Foxconn Industrial Internet H1 net profit nearly doubles to 23.7B yuan, AI server revenue up 2.3x

In an era of macroeconomic uncertainty, a listed company's half-year report showing a net profit year-on-year growth of 95.99% stands out as exceptionally bright.

On the evening of September 18, 2026, Industrial Foxconn (SH601138) officially released its half-year results. The company achieved operating revenue of 557.861 billion yuan, up 54.63% year-on-year; net profit attributable to shareholders reached 23.740 billion yuan, nearly doubling year-on-year; and non-GAAP net profit exceeded 22.9 billion yuan, up 96.99% year-on-year.

An even more striking development occurred outside the secondary market: on the same day the financial report was disclosed, Industrial Foxconn announced that it had completed its share buyback plan, spending a cumulative total of approximately 1 billion yuan at an average price of 63.16 yuan per share, representing 0.08% of total share capital.

On one hand, there is a comprehensive surge in performance; on the other, there is tangible capital deployment. This manufacturing giant, which holds a core position in the global supply chain, now stands at the crest of the AI computing infrastructure wave. What investors should truly focus on is the logic behind profit distribution within this industrial chain.

The Powerful Engine of the Cloud Computing Segment

Focusing solely on total revenue figures risks overlooking the structural transformation occurring within Industrial Foxconn.

Reviewing the half-year report, traditional ICT business, while massive in scale, has seen its overall growth rate flatten. The real driver of profit growth is the "Cloud Computing" segment. Data shows that operating revenue for this segment surged 75.7% year-on-year, becoming the absolute主力 in driving the company's overall performance growth.

Supporting this segment's rapid ascent is the global boom in data center construction. As AI technology transitions from cloud-based training to inference deployment, demand for computing power from major cloud service providers has exploded. According to company disclosures, revenue from AI servers for cloud service providers grew 2.3 times year-on-year, with shipment volumes showing exponential growth.

Specifically:

• Shipments of GPU AI cabinets, representing high-end computing power, grew 3.2 times year-on-year.
• Shipments of ASIC AI cabinets, used for specific algorithm scenarios, grew 3 times year-on-year.

This multiplication in shipments is no coincidence—it directly corresponds to the easing of global AI chip production bottlenecks and the arrival of a new peak in capital expenditure by major internet giants.

Beyond complete machines and cabinets, key components serving as the computing foundation are also accelerating in volume. The company noted in its report that revenue from high-speed network equipment doubled year-on-year. In particular, shipments of data center switches supporting ultra-high bandwidth (800G and above) grew 1.4 times year-on-year. This indicates that Industrial Foxconn is not merely selling servers but is deeply involved in the systematic construction of global computing infrastructure—from chip-level modules to cabinet-level integration and network communication equipment, deepening its participation across the entire AI hardware chain.

The Dividends and Challenges of the "OEM" Model

However, it must be clearly recognized that despite its "technology manufacturing" label, Industrial Foxconn's essence remains manufacturing.

While the 23.7 billion yuan in profit is certainly enticing, when placed against the backdrop of 557.8 billion yuan in revenue, its net profit margin remains only around 4.25%. This means that, amid massive revenue flows, every yuan of profit requires extreme cost control, exceptionally high yield rates, and extremely stable economies of scale.

This is precisely the moat of a super OEM (JDM/ODM) enterprise like Industrial Foxconn. When upstream chip manufacturers like NVIDIA and downstream cloud providers vie for influence, the ability to transform complex manufacturing processes into high-yield products is itself the hardest barrier. After all, an AI chip worth thousands of dollars is useless if it cannot be installed in a well-cooled, stable-power-supply cabinet.

Internally, the company has built 9 certified "Lighthouse Factories" and externally empowered 13 more. With a patent portfolio exceeding 7,600 items, it provides underlying support for its smart manufacturing capabilities. This "soft-hard combination" allows it to command a technical premium when undertaking high-end AI server OEM orders that ordinary foundries cannot match. Yet, it must also be noted that the OEM model inherently makes it difficult to achieve the high brand附加值 and profit margins enjoyed by chip design companies.

Risks also follow. Heavy reliance on orders from a few major customers means that once the capital expenditure cycle of global cloud computing companies fluctuates, or if a disruptive change occurs in the next-generation AI architecture, profit margins in the midstream manufacturing segment are easily squeezed. The current prosperity is built on the historical window where the AI computing power gap remains unresolved. As more production capacity comes online in the coming years, whether supply-demand dynamics will reverse remains a question mark hanging over the entire sector.

The Financial Strength Behind the 10 Billion Yuan Buyback

Alongside strong performance, the capital market has also voted with its feet. After a period of stock price appreciation, Industrial Foxconn chose to increase its commitment.

According to its results announcement submitted to the Shanghai Stock Exchange, as of September 18, the company had spent a cumulative total of approximately 1 billion yuan from its own funds to repurchase 15.8334 million shares, representing 0.08% of total share capital. The repurchase price ranged from 59.91 yuan/share to 66.40 yuan/share, with an average transaction price of 63.16 yuan/share.

Notably, even under prevailing market conditions, its highest purchase price was nearly 6% above the closing price, demonstrating the company's strong willingness to support its stock price and its financial strength even at elevated levels. Looking at the repurchase range, the lowest purchase price was close to the 60 yuan mark, indicating that management has a clear judgment framework for the company's long-term valuation floor.

This 1 billion yuan buyback is not intended for long-term equity lock-up. According to subsequent arrangements in the announcement, these repurchased shares are planned to be sold through centralized竞价 trading within 12 months after the results announcement; if not sold within 3 years, they will be cancelled. This is somewhat nuanced in business logic: on one hand, it demonstrates management's confidence in the company's fundamentals; on the other, it utilizes idle funds to enhance returns or adjust the equity structure.

For shareholders, the fact that a manufacturing enterprise once viewed as a traditional cyclical stock achieved net operating cash flow of 7.391 billion yuan, surging over 425% year-on-year, is a more reassuring fundamental reality. This indicates that the company's profit quality is quite healthy, with the vast majority converted into genuine cash inflows.

Conclusion: Key Indicators on the Long Slope of Computing Power

The wealth-creation myth brought by AI has not remained on PowerPoint slides but has been tangibly realized as hundreds of billions in revenue and over twenty billion in profit on Industrial Foxconn's books. However, as the industry enters its second half, investors should not be blindly optimistic.

In the coming quarters, it is recommended to focus on the following three indicators:

1. Gross margin trend in the Cloud Computing segment. As scale expands, will production bottlenecks be broken? Or will rising raw material and labor costs compress profit margins? Gross margin is the core signal for judging whether the OEM premium can be maintained.

1. The pace of transition to next-generation AI server architectures. Whoever can complete the R&D and mass production of the new computing platform first will lock in the next round of超额 profits. The industry has already begun discussing the application prospects of higher-bandwidth interconnects and new packaging processes.

1. Signals of a turning point in overseas capital expenditure cycles. Currently, the world's largest cloud providers are still frantically stockpiling computing power. But how long this spending spree can last will determine how many more years this feast can continue. Monitoring the quarterly guidance of core customers like Microsoft, Google, and Amazon is the most direct tracking method.

There is no doubt that Industrial Foxconn has secured its first ticket. But on the long track of computing power, it still has a long way to go.