyond international platforms such as Intel/NVIDIA, continues to invest in a diversified heterogeneous route. In simple terms, Inspur wants to move from “making servers” to “turning anyone's chips into delivered full systems as quickly as possible.”

The 2025 financial report shows both the achievements and shortcomings of this route. Full-year revenue was RMB 164.782 billion, up 43.25% year over year, while net profit attributable to shareholders was RMB 2.413 billion, up only 5.20%. Server product revenue was RMB 154.605 billion, or 93.82% of total revenue, but server product gross margin fell from 6.76% in 2024 to 4.52%, and consolidated gross margin was about 4.88%. Media reviews outside the annual report also noted that inventory reached RMB 46.508 billion at the end of 2025, with inventory impairment provisions exceeding RMB 1.3 billion. The revenue ceiling was raised, but the margin floor moved lower.

What really explains capital-market sentiment is cash flow. In 2025, net operating cash flow was RMB 5.453 billion, up 7,183.71% year over year, so the market briefly began to believe that the old problem of “low margins but cash can return” might be easing. Yet in Q1 2026, revenue declined 24.30% year over year, net profit attributable to shareholders rose 30.74%, and operating cash flow turned negative again at RMB -7.772 billion. Accounts receivable increased 32.55% from the beginning of the period, and short-term borrowings increased 522.47%. This suggests that the 2025 cash-flow repair was more like a confluence of delivery and collection timing than a completed transformation of the business model.

In stock-price narrative terms, the market in 2025 had already traded the “domestic compute theme” to a very high level. The company touched a historical high of RMB 80.80 on 2025-10-09. By 2026-06-12, the share price had returned to RMB 57.86, a drawdown of about 28%. This correction does not mean the market rejects domestic substitution. It means investors are attaching a more accurate price tag to the company: it may still be an important industrial participant, but industrial position and the quality of shareholder returns are not the same thing.

Key Data Table

The 2020-2022 data in the table come from the 2022 annual report summary, the 2023-2024 data come from the 2024 annual report, and the 2025 data come from the 2025 annual report and annual-report commentary. All figures are organized in RMB 100 million.

The most important row in this table is operating cash flow / net profit, not revenue. On a consolidated basis over 2021-2025, the ratio of cumulative operating cash flow to cumulative net profit attributable to shareholders was roughly close to zero. In the last cycle, profit almost never turned into cash on a stable basis. Excluding the severe stocking year of 2021, the cumulative cash flow / net profit ratio over 2022-2025 was about 0.9x. This means the issue is not that the company can never earn cash, but that cash conversion is highly back-ended, and in upcycles it often consumes working capital first. For a low-margin full-system vendor, this matters more than the RMB 2 billion-plus net profit shown on the income statement.

Business Model, Industry, and Horizontal Competition

Inspur Information's real business machine is simple: it sells full server systems and related infrastructure products, and earns money through scale, delivery, and supply-chain scheduling. In 2025, server product revenue was RMB 154.605 billion, accounting for 93.82% of total revenue; storage, switching, and other product revenue was RMB 9.771 billion, accounting for 5.93%. In other words, this company relies on massive full-system shipments, not high-margin software subscriptions or chip IP. Server product gross margin was 4.52% in 2025, and consolidated gross margin was about 4.88%, which makes the boundary of this business model very clear.

The cost structure of such a business has almost no mystery. In Q1 2026, the company generated operating revenue of RMB 35.470 billion and operating costs of RMB 33.115 billion. Based on this, I estimate consolidated gross margin at about 6.64% and net margin at about 1.71%. Most of its costs come from chips, boards, memory, interconnects, chassis, power supplies, thermal systems, and manufacturing delivery. R&D is important, but it has never been the decisive drag on margins. What really determines profit swings are BOM cost, customer mix, price competition, and working capital. Once revenue falls, profit thins quickly, because a low-margin business has almost no buffer.

So Inspur Information's moat should not be sought in “pricing power,” but in “delivery rights.” The first moat is scale and supply-chain organization. The company's website says its products and services cover more than 100 countries and regions and that it has a global R&D and manufacturing network. This means it can coordinate chips, thermal systems, interconnects, and customer delivery at large scale. The second moat is heterogeneous ecosystem adaptation. The company says YuanNao AI supports more than 30 AI chips, making it easier to absorb orders during the phase when international solutions are restricted and domestic solutions substitute rapidly. The third moat is experience in liquid cooling and high-density clusters. Citing IDC, the company says its sales and shipment shares in China's liquid-cooled server market in 2024 were 35.5% and 34.2%, respectively, both ranking first. The fourth moat is key-customer trust. In 2025, the market paid close attention to the company as a candidate winner in Industrial and Commercial Bank of China's domestic-chip server procurement project.

But none of these is a moat that “no one can take away.” The real issue is that most of these advantages show up in winning and delivering orders, not raising prices. The strongest counterevidence comes from the 2025 financial report: annual revenue rose sharply by 43.25%, yet server gross margin fell from 6.76% to 4.52%. If a company cannot turn scale into meaningful gross-margin expansion even in the hottest year, its moat looks more like execution capability than pricing power. For long-term investors, these are two very different things.

The industry structure is therefore clear. In the AI infrastructure value chain, the main profit pools are in upstream chips, networking interconnects, and some system software, not in midstream full-system assembly. This is not an abstract judgment. Inspur Information's server gross margin in 2025 was only 4.52%. Dell recorded USD 113.5 billion of revenue and USD 11.2 billion of operating cash flow in FY2026 ended 2026-01-30, and its single-quarter AI-Optimized Servers revenue had already reached USD 9.0 billion. SMCI generated USD 10.2 billion of revenue in Q3 FY2026, but to meet USD 39.0 billion of orders it announced in June 2026 that it would raise USD 7.0 billion of equity and equity-like capital. Full-system vendors can have orders and scale without easy free cash flow.

This industry also combines four cycles. The first is the capex cycle: once cloud vendors, financial institutions, telecom operators, and government and enterprise customers build intelligent-computing centers intensively, orders explode; when construction slows, full-system vendors are hit first. The second is the policy cycle, where Xinchuang, AI+, and centralized procurement by financial institutions and telecom operators all change volume. The third is the technology iteration cycle, with fast evolution in training and inference architecture, liquid cooling, and high-speed interconnects. The fourth is the geopolitical cycle, where the Entity List and export controls directly change supply-chain availability. The company sits at the intersection of all four cycles. The U.S. placement of Inspur Information on the Entity List, H20 supply fluctuations, and the accelerated shift of the Chinese market toward domestic chips are long-term structural constraints, not one-off shocks.

Looking horizontally at competitors, Inspur Information occupies a very specific niche: it is neither the highest-margin player nor the most expensive by valuation, but it is one of the stocks most easily treated by the market as a “total-volume proxy for domestic AI servers.” The table below starts with the numbers, then looks at the group profile.

† For A-share companies, FY2025 is used; for Dell, FY2026 is used; for SMCI, the midpoint of FY2026 revenue guidance is used, and Q3 FY2026 year-over-year growth reflects current momentum. ‡ Market capitalization and share prices in the table are as of the 2026-06-12 close; Dell and SMCI market capitalization and revenue have been converted using the 2026-06-12 CFETS USD/CNY central parity rate of 6.8109. P/S in the table is self-calculated from market capitalization and latest annual or guided revenue.

Sugon has evolved into something closer to a “high-performance computing and domestic compute solution platform.” Its 2025 revenue was only RMB 14.964 billion, far smaller than Inspur Information's, but the market gives it about 54x PE and more than 8x P/S, a significantly higher valuation. Investors are buying more than server shipments; they are buying a solution premium linked to research, government, liquid cooling, high-performance computing, and the Hygon ecosystem. Customers often choose Sugon for overall solution fit in specific scenarios, not because it is “cheaper.”

Unisplendour's H3C has evolved into an “enterprise ICT platform provider.” Its 2025 revenue was RMB 96.748 billion, with net profit of RMB 1.686 billion. It has large scale and deep channels. Customers choose it more because networking, servers, storage, security, cloud, and industry solutions are offered as an integrated package. Compared with Inspur Information, H3C's story is about enterprise channels and integration capability capturing a more stable profit pool, rather than betting on a single AI server volume cycle. The market therefore gives it a higher P/S than Inspur Information, while not treating it as a highly elastic pure AI theme stock.

Dell is a completely different reference point. Its AI server business is growing quickly, but what customers really pay for is its global supply chain, service capability, integration of storage and client solutions, and stronger cash generation. In FY2026, Dell recorded full-year revenue of USD 113.5 billion, operating cash flow of USD 11.2 billion, ISG revenue of USD 60.8 billion, and Q4 AI-Optimized Servers revenue of USD 9.0 billion. Customers choose Dell because it behaves like a mature enterprise infrastructure machine, not only because it can assemble servers. Inspur Information and Dell are similar in “selling servers,” but very different in “returning profit and cash to shareholders.”

SMCI is the global mirror that Inspur Information should watch most carefully. It represents the high-elasticity version of AI servers: on the way up, revenue surges, theme concentration is high, and the market pays attention; on the way down, it must raise large amounts of capital to secure components and meet orders, exposing shareholder dilution and free-cash-flow pressure at the same time. In June 2026, SMCI announced about USD 7.0 billion of equity and equity-like financing to support component procurement for advanced AI servers; before that, the company said it had secured more than USD 39.0 billion of orders. This story reminds investors that the central contradiction in the AI server track is often that profit and cash are not thick enough, not that orders are insufficient. If Inspur Information also reaches a future state of “many orders but tighter funding,” the stock price will not be more forgiving than the financial statements.

Current Fundamentals and Valuation Analysis

Looking at the most recent four quarters, Inspur Information's operating condition has two distinct sides. The first is that demand remains resilient. In H1 2025, revenue was RMB 80.192 billion, up 90.05% year over year, and server product revenue was RMB 75.286 billion, up 99.5% year over year. Full-year revenue then reached RMB 164.782 billion. The second is that margins and cash flow did not improve in tandem. H1 2025 consolidated gross margin fell to 4.51%, server gross margin was 4.32%, inventory rose to RMB 59.522 billion, and short-term borrowings also increased sharply; full-year server gross margin was still only 4.52%. This shows that the high growth in 2025 was mainly volume growth, not quality transformation.

The Q1 2026 data pushed the conflict between “volume” and “quality” one step further. Revenue fell 24.30% year over year, but net profit attributable to shareholders rose 30.74%. Based on the income statement, I estimate consolidated gross margin recovered to about 6.64%, and net margin to about 1.71%. Looking only at the income statement, the market could easily reach an optimistic interpretation that “a gross-margin inflection point has arrived.” The cash-flow statement gives a different answer: net operating cash flow was RMB -7.772 billion, accounts receivable increased 32.55% from the beginning of the period, and short-term borrowings increased 522.47%. This means profit improvement has not yet flowed through to cash, and the company's fundamentals are now in a phase that is easy to read optimistically and just as easy for cash flow to disprove.

The market is currently trading two main themes. The first is domestic AI compute substitution. Reuters reporting indicates that supply of international GPUs such as H20 has been restricted and NVIDIA's share of China's AI chip market has been rapidly squeezed. This would push more orders from domestic government and enterprise, financial, and telecom customers toward local solutions, raising the role of full-system vendors. The second is whether valuation re-rating for low-margin full systems can hold. The market is willing to pay attention to Inspur Information because it looks like a “total-volume ticket for domestic AI servers.” Yet the market has been reluctant to assign a higher valuation because its financial statements keep reminding investors that this remains a working-capital-heavy business with weak pricing power.

This also explains why the bull-bear divide is so concentrated. The strongest bullish evidence is that the company has already built out both international and domestic ecosystems at the product level, not the 2025 high revenue growth by itself. The company released SD200 and HC1000, supports more than 30 AI chips, and became a candidate winner in ICBC's domestic-chip server project. If domestic compute procurement by financial institutions, telecom operators, government, and enterprise customers expands further, Inspur Information is one of the easiest vendors to route orders to. The strongest bearish evidence is that server gross margin fell to 4.52% in 2025 and Q1 2026 cash flow turned sharply negative again, showing that scale expansion still needs inventory, receivables, and financing support. Industrial position has not automatically turned into shareholder returns.

Sell-side views are also splitting toward these two ends. Some market participants still regard Inspur as a core beneficiary of domestic AI infrastructure, but in May 2026 media reports said Goldman Sachs downgraded the company's rating and target price. The core logic was that higher penetration of domestic AI chips would compress full-system ASP and profit space. This divide is the key to whether valuation can continue rising, not sentiment noise. If you believe the company will move from “selling more” to “earning thicker margins,” the current valuation remains debatable. If you believe it is merely converting the gap left by international chips into higher-volume but cheaper domestic-solution delivery, the current share price is already not cheap.

From a historical valuation perspective, the current price is awkward. Based on the 2026-06-12 close, the company's PE(TTM) was about 33.25x, P/B about 3.81x, and P/S about 0.52x. Although the share price is about 28% below the RMB 80.80 high reached on 2025-10-09, this looks more like a move from mania back to a somewhat expensive mid-to-high range, not a return to obvious undervaluation. For a company with only 4.52% server gross margin in 2025 and operating cash flow turning negative again in Q1 2026, a TTM PE around 33x still embeds a substantial premium for being a “domestic AI infrastructure gateway.”

After looking through cash flow, this conclusion becomes sharper. Over 2021-2025, the company's aggregate operating cash flow to net profit ratio was roughly -0.04x, almost equivalent to “profit over the past five years did not convert into cash on a stable basis.” Even looking only at 2022-2025, the ratio was only about 0.92x, with huge volatility. More importantly, the company's capital expenditure is not heavy: cash paid to acquire or construct fixed assets, intangible assets, and other long-term assets was about RMB 307 million in 2023 and RMB 248 million in 2024, and RMB 28 million in Q1 2026. This means the free-cash-flow problem mainly comes from violent swings in inventory, receivables, and payables, not capex. Based on this, I roughly estimate 2025 maintenance capex at about RMB 200-300 million, implying 2025 owner earnings of about RMB 5.2 billion and an owner-earnings yield of about 6% on the current market capitalization. But that is a single-year high point. On a cumulative 2021-2025 basis, owner earnings are close to zero or slightly negative. Reported PE tells you it is not extremely expensive; the cash-flow lens reminds you that the cash convertibility of this business is far more fragile than PE suggests.

For the valuation scenarios below, I use P/S as the main anchor and PE only as a cross-check. The reason is simple: in an industry where net margin is often only 1%-2%, point-estimated PE can be easily distorted by one quarter of working-capital change and foreign-exchange gains.

This is not investment advice, only a decomposition of the linkage among revenue, margin, cash flow, and valuation multiples under the research framework. The endpoints of each range above are intended to provide a consistent basis for the price signals below, not to predict that the market must reach any specific scenario.

Looking at margin of safety separately, the conclusion is poor. The current price of RMB 57.86 is a clear premium to the bearish scenario of RMB 36-42, giving no margin of safety. In the base scenario, the most fragile assumption is net-margin improvement. If the net-margin assumption in the base scenario is cut by 30%, fair value would shrink from RMB 48-58 to roughly RMB 40-49. If earnings show zero growth over the next three years and valuation does not expand, the earnings yield implied by the current TTM PE is only about 2.8%-3.0%, slightly above China's 10-year government bond yield of 1.74% on 2026-06-12, but far below the return most equity investors should require for a volatile hardware stock. My conclusion: this is a company worth researching, tracking, and even continuing to observe for investors who already hold it, but the current buying price does not provide enough margin of safety.

Risk Catalysts and Tracking Indicators

The biggest business risk is price competition and product-mix downshift. I assign a medium-to-high probability and a high impact. The evidence is direct: server product gross margin had already fallen to 4.52% in 2025, showing that even in the hottest year the company retained little profit. If international restrictions continue to reduce the share of high-end international solutions while domestic customers favor a more cost-performance-oriented route, Inspur Information may keep winning orders without capturing higher per-unit gross profit. The signals to watch most closely are whether consolidated gross margin falls back below 5% for two consecutive quarters and whether server product gross margin can return above 5.5%. If the former happens, the stock narrative will switch from “domestic substitution volume growth” back to “more volume without more profit.”

The second risk is loss of control over working capital. Probability is high, and impact is also high. Servers are a classic “stock first, deliver later, collect last” business, and Inspur Information has shown this clearly over the past few years: inventory was RMB 46.508 billion at the end of 2025, accounts receivable rose another 32.55% from the beginning of the period in Q1 2026, operating cash flow had a net outflow of RMB 7.772 billion in the same quarter, and short-term borrowings rose 522.47% from the beginning of the period. The danger of this kind of risk is that once inventory turnover slows and customer payments slow, the company must use more short-term debt to fill the operating gap, forcing shareholders to bear a liquidity discount. The main issue is not whether profit falls by a few hundred million in one quarter. The most important signals are whether operating cash flow / net profit stays below 0.5x for two consecutive quarters and whether receivables and inventory growth remain clearly faster than revenue growth.

The third risk is further escalation of Entity List and export controls. Probability is medium, impact is high. Inspur Information's English legal-entity name has already been placed on the U.S. Entity List with footnote 4, making it more difficult to obtain U.S. items subject to the EAR and some foreign-produced items. Even as the company advances domestic substitution, international platforms will not disappear completely in the short term. If access to international high-end chips, EDA, accelerator cards, and certain supporting components is tightened further, the company's product mix, delivery cadence, and customer expectations will all be affected. The signals to watch most closely are BIS list changes, changes in NVIDIA China restriction rules, and the company's public statements on international-platform delivery.

The fourth risk is that competitors take the profit pool in the domestic ecosystem, rather than taking all the revenue. Probability is medium, and impact is medium to high. Sugon looks more like a “high-performance computing and liquid-cooling solution provider,” while Unisplendour/H3C looks more like an “enterprise channel and integration platform.” Their competition with Inspur Information does not always show up in shipment volume. It more often shows up in who can win higher-value integrated projects from financial, telecom, government, and enterprise customers. If future domestic large orders are defined more as “compute platform solutions” than as “server supply,” Inspur Information's scale advantage will remain, but margins may be captured by others. The signals to watch most closely are changes in the company's winning role in large financial, telecom, and government projects, as well as public progress in liquid cooling, high-density clusters, and platform software capability.

The fifth risk is valuation risk itself. Probability is medium to high, and impact is high. Although the current share price has pulled back from the high, it still corresponds to 33.25x TTM PE and about 0.52x P/S. If market style shifts from “domestic AI infrastructure” back to “cash flow and shareholder returns,” the valuation ceiling for low-margin hardware stocks of this type will contract quickly. The signals to watch most closely are whether P/S can hold the 0.4-0.5x range and how tolerant the market is after earnings toward combinations such as “profit improves but cash flow worsens.”

Positive catalysts are concentrated in three categories. The first is financial-report catalysts: if consolidated gross margin can stabilize above 6% and operating cash flow turns positive at the same time, the market will be more willing to believe that the 2025 scale expansion is turning into profit quality. The second is order catalysts: domestic compute centralized procurement by financial, telecom, and large government and enterprise customers continues to land, especially landmark orders similar to the ICBC project. The third is product-mix catalysts: domestic training/inference products such as SD200 and HC1000 enter real scaled delivery, rather than remaining launch-event narratives.

The negative catalysts are equally clear. The most dangerous combination is “slowing revenue + falling gross margin + continued operating-cash-flow deterioration.” If this is accompanied by tighter BIS rules, further international supply-chain restrictions, or a wait-and-see phase among domestic customers, the story will quickly shift from “domestic substitution” to “cycle digestion.” At that point, valuation will be hit first, and the financial statements will come later.

Tracking Dashboard

The latest values in the table come from the company's Q1 2026 report, 2025 annual report, and 2026-06-12 market data. The thresholds are tracking disciplines set in this article based on historical volatility.

There are really only three things in this table worth tracking over the long term. First, can gross margin hold up? Once gross margin fails, all valuation imagination around being an “AI compute gateway” becomes hard to sustain. Second, can cash flow keep up? Inspur Information's free-cash-flow problem almost never comes from capex; it comes from working capital. Third, do high-value domestic orders continue? This determines whether the company remains a large-scale full-system vendor or has a chance to capture a thicker profit layer. The main tracking channels are periodic reports, investor-relations activity records, major tender announcements, and public information from BIS/CFETS and similar sources.

Cross-Sectional and Longitudinal Summary

Viewed longitudinally, the capability Inspur Information has proved over time is the ability to rapidly industrialize, engineer, and scale complex compute infrastructure, not technological monopoly. It has moved from a state-backed hardware enterprise in 1998 to today's leader in servers and diversified compute full systems through supply-chain organization, delivery systems, customer resources, and scenario adaptation, not a single high-margin product. This capability matters greatly in China's AI infrastructure buildout cycle, because real large orders are often won by whoever can quickly combine imperfect chips, networks, liquid cooling, cabinets, and software stacks into usable clusters, rather than by “who has the best chip.” Inspur Information has already delivered evidence on that front.

The other longitudinal side is equally clear: much of its past success came from era dividends and industrial cycles, not from an inherently elegant business model. During the cloud-computing buildout, Xinchuang push, AI server ramp, and domestic compute substitution phase, the company could quickly scale revenue. Yet it has also repeatedly shown that scale does not automatically produce high-quality profit. There was a large operating-cash-flow outflow in 2021; only RMB 98 million of operating cash flow on RMB 114.767 billion of revenue in 2024; record-high revenue in 2025 but server gross margin falling to 4.52%; and Q1 2026 profit improvement accompanied by a large negative cash-flow swing. The company is good at capturing demand cycles. It has not yet proved it can lock those cycles into long-term ROIC.

Horizontally, Inspur Information's real advantage versus peers is that it looks more like a “total order intake gateway.” Sugon leans more toward high-performance computing and solutions, Unisplendour/H3C toward enterprise channels and integration, Dell toward global services and cash generation, and SMCI represents AI server high elasticity in a more extreme form. Inspur Information sits among these roles. Its strongest areas are scale, delivery, customer coverage, and diversified compute adaptation. Its weakest areas are pricing power and financial stability. This weakness is a structural shortcoming of full-system infrastructure businesses, not a short-term management error. As long as the profit pool remains in upstream chips and interconnects, Inspur Information will be hard to value as a true “hard-tech high-margin platform.”

This is precisely where the market is most likely to misjudge it. Many investors extrapolate directly from “the major trend of domestic compute substitution” to “Inspur Information's margin must improve.” I do not. The trend is real, and orders may be real, but margin improvement requires two more demanding conditions at the same time: competition inside the domestic ecosystem must not push prices down first, and the company must keep inventory, receivables, and payment terms under control. If either condition fails, revenue growth remains only scale growth. The current valuation already rewards it for standing on the right track, but there is still insufficient evidence that it can turn track position into sustained shareholder returns.

The key variables differ over the next 1 year, 3 years, and 5 years. Over the next 1 year, the most important question is whether gross margin and operating cash flow can improve together. If only the income statement improves while cash flow does not, the stock will struggle to sustain further valuation expansion. Over the next 3 years, the key is whether order share and customer stickiness in the domestic chip ecosystem can settle into a stronger platform position, rather than simply following tenders again and again. Over the next 5 years, the key is whether the company can truly move from “leading server full-system vendor” to “diversified compute infrastructure platform provider,” meaning it can combine product mix, liquid cooling, interconnects, and software-like capabilities into a thicker profit layer. If it cannot, it will remain an important but less compelling infrastructure contractor for a long time. If it can, then a higher valuation center becomes discussable.

Better investment timing would appear under two conditions. The first is that price comes down first, falling into the bearish scenario's ideal buy zone, roughly RMB 36-42. The second is that the price is not cheap, but two to three consecutive quarters prove that gross margin, cash flow, and inventory turnover improve together. At that point, the purchase would no longer be only a theme, but quality improvement. Conversely, if the next two quarters continue to show a combination of “decent profit, poor cash,” or if Entity List and international supply-chain constraints escalate further, the original research framework should be overturned quickly, because that would mean the “quality transformation” the market has been waiting for has not happened.

Bull Case

• Domestic compute substitution is not just a concept. The company has already launched SD200 and HC1000 and turned diversified heterogeneous adaptation into a real product line.

• Domestic server procurement by large customers such as financial institutions is happening, and the company's role in ICBC's domestic-chip server project reinforces its order-absorption capability.

• The company ranks first in both sales and shipment share in China's liquid-cooled server market, showing that it is not merely a follower in high-density compute infrastructure.

• Operating cash flow briefly recovered to RMB 5.453 billion in 2025, showing that this business is not incapable of collection forever; its conversion rhythm is simply poor.

Bear Case

• Server product gross margin fell to 4.52% in 2025, proving that the company did not obtain enough pricing power even in the strongest year.

• Q1 2026 net operating cash outflow was RMB 7.772 billion, accounts receivable rose 32.55% from the beginning of the period, and short-term borrowings rose 522.47%, showing cash quality deteriorated again.

• The U.S. Entity List now directly covers Inspur Electronic Information Industry Co., Ltd., turning international supply-chain constraints from background noise into an operating variable.

• The current valuation still stands at 33.25x TTM PE and about 0.52x P/S, which is not cheap for a low-margin full-system vendor.

Pre-mortem

The first loss scenario unfolds over the next 12-18 months. Suppose that starting in H2 2026, domestic large customers enter a digestion phase for training-cluster procurement, and Inspur Information cuts prices to preserve share. Consolidated gross margin falls back below 5%, server gross margin stays at 4%-4.5%, and receivables and inventory do not fall in sync. By then, the market would realize that 2025 was a revenue peak, not a quality inflection point. P/S could be compressed from 0.52x to around 0.35x, corresponding to a share price in the RMB 36-42 range, down roughly 30% from the current level.

The second scenario is worse and plays out over 2-3 years. Suppose the Entity List and related export controls continue to escalate, high-end international platforms become further restricted, and in the domestic chip ecosystem more of the profit layer at the solution level is captured by other full-system vendors or strong integrators. Inspur Information can only deliver lower-ASP general-purpose full systems. At that point, revenue may not decline sharply, but margins could remain around 1% for a long time, cash flow would keep swinging sharply, and capital markets would stop paying a premium for the “domestic AI core gateway” narrative. Valuation could move toward a more typical hardware outsourcing stock. If P/S is compressed toward 0.3x, the share price could fall toward the low RMB 30s, resulting in a near-halving outcome.

My final judgment on Inspur Information is clear. It is an important industrial company, and it is not an easy stock to treat as a comfortable long-term holding. Its era position in domestic compute substitution is real, and its products, customers, and delivery capability are not empty shells. But its financial statements repeatedly remind investors that this business has thin profit and that cash flow can be swallowed easily by inventory and receivables. The current share price is not absurdly overvalued, but it is also not cheap enough to cover these structural risks. For existing holders, it looks more like a stock waiting for evidence. For investors not yet in the stock, the market has already paid one round for being on the “right track”; what matters next is waiting for “quality improvement” or “price decline.”

If the company can prove three things over several consecutive quarters, I would be more willing to raise my view: first, consolidated gross margin stabilizes above 6% rather than rebounding for one quarter; second, operating cash flow keeps up with net profit, at least recovering to above 0.8x; third, large domestic-customer orders not only scale, but also avoid crushing ASP and profit. Without these three things, Inspur Information is better treated as an “important but high-volatility infrastructure order absorber,” not a high-quality growth stock.

【Company Profile Scorecard】

• Fundamental quality: Medium

• Growth: Medium

• Moat: Medium

• Financial resilience: Medium

• Management credibility: Medium

• Valuation attractiveness: Low

• Risk level: High

• Suitable investor type: Cyclical

【Investment Rating】

• Rating: Hold

• One-line investment thesis: Domestic compute substitution can amplify revenue, but low margins and cash-flow volatility limit valuation upside.

• Three price signals: Ideal buy price: RMB 36-42

• Holdable price: RMB 48-58

• Clearly overvalued price: RMB 64-76

• Current price classification: Holdable

• Whether it is worth waiting for a better price: Yes. Better triggers would be the share price returning below RMB 42, or, without a pullback, two consecutive quarters validating “gross-margin improvement + operating cash flow turning positive + receivables and inventory converging.” The opportunity cost of waiting is that a single large domestic compute order could drive a thematic rally that investors miss.

• Target holding period: 1-3 years

• Expected annualized return: bearish -27% to -38%; base -17% to 0%; bullish +11% to +31%

• Maximum loss risk: about 35%-50%. Triggers would be gross margin falling below 5% again, operating cash flow remaining negative for consecutive quarters, Entity List constraints escalating, and the domestic ecosystem profit layer being taken by competitors.

• Signals that trigger reassessment: Consolidated gross margin below 5.5% for two consecutive quarters.

• Operating cash flow / net profit below 0.5x for two consecutive quarters, or operating cash flow remaining negative.

• Accounts receivable and inventory growth significantly faster than revenue for two consecutive quarters.

• BIS / Entity List rules continue tightening and affect international-platform delivery.

• Landmark domestic compute orders from financial institutions, telecom operators, and similar customers give more high-value links to Sugon, H3C, and other competitors.

【Ideal/Fair Buy Price】36-42 CNY Basis: this corresponds to 0.35x-0.40x P/S under the bearish scenario and a margin-of-safety requirement of at least 20%. In this range, the market would be closer to pricing the reality of “low margin, strong cyclicality, weak cash conversion,” rather than paying for “track gateway” status.

【Valuation Range】

• current: 57.86 (as of the 2026-06-12 close)

• bear (conservative · ideal buy zone): [36, 42]

• base (reasonable · acceptable holding zone): [48, 58]

• bull (optimistic · above the clearly overvalued line): [64, 76]

Research Uncertainties

First, the detailed items in the full 2025 cash-flow statement are less smoothly visible in the public web environment than in the PDF text, so maintenance capex can only be inferred as a range based on cash paid for fixed-asset construction and acquisition in 2023, 2024, and Q1 2026. Second, freely available A-share sell-side consensus data are incomplete, so this article does not mechanically use “market consensus expectations” as the core anchor, and instead relies more on company disclosures and cash-flow look-through. Third, there is no unified, continuous, authoritative quarterly public disclosure of the exact shares of domestic AI servers across the Hygon, Ascend, and Cambricon routes, so the judgment on ecosystem position is based more on cross-checking product implementation, tenders, and industry reporting. Fourth, quarterly cash flow for full-system vendors is heavily disturbed by order and collection timing, so single-quarter data are more useful as signals than as precise values. Fifth, overseas comparable companies differ materially in accounting standards, fiscal-year timing, and product mix, so horizontal valuation is more suitable as a reference than as a mechanical mapping.

Reference Sources

This article was mainly written by cross-checking the following public materials: Inspur Information's 2024 annual report, 2025 annual report summary, 2025 Q1 report, 2026 Q1 report, and investor-relations records; the company's website and product-release materials; U.S. BIS/eCFR and Federal Register Entity List documents concerning Inspur; exchange-rate and government-bond-yield data from China Foreign Exchange Trade System and ChinaBond; Reuters, Dell IR, SMCI IR, and periodic reports and market data from comparable companies including Sugon and Unisplendour.

Other Securities Mentioned in the Report

• 603019.SHG - Sugon, the direct comparison for domestic high-performance computing and liquid-cooling solutions.

• 000938.SHE - Unisplendour, the comparison for H3C's channels and enterprise ICT platform capability.

• DELL.US - A global server and enterprise infrastructure platform reference, used to compare profit and cash-flow quality.

• SMCI.US - The global mirror for high-elasticity AI servers with high financing needs.

• NVDA.US - The upstream GPU profit pool, used to illustrate the industrial structure that limits full-system vendor margins.

• 688041.SHG - Hygon Information, an important upstream player in the domestic compute ecosystem, linked to the chip routes of domestic server solutions.

• 688256.SHG - Cambricon, an important representative of the domestic AI chip route, affecting Inspur Information's heterogeneous full-system ecosystem position.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.