Terry Gou, known as the "King of OEM", has finally realized his 20-year-long obsession with transformation, and the valuation of this trillion-yuan leading computing power giant is entirely sustained by the booming AI industry.
In 2015, Terry Gou spun off high value-added assets from Hon Hai to establish Foxconn Industrial Internet (FII). This enterprise does not carry a short-term transformation plan, but a two-decade-long de-foundry strategy set back in 2005. Eleven years later, the 2026 semi-annual report was released: cloud computing business surged 75.7% year on year, AI server revenue rose 2.3 times, net profit attributable to shareholders nearly doubled, and the performance delivery intensity reached an unprecedented level.
However, the core controversy in the capital market has never been the current book profit, but that this round of doubling rise is entirely supported by the single booming story of AI computing power. FII started from the phase low of 13.61 yuan in April 2025, and hit the all-time high of 84.95 yuan in June 2026, with a cumulative increase of over 520% in the interval. The current valuation continues to run at its historical high range, and the lowest share price for the whole year of 2025 was only 14.58 yuan.
Objectively, Terry Gou's 20-year de-foundry strategy has been implemented, and the business identity breakthrough has been completed by spinning off FII — completely divesting low-margin mobile phone assembly and turning to high-end system integration businesses such as AI full racks and high-speed switches, with the cooperation mode upgraded to the joint development JDM model. However, this transformation only changed the manufacturing content, and failed to reshape the position in the industrial chain. No matter how far the full rack integration and self-developed liquid cooling are promoted, the pattern that upstream GPU/HBM chips are monopolized by NVIDIA and downstream overseas cloud vendors control procurement orders and bargaining power has not changed. The company is still a mid-stream manufacturing service provider dependent on customer capital expenditure, has not jumped out of the underlying logic of the foundry business, and all performance increments are tied to the computing power expansion plans of overseas giants.
Therefore, whether the share price can stabilize at a trillion-yuan valuation and open up new upside space in the future essentially depends on only one thing — how long the booming narrative of global AI capital expenditure can last.
Terry Gou's 20-year obsession against OEM, spinning off the main entity to bet on computing power to complete identity breakthrough
Terry Gou's entrepreneurial history is full of unwillingness for the meager profits of OEM, and this obsession finally gave birth to FII as the transformation carrier.
Hon Hai started in 1974, relying on consumer electronics connectors and complete machine assembly to seize global manufacturing share, but the dilemma at the bottom of the smile curve lasted for decades. Assembling a ten-thousand-yuan iPhone can only get dozens of yuan of processing fees. When labor and raw material costs rise, the quotation is firmly locked by Apple, and the profit margin will be directly compressed.
In 2005, Hon Hai's revenue exceeded NT$1 trillion. Terry Gou publicly put forward the long-term transformation strategy at the annual shareholders' meeting, proposed the "Foxconn inside" plan, and clearly refused to be trapped in the terminal product assembly link forever.
In the same year, Hon Hai officially entered the server ODM track, broke away from the single mobile phone assembly business and laid out high-end computing power hardware. The core demand was to extend to upstream core components and high value-added equipment, getting rid of the inherent label of "low-end foundry". At that time, he said bluntly that the pure assembly business had no growth ceiling, and the two-way squeeze from upstream and downstream would permanently compress profits, which was the official starting point of the 20-year de-foundry layout.
After 2010, the domestic demographic dividend faded and the competition in Apple industrial chain OEM intensified. Terry Gou fully recognized that the pure labor-intensive assembly has no long-term way out, and it is necessary to divest low value-added businesses and build a high-end technology manufacturing platform independently.
In 2015, Terry Gou overcame internal differences in the group, established Fujiang Technology in Shenzhen, injected all high-margin assets under Hon Hai including servers, high-speed communication equipment and industrial robots, and separated them from the traditional mobile phone assembly business. At that time, he made it clear to the outside world that the outside world had long labeled Hon Hai as a foundry, and spinning off the new entity was to tear off this label, no longer only do hardware assembly, and the goal was to become a supplier of integrated solutions for computing power and industrial internet.
In 2017, the entity was renamed Foxconn Industrial Internet Co., Ltd. In 2018, it set the record of the fastest IPO approval in A-share market at that time with only 36 days, and landed in the capital market. The market value on the first day of listing was close to 400 billion yuan, and Terry Gou defined FII as the core carrier of Hon Hai's transformation in the next 30 years.
However, the first five years of transformation lacked performance proof, and the market still classified it as an affiliated enterprise of Apple industrial chain. The share price fluctuated below 15 yuan for many years, and the narrative of "computing power service provider" was difficult to implement. It was not until the outbreak of the global large model wave in 2022 that the AI servers, full racks and heat production lines laid out in advance in the early years fully released their value. Terry Gou drew a clear upgrade path for FII: break away from single machine assembly, develop core components such as liquid cooling, power supply and high-speed switches independently upstream, and provide full-process services of integrated design, integration and delivery of full racks downstream, trying to continuously increase gross profit margin relying on vertical integration to completely dilute the OEM attribute.
But after 20 years of breakthrough layout, FII still has not been able to get rid of the restraint of overseas customer demand, and the initiative of enterprise growth is never in its own hands.
Cloud computing is exploding in full, all growth is anchored on overseas cloud vendors' AI capital investment
The eye-catching data of the 2026 semi-annual report are all phased dividends brought by the computing power expansion of overseas tech giants, and the cloud computing sector has become the absolute revenue pillar. In the first half of the year, the total revenue was 5.57861 trillion yuan, cloud computing business revenue accounted for more than two-thirds, and the 75.7% year-on-year growth rate was significantly higher than the company's overall revenue growth rate of 54.63%, which was the only core driving force for the nearly doubled profit.
The shipment data of subdivided products are fully exploding. The revenue of AI servers for overseas cloud service providers increased by 2.3 times year on year, the shipment of GPU AI full racks increased by 3.2 times, and the ASIC customized racks corresponding to Google TPU and Amazon Trainium also increased by 3 times synchronously. The revenue of high-speed network business in data centers doubled directly, the shipment of 800G switches and SuperNIC network cards increased by 140%, and the CPO all-optical switch completed prototype delivery, locking in subsequent orders from leading overseas cloud vendors in advance.
Industry data shows that the total annual capital expenditure of the world's nine major cloud vendors in 2026 exceeded 886.7 billion US dollars, with a year-on-year increase of 90%, and the five companies of Microsoft, AWS and Google in North America accounted for nearly 90% of the total. With a global market share of more than 40% in AI servers, FII has taken half of the overseas computing power infrastructure OEM orders this round.
In order to weaken the OEM label, the company continues to increase investment in vertical integration. The construction in progress of the cloud computing sector is 8.58 billion yuan, accounting for 83.4% of all construction in progress. All new funds are invested in self-developed production lines of high-value components such as liquid cooling and power supply. The delivery mode has been upgraded from single machine assembly to full rack system integration, and it is simultaneously positioned on NVIDIA's next-generation Vera Rubin platform, which significantly increases the value of a single machine, driving the comprehensive gross profit margin to rise from 6.60% to 7.15%.
However, the underlying logic of growth has not changed. Upstream core chips of GPU and HBM are monopolized and priced by NVIDIA, and downstream overseas cloud vendors hold procurement decision-making power. The company only undertakes the mid-stream links such as integration, assembly and testing. At present, the customer's raw material procurement mode continues to change from the buyout system to the consignment system. High-value chips such as GPU and HBM are supplied by overseas customers themselves, and FII completely loses the profit from raw material price difference, only charging fixed integrated service fees. Even if it enters the high-end AI track, it still cannot jump out of the inherent logic of thin profit in OEM.
Share price increase has overdrawn short-term expectations, valuation ceiling is fully tied to the AI narrative cycle
The AI computing power narrative has spawned the epic rising market of FII. Under the huge increase, the core foothold of market divergence is whether this high valuation can be continuously supported by the long-term AI capital expenditure logic.
Looking back at the starting point of the market, the phase low of 13.61 yuan in April 2025, along with overseas cloud vendors continuously raising AI capital expenditure guidelines and AI server orders being delivered in batches, capital continued to pour in to push up the share price. In June 2026, the intraday price hit the all-time high of 84.95 yuan. In just over a year, the cumulative increase in the interval exceeded 520%, and the total market value at the peak stage exceeded 1.68 trillion yuan, once ranking in the first echelon of A-share market value. By mid-August, the share price fell back to around 65 yuan. Compared with the low of 14.58 yuan for the whole year of 2025, the overall increase is still at a very high level, and the valuation has long been running above the 90th percentile of its own history.
The capital structure in the secondary market clearly reflects the market game mentality. In one year, the number of shareholders surged from 310,000 to 821,500, and retail capital deeply participated in the market, mainly betting on the long-term dividend of AI computing power. The institutional camp is clearly differentiated: Goldman Sachs raised its target price to 121 yuan, optimistic about the continuous profit improvement of liquid cooling, CPO and full rack integration, and judged that the AI capital expenditure cycle can last for many years. A large number of public offering and private placement institutions choose to reduce their positions in batches to cash out chips, and the core concerns focus on the mismatch between valuation and fundamentals.
The essence of institutional divergence is two completely opposite predictions of the "duration of AI narrative". Optimistic capital assumes that global cloud vendors will continue to increase investment in computing power infrastructure in the next 3-5 years, and FII can continuously release performance relying on its share, so the current high valuation can be digested by annual profit growth. Cautious capital points out that the current trillion-yuan market value has already priced in most of the order expectations for the next two years in advance, and the comprehensive gross profit margin of about 7% and net profit margin of 4.26% are still at the standard OEM profit level, which does not match the valuation of pure tech growth stocks.
This means that the core variable that determines how far FII's share price can go and whether its valuation can stay at a high level is not the short-term performance growth rate of a single quarter, but the length of the boom cycle of AI computing power capital expenditure. Once overseas cloud vendors lower their computing power expansion plans and the global AI capital expenditure growth rate drops significantly, the core narrative supporting high valuation will loosen directly, and there is significant room for downward revision of the valuation center. Short-term excellent performance can only maintain the bottom of the share price. To open up upside space again, it is necessary to continuously verify that the long-term AI expansion logic has not been falsified.
Four structural hidden dangers dependent on overseas capital, long-term ceiling is subject to narrative boundary
All current performance dividends and trillion-yuan valuation premiums are entirely based on the single premise that overseas cloud vendors continue to increase AI investment. Four major structural risks are hidden in the financial statements and industry pattern. Once the AI boom narrative cools down, all contradictions will be exposed intensively.
First, performance growth is completely tied to the overseas capital expenditure cycle, and there is no hedging buffer. Multiple institutions have consistently estimated that the capital expenditure growth rate of leading overseas cloud vendors will drop sharply from nearly 90% in 2026 to about 50% in 2027, and the procurement demand for computing power hardware will slow down synchronously. The top five overseas customers of the company account for more than 60% of revenue. Once Microsoft, Amazon and Google reduce the expansion scale of data centers, new AI server and switch orders will shrink directly. At the same time, Google and Amazon continue to increase the layout of self-developed ASIC chips, reducing the procurement of NVIDIA GPU in the long run, and the company's orders on both computing power routes will be under pressure synchronously. Referring to the history of sharp profit decline of Apple industrial chain enterprises in the downward stage of the smartphone OEM cycle, high-end AI OEM is only a track switch, and the underlying logic of cyclical fluctuation is completely consistent.
Second, two-way squeeze from upstream and downstream makes it difficult for gross profit margin to break through the natural upper limit of OEM. Upstream NVIDIA monopolizes the pricing power of high-end chips, and downstream overseas cloud vendors have strong bargaining power. Even if the full rack and liquid cooling business are expanded, the company's maximum comprehensive gross profit margin is only 7.35%, and the gross profit margin in the second quarter when orders are released has fallen back to 6.99%. Quanta, Wiwynn and Inspur continue to expand the production capacity of AI servers, and the industry price involution intensifies. It is difficult for the long-term gross profit margin to exceed 10%, and the model of earning processing fees by scale has not changed.
Third, hundreds of billions of inventory and high liabilities form capital shackles, and the quality of cash flow is difficult to improve synchronously. In order to meet the mass pre-stocking demand of overseas customers, the ending inventory is as high as 192.109 billion yuan, an increase of 27.30% compared with 150.913 billion yuan at the end of 2025, and nearly half of which are imported chip raw materials. In the second quarter of 2026, the net cash outflow from operating activities was about 17.6 billion yuan, and the overall operating cash flow in the first half of the year still recorded a net inflow of 7.391 billion yuan. The annual net profit of 23.7 billion yuan only corresponds to a net inflow of 7.391 billion yuan, and the feature of profit being book-based is obvious. Short-term loans exceeded 102.3 billion yuan, and the financial expenses in the first half of the year surged 750% year on year. The exchange losses brought by overseas business continue to erode profits, and the expansion of orders raises leverage pressure simultaneously.
Fourth, geopolitical and supply chain trust disturbances amplify the risk of narrative fluctuation. Overseas business accounts for more than 60% of the company's total revenue, and changes in export control and global trade policies may impact overseas order delivery at any time. Although the founder's private public opinion will not cause large equity division, it will affect the judgment of Apple and overseas cloud vendors on the stability of the supply chain, and there is a possibility that orders will be diverted to Quanta and Luxshare, further reducing the market share.
Overall, Terry Gou's 20-year goal of getting rid of the OEM identity has not been fully realized so far. All the current eye-catching performance is phased dividends spawned by huge overseas AI capital expenditure. AI full racks and high-speed switches are only upgrades of assembled products, and the passive position in the mid-stream of the industrial chain has not changed. As long as the enterprise's growth is completely dependent on the computing power investment plans of overseas customers, the valuation height and performance growth rate are all tied to the duration of the AI narrative. If the narrative cycle is extended, the valuation will be digested continuously. Once the prosperity recedes in advance, the growth logic of trillion-yuan market value will fail quickly.