Foxconn

IThe business

Hon Hai Precision Industry is the company most people know as Foxconn. It builds other companies’ products. In the four quarters to June 2026 it had revenue of NT$9.31T and an operating profit of NT$327B, a margin of 3.5% C (calculated).⁴²⁵ That is about US$284B of sales at NT$32.74 to the dollar, and about US$10.0B of operating profit C (calculated).³ Out of every NT$100 it collects, NT$94 goes to the cost of what it sells, NT$2.60 to selling, research and administration, and about NT$3.50 is left.

This page is about that last number. A company that keeps 3.5% of its sales has a different problem from one that keeps 30%. It has to move enormous volume, collect from customers who are bigger than it is, and finance the parts it buys until it is paid. When sales rise by a third, as they are doing now, the cash tied up in parts and receivables rises with them. Margins decide what the revenue is worth. Revenue alone does not.

What it makes

Hon Hai reports four product lines.¹¹ Cloud and networking is servers, racks, switches and communications equipment; it was 40% of 2025 revenue and 50.5% of the June 2026 quarter.¹⁵² Smart consumer electronics is smartphones, televisions and game consoles; it was 38% of 2025 revenue. Computing is notebooks, tablets, desktops and printers, 15%. Components and other is connectors, precision parts, electric vehicle parts, semiconductors, and services, 7%.¹⁵ The company has 240 sites in 24 countries and about 900,000 employees.¹²¹⁵

The shift is fast. In 2022 cloud and networking was 23% of revenue and smart consumer electronics 53% C (calculated).¹⁶ In the second quarter of 2026 cloud and networking was NT$1,276B of NT$2,526B, and the chairman said in March that it had passed consumer electronics for the first time in the fourth quarter of 2025.²¹⁷ Consumer electronics did not shrink in the last year; it rose 10% in the first half of 2026, partly on higher prices.²⁶ The cloud line grew 77% in the same half, seven times as fast.

How the money is made

Hon Hai is a contract manufacturer. A customer designs a product, or increasingly co-designs it with Hon Hai, and Hon Hai builds it in volume, buys most of the parts, and ships it. The profit is the gap between what the customer pays and what the parts, labour, plants and engineers cost. In 2025 that gap was 6.15% of revenue before operating costs and 3.20% after them C (calculated).⁴

Two features of the business explain most of what follows on this page. The first is that revenue contains a lot of other people’s parts. When Hon Hai buys a graphics processor or memory chip, builds it into a server and sells the server, the whole price of the chip is revenue and nearly all of it is cost. This is why gross margin fell as cloud grew: the fourth-quarter call said the "buy and sell" server model was diluting it.¹⁷ Some customers now supply their own chips and pay Hon Hai a fee to build around them. Hon Hai calls this consignment. The company said it has moved part of its server business to consignment in 2026, starting with custom chip (ASIC) projects, to cut the working capital it ties up.¹⁷¹² Consignment lowers revenue and working capital by about the same amount and barely changes profit. Revenue growth rates are therefore a poorer guide to this company than they look.

The second feature is that the profit is made in one place. Hon Hai reports its results by group: the parent group, Foxconn Industrial Internet (FII), which is listed in Shanghai, and FIH Mobile, which is listed in Hong Kong. In the first half of 2026 FII earned NT$136.3B of the NT$170.5B group operating income, 80% of it C (calculated). The parent group’s own operating income fell from NT$38.0B to NT$29.1B, down 23% C (calculated).²

FII makes cloud and networking products, which is where the AI racks are. Outside holders receive about 16% of FII’s comprehensive income C (calculated).² That share leaves the group before it reaches Hon Hai shareholders. In the June 2026 quarter minority holders took NT$10.1B of the NT$70.1B of profit, 14.4%, against 10.4% a year earlier C (calculated).² I model a 14% share.

What changed in 2026

The second quarter of 2026 was the best in the company’s history: revenue of NT$2,526B, up 41%; operating profit of NT$94.8B, up 68%; net profit to the parent of NT$60.0B, up 35%; earnings per share of NT$4.27.²¹² The gross margin was 6.12%, down 0.21 point from a year before. Operating profit grew faster than revenue because gross profit rose 36% and costs below it rose 5%.¹² September revenue then reached NT$1,158.6B, the first month above NT$1 trillion, and the third quarter came to NT$3,026.9B, 47% more than a year before.⁶

The things around the core

Hon Hai holds NT$203B of equity-method investments, NT$175B of shares at fair value through other income and NT$102B of other holdings at fair value, at 30 June 2026.² The best-known holding is Sharp. Hon Hai completed its investment in Sharp on 12 August 2016, and on the second-quarter call it said it would keep supporting the Japanese company.¹² In electric vehicles it owns 46.25% of Foxtron Vehicle Technologies and consolidates it because it directs the company’s decisions.⁴ It also paid in US$22.7M for ordinary shares and US$30.0M for preferred shares of NuRide (formerly Lordstown Motors) of the US$170M it agreed to invest in 2022, and is a defendant in a suit by Lordstown’s bankruptcy estate.⁴¹⁹

Electric vehicles are small. The company gives no revenue for them. In March the chairman said the point at which carmakers start to outsource had come later than the company expected, and in August management said its Mitsubishi Motors vehicle would begin delivery in the second half of 2026.¹⁷¹² I treat electric vehicles as something I cannot judge from the filings, and my model has no line for them.

Hon Hai is a very large assembler whose profit now comes from one group and one product line. Its margin is thin and its customers are big. That is the whole business.

IIThe moat

I think Hon Hai has a narrow protection that works on its customers’ side of the table and not on its own. It is very hard to replace as a supplier, because few companies can build phones or AI racks in huge volume on a deadline. It is not hard to bargain down, because ten years of margins show customers keep most of the value.

Evidence it holds

Scale nobody else has. The annual report says Hon Hai has more than 40% of the electronics manufacturing services market and the largest share in consumer electronics.¹¹ Revenue of NT$9.31T in the last twelve months is 1.9 times TSMC’s NT$4.95T in its twelve months to September 2026 C (calculated).²²⁰ A customer launching a product needs the plants, the people and the logistics ready months ahead. Few suppliers can promise that.

Depth in AI racks. On the second-quarter call management said it supplies "the entire system, not a single product": components, GPU modules, motherboards, switch and compute trays, rack assembly, and the links between racks.¹² It says the in-house share of components in a rack is above 50% and that compute-tray assembly is fully automated in the next generation.¹² On the fourth-quarter 2025 call it described itself as "a core manufacturing and system integration partner for NVIDIA."¹⁷ These are management claims. I could not check the in-house share or the automation rate, and I did not open a competitor’s filing for this page.

Old contracts with the biggest names. The annual report’s list of important contracts has sale-and-purchase agreements with Apple, Cisco, Dell, HP, IBM, Lenovo, Microsoft and Sony. The Apple one is dated 7 December 1999 and runs until one side ends it.¹¹ The filings do not name their largest customer. They call it Client E, and it bought NT$3,574B in 2025, 44.1% of revenue.⁴ I infer, and the company does not say, that Client E is Apple. Every one of those product contracts has the same restriction column: the customer "reserves right to adjust delivery dates and order".¹¹ The relationship is long, and the customer keeps the right to change the order.

Evidence it narrows

The margin has not moved. Gross margin was 7.4% in 2016, 6.3% in 2024 and 6.1% in 2025 C (calculated).⁷ A supplier with pricing power would show a rising line while its sales rose by 86%. Hon Hai’s line fell. What rose was the operating margin, from 2.1% in 2020 to 3.2% in 2025, and that came from costs below gross profit growing slower than sales.⁷¹⁷

The biggest customer is flat. Client E bought NT$3,673B in 2022 and NT$3,574B in 2025. Everyone else bought 53% more over the same three years C (calculated).¹⁶⁴ A protected position grows with the customer. This one stayed about level for three years while the rest of the company kept growing. The customer is also big enough to move: a fifth of its volume is NT$715B of revenue C (calculated).

One supplier is more than a third of purchases. The annual report names a single vendor, "Vendor L", as 37.0% of 2025 purchases and 39.3% of 2024 purchases.¹¹ The report does not say who it is. A company whose sales go to one customer and whose purchases come from one supplier is in the middle of two negotiations it does not control.

Others build parts of the same racks. Management said on the second-quarter call that most peers supply only part of the AI server chain, and that standardisation does not lower barriers to entry.¹² That is its own view. It also said it is still moving toward a 50% share of racks, and the CFO put the aim for custom-chip servers at a share above 40%.¹² A company aiming for 50% has rivals for the rest.

The position is hard to dislodge and hard to improve. The margin says the customer decides how much of the value stays here.

IIIThe competitive advantage

Hon Hai’s advantage is a stack of capabilities that a customer would have to rebuild elsewhere: component making, board assembly, automation, rack integration and plants in 24 countries. The question this page keeps asking is how much of the stack’s value Hon Hai keeps. The table puts it beside three companies on this site.

Revenue is not value

Hon Hai sells about as many dollars of product as Nvidia: US$284B against Nvidia’s US$303.0B in its latest twelve months C (calculated).³¹⁰ Nvidia’s market value on its page is US$4.97 trillion. Hon Hai’s is US$106B at today’s price, so Nvidia is worth 47 times as much C (calculated).¹⁰¹ The gap is the margin. Nvidia keeps 65.2% of sales as operating profit and Hon Hai keeps 3.5%. TSMC, the other large Taiwan company on this site, had revenue about half of Hon Hai’s and a market value 18 times as large.²⁰ Each company’s value tracks the profit left over after its customers and suppliers are paid, not the size of the flows it handles.

Against its customers

Hon Hai’s customers are companies like these. It calls itself a core manufacturing and system integration partner for Nvidia, and its list of important contracts names Apple.¹⁷¹¹ Apple has an operating margin of 33.2% and Nvidia 65.2%.²¹¹⁰ Hon Hai makes 3.5% on the products it builds for such customers. A phone or a rack gets built for a few percent of its price. This is not a flaw in Hon Hai’s management. It is how the work is priced: the customer designs the product, owns the brand and the software, and could in principle move the work to another plant.

Where Hon Hai wins

It wins on being the safe choice for a big new launch. When Nvidia, a cloud company or a phone maker needs large volume of something that has never been built, Hon Hai is the supplier whose delivery risk is lowest, in my judgment. It wins on the cost of switching: a customer that moves a design to another builder has to qualify a new plant. I expect that to take months but have no source for the time. It wins on the geography of the plants. The company says it has put production in the countries its customers sell to, and it points to 240 sites in 24 countries.¹² And it wins on capital: an S&P rating of A- with a positive outlook, net cash of NT$219.8B at June, and a cash conversion cycle of 42 days, or 52 days before the factoring of receivables.¹²

Where it loses

It loses on price. Gross margin has not risen in a decade. It loses on concentration: the filings show one customer at 44% of sales and one supplier at 37% of purchases.⁴¹¹ It loses on the cost of growth: the faster Hon Hai grows, the more cash it ties up. And it loses in one category the company itself has flagged: electric vehicles, where the outsourcing it expected has been slow to come.¹⁷

Hon Hai has the widest stack in electronics assembly and the thinnest margin on it. The stack is the reason it is hard to replace. The margin is the reason it is worth a fraction of its customers.

IVManagement

Who runs it and what they own

Young Liu has been chairman since June 2019 and is also general manager. He founded a motherboard company and two chip companies in the United States, one of which listed on Nasdaq in 2001, and joined Hon Hai in 2007 when the founder, Terry Gou, recruited him.¹³ Hon Hai also runs a rotating CEO program, started in April 2024 to build a bench of successors; the second-quarter call was led by the rotating CEO, Michael Chiang, with CFO David Huang.¹³¹² The founder is still the largest shareholder: Terry Gou held 1,739,698,518 shares, 12.40%, at 31 March 2026, which is NT$427,096.0B at today’s price C (calculated).¹⁴¹ The annual report lists nine directors, five of them independent, and does not list him among them.¹³ The next largest holders are Taiwan’s New Labor Pension Fund at 1.77%, an index fund at 1.68%, and LGT Bank, Vanguard funds, Norway’s central bank and Singapore’s government fund at about 1% to 1.3% each.¹⁴ I show these as holdings, not as views.

How they are paid, and what that pays them to do

Liu’s 2025 pay was NT$1.37B, or about US$42M, of which NT$1.18B, 86%, was employee compensation: a share of profit.¹³ The report says managers’ pay is salary, a performance bonus allocated by each group’s contribution to the company’s profit, and employee compensation tied to the year’s result. The company set aside NT$11.75B for employee compensation in total.¹³ The independent directors get a fixed fee and attendance fees and no variable pay.¹³ So the people who run the company are paid for profit dollars. They are not paid for the return on the capital those dollars needed. A year with 30% more profit on 50% more capital lifts their pay and may lower the return to shareholders.

The company’s stated targets point the other way: operating margin above 3% over the medium term, return on capital above 12% and net debt below 1.5 times EBITDA.¹² On the second-quarter call the rotating CEO said "we will definitely continue to pursue the operating objective of maximizing earnings", in the company’s translation.¹² Earnings, again, not margin.

How much of their own money is in

Not much. Liu held 656,219 shares at 31 March 2026, worth NT$161M at today’s price C (calculated), and one year of his pay is 8.5 times that.¹³¹ The report’s financial director held 708,126 shares.¹³ I take the financial director to be the CFO named on the calls. The independent directors held none. Their wealth is in their salary and profit share, not in the stock. The founder is the exception, and he is not an executive.

What they did with capital over a cycle

From 2016 to 2025 Hon Hai earned NT$1,399B of profit for its owners and spent NT$934B on property and equipment C (calculated).⁷ Capital spending went from NT$53B to NT$174B a year. From 2019 to 2025 it paid NT$470B in cash dividends C (calculated).²²⁸¹⁶⁴ The dividend per share rose from NT$4.0 for 2018 to NT$7.2 for 2025, a yield of 2.9% at today’s price, and the NT$100.9B payment was the seventh year in a row above 50% of profit.⁷¹²¹

It has not bought back shares on the market that I could see. Treasury stock was NT$15M, held by a subsidiary since 1998, at 30 June 2025, 31 December 2025 and 30 June 2026.² It did return cash through a capital reduction: in August 2018 it paid out NT$34.7B, NT$2 a share, and cancelled 20% of its shares, taking the count from 17,329 million to 13,863 million.¹³ Issued shares were then 13,863 million until late 2024 and 14,073 million in June 2026, up 1.5%, from conversions of overseas convertible bonds and a share exchange for TECO Electric shares in October 2025.¹³² Since 2019 earnings per share rose from NT$8.32 to NT$13.61, up 64%, total equity per share, minority holders included, from NT$101 to NT$142, and the year-end price from NT$90.80 to NT$230.50.⁷²³ I start at 2019 because per-share figures before the reduction are on a different share count.

The large bets are the ones outside the core. Sharp was a 2016 investment the company still supports; it lifted results in 2025 when Sharp recovered.¹⁷¹² The NuRide investment in electric vehicles is still open, and the lawsuit is still running.⁴¹⁹ The record on capital allocation is mixed: the core spending has raised sales and profit, though not cash, and the side bets have not paid yet.

What I can’t judge

I cannot tell how much of the profit share goes to which people below the nine managers listed, or how the company decides to scale capital spending against the profit it earns in a given group. I cannot see the electric vehicle economics at all, because the company gives no revenue for it. And I cannot judge the rotating CEO program: the program started in April 2024 and has little record yet.

The managers are paid a share of the profit and own very little stock. The founder owns 12.4% and does not run the company. That is a reasonable structure for profit. It is a weaker one for return on capital.

VFinancials

Ten years of numbers tell one story: sales have risen by 86%, profit has risen by 27%, and cash has been the hard part. The figures come from Hon Hai’s own workbook and its consolidated reports; I checked the cash flow lines against the reports for 2016, 2017, 2019 to 2025.

Revenue and profit

Net income to the parent was NT$148.7B in 2016 and NT$189.4B in 2025, up 27% while revenue rose 86% C (calculated).⁷ Earnings per share rose from NT$8.32 in 2019 to NT$13.61 in 2025, up 64%. Before August 2018 the company had 17.33 billion shares; a cash capital reduction cancelled 20% of them, so I compare per-share figures only from 2019 on.⁷¹³ The profit for the four quarters to June 2026 is NT$212.8B, and earnings per share are NT$15.21 C (calculated).⁵²

Margins

The margin is the thing to watch. Operating margin was 4.0% in 2016, 2.1% in 2020, 3.2% in 2025 and 3.7% in the first half of 2026 C (calculated).⁷² Costs below gross profit were 3.4% of revenue in 2016, 2.95% in 2025, and 2.48% in the first half of 2026 C (calculated).⁷⁴² The company says operating margin will beat last year’s 3.2% for 2026 and that its medium-term target is above 3%.¹² Each 0.1 point of operating margin on today’s revenue is NT$9.3B of profit, about 3% of the operating profit C (calculated).

Cash, which is the number that matters

Operating cash flow in the four quarters to June 2026 was NT$136B and capital spending NT$177B, so cash after capital spending was −NT$42B against a profit of NT$213B C (calculated).⁴² In the first half alone the operating cash flow was an outflow of NT$69.1B, because receivables rose NT$178.6B and inventory NT$221.6B while payables rose NT$157.3B.² Management called it a phase and pointed to EBITDA of NT$227.7B for the half, up 52%.¹² The same pattern is in the record. In 2021, when revenue rose 12%, operating cash flow was an outflow of NT$98.2B. In 2023, when revenue fell 7%, it was an inflow of NT$445.6B.⁷⁸¹⁶ When sales grow quickly, cash goes into goods and receivables. When they stop, it comes back.

Over ten years cash after capital spending was 45% of profit C (calculated). Over seven, from 2019, it was 73%, and the dividends paid were 65% of that cash C (calculated).^7228164 Stock pay is NT$1.4B to NT$4.4B a year, so it does not explain the gap.^2281694 Neither does net interest: interest received less interest paid was between about −NT$0.3B and +NT$16.9B a year C (calculated). A company that is not growing shows little gap between profit and cash. This one grew, so cash that would have shown up as profit went into the next quarter’s goods.

Debt, cash and what the net cash is

At 30 June 2026 Hon Hai had NT$1,522B of cash and time deposits and NT$1,305B of loans, notes and bonds, a net cash balance of NT$217.6B on my count and NT$219.8B on the company’s C (calculated).²¹² A year earlier the company’s figure was NT$243.6B.¹² Short-term loans rose from NT$512B to NT$845B in a year.² Receivables were NT$1,385B, inventory NT$1,370B, up 46% in a year, and trade payables NT$1,528B.² The company’s cash conversion cycle was 42 days, and 52 days if the receivables it sells to banks are put back.¹² At the end of 2025 net cash was NT$339.1B, after rising NT$248B in the fourth quarter, mainly because more receivables were sold to banks.¹⁷ I treat the net cash as real but not free: the bank lines that fund goods in transit are part of the same balance sheet.

The debt ratio is 62%, S&P rates the company A- with a positive outlook, and the company says it will keep net debt below 1.5 times EBITDA and does not plan to raise equity.¹² The financing is not a worry on my numbers. The point is the size of the claims ahead of the owners: NT$3.5 trillion of liabilities against NT$5.6 trillion of assets.²

Quality of earnings

Three things sit between operating profit and what reaches shareholders. Tax. The effective rate was 26.7% in 2025 and 26.2% in the June 2026 quarter, up from 24.0%, because the company is moving profit between countries and the minimum tax on multinationals applies.¹⁹¹² Minority holders. Outside owners of FII and others took 12% of net income in 2025 and 13.5% in the first half of 2026 C (calculated).¹⁹² Investment gains. Non-operating income was NT$34.2B in 2025, from funds, equity-method investments and Sharp, and a loss of NT$1.6B in the first half of 2026.¹⁹² Net income is a better measure than operating income for what the owners keep, but it moves for reasons that have nothing to do with building servers.

Who the customers are, and where the assets are

The geography table shows two different maps. Revenue is booked to customers’ registered locations: the United States 45%, Singapore 22%, Ireland 11%, China 4.6% and Taiwan 2.5% C (calculated).⁴ Singapore and Ireland are where large customers book their purchases; they say nothing about where products are shipped. The long-lived assets are somewhere else: 42% in China, 14% in Vietnam, 11% in the United States, 11% in India, 8% in Taiwan and 8% in Mexico C (calculated).⁴ China’s share of these assets was 65% in 2020 and 42% in 2025 C (calculated).⁸⁴

What the market has done with the multiple

For eight years the stock traded at 9 to 13 times earnings. It is 16.1 times the last twelve months now C (calculated).²³⁷⁵¹ The profit per share has risen 33% from 2022 to 2025, and the price has risen 131%. Whoever bought at the 2022 close of NT$99.90 has been paid for the AI boom and the rerating. Anyone buying at NT$245.50 pays the higher multiple again, and it holds only if the earnings keep rising.

VIValuation

First, how this goes badly wrong

Before the numbers, the three ways I lose money, most likely first.

The margin gives way. Hon Hai keeps 3.5 cents of each dollar of sales. If customers or rivals take half a point of that, the profit falls by 14%. In my model the operating margin on its bear path, which falls to 2.5%, takes NT$66 a share off the base E (estimate). That is the most likely failure, because the ten-year record says the margin has been set by customers and the AI racks are new, large and priced against rival builders.⁷¹² If cloud and networking revenue followed its bear path instead, with the margin unchanged, the loss would be NT$24.

Growth eats the cash. Revenue is rising by a third a year and working capital is about 12.5% of it. In the base case owner cash is −NT$92B in year 1 and does not turn positive until year 2.² The cash comes back when sales stop growing, as in 2023, but the loans that fund the wait are real, and short-term loans rose from NT$512B to NT$845B in a year.² If a customer’s demand is cut after Hon Hai has bought the parts, the inventory of NT$1,370B is Hon Hai’s problem.

One customer or one place fails. The largest customer is 44% of sales, and the advanced chips in the racks depend on a few suppliers, the best known of them in Taiwan. I price the large shocks in the geopolitical section below. A fifth of the largest customer’s volume is NT$715B, and in my model it costs only 3% of value if the margin on the rest holds, and 5% with idle plants, because the margin is thin and the working capital comes back. A blockade around Taiwan takes about 40%.

What an owner can take out

Owner cash is the cash left after tax, after the capital needed to run the business and after the working capital growth needs. Stock pay is small, NT$2.7B in 2025, and is already in the costs I start from.⁴ In the four quarters to June 2026 cash after capital spending was −NT$42B against profit of NT$213B C (calculated).⁴² In 2025 it was NT$53.1B against NT$189.4B, 28% of it.⁴⁷ Over ten years it was 45%. The gap between what the accounts show and what an owner could take out is NT$254B for the last twelve months C (calculated).

That number is a growth number, not a standing one. Suppose sales stopped growing, working capital stopped growing and capital spending fell to the level of depreciation. Then owner cash would be about NT$208B a year: the NT$327B operating profit, less 26% tax, less the 14% share of the minority holders E (estimate). On a market value of NT$3,455B that is 6.0%. The Treasury ten-year yield was 4.91%.²⁵ So even with the growth stopped, the price pays a little more than a bond for owner cash at today’s margin, and the rest of the price is a bet that growth and margin continue.

What I can’t judge

I cannot see the margin on AI racks. Hon Hai reports profit by group, not by product, and says its gross margin is shaped by mix and by the buy-and-sell or consignment model.¹² I use one operating margin for the whole company. If I am off by half a point in every year the value moves by about NT$42 a share E (estimate), which is why I show the range and do not rely on a single number. I cannot judge the electric vehicle business, for which the company gives no revenue, so I leave it out. I cannot judge how much of the largest customer’s volume is secure. And I cannot judge the exchange rate: the New Taiwan dollar rose from about 32.8 to 28.8 per US dollar in 2025 before it fell back, and the company said the rise cost gross margin that year.¹⁹¹⁷ I hold the rate at NT$32.74, as I do on my TSMC page, which means my values assume no repeat.³²⁰

What the price implies

At NT$245.50, Hon Hai’s market value is NT$3,455B (US$105.5B) on 14,073 million shares.¹²³ That is 16.1 times the last twelve months of earnings per share, NT$15.21, and a dividend yield of 2.9% on the NT$7.2 that went ex-dividend on 2 July C (calculated).⁵²¹²²⁶ Take out net cash and investments to the parent after my haircuts, NT$394B or NT$28 a share, and the operating business is priced at NT$3,061B, NT$218 a share C (calculated).² That is 14.7 times the NT$208B of after-tax operating profit that belongs to the parent C (calculated).

The price needs the earnings to grow, and my base says they do: operating profit rises from NT$327B to NT$460B in the first year E (estimate). The price pays for that growth in profit and for a wider margin as well.

Run my model backwards

Hold my base revenue path and ask what the price needs. The operating margin from year six has to be 4.03% instead of 3.3%. Or the operating margin has to be 3.90% from year two. Or the cloud growth has to be 2.0 times my base path in every year after the first. Or the discount rate has to be 9.30% instead of 10.5%.²⁷ A margin of 4.0% from year six would match the best year of the last ten, 4.0% in 2016, and hold it for a decade.⁷

Margin is the lever. Each 0.1 point of operating margin, in every year, is worth NT$8 a share E (estimate). In my one-at-a-time test the operating margin’s bear and bull paths move the base by −NT$66 and +NT$58, the cloud revenue paths by −NT$24 and +NT$23, and the working capital by about NT$10 either way. Revenue in the other product lines hardly matters at all, ±NT$3, because the margin on it is thin and the working capital comes back.

The discount rate

I use 10.5%, in dollars, the same rate as my TSMC page so the two compare.²⁰ The latest US ten-year yield, for 2 November, was 4.91%; adding the implied equity risk premium of 4.20% at Hon Hai’s measured beta of 1.03 against the Taiwan index gives 9.2%.²⁵²⁷²⁸ Against the S&P 500 the beta is only 0.27, which tells me the stock moves with Taiwan and not with the US market. I add about 1.3 points for the dependence on Taiwan and for one customer. The model is built in New Taiwan dollars at a constant NT$32.74 to the US dollar, so it is a dollar model; no change in the exchange rate is assumed.³

One point on the rate moves my base by 13% to 17%. At 9.5% the base is NT$237 and at 11.5% it is NT$177. Growth after year 15 matters little, NT$200 at 2.0% and NT$206 at 4.0%, because I charge the capital that growth needs.

Bear, base and bull

The model runs fifteen years, from July 2026 to June 2041. Revenue has two pieces: cloud and networking, which starts at NT$4,250B in the twelve months to June 2026, and everything else, NT$5,061B. A single operating margin applies to the total. Depreciation comes from the property, equipment and intangible base of NT$678B, at 16% a year, and new capital spending is added to it. Tax is 26% of operating profit. Working capital is 12.5% of annual revenue; at 30 June it was 13.7% of the trailing revenue. Each year’s owner cash is operating profit after tax, plus depreciation, less all capital spending and the build in working capital, times 86% for the minority holders. After year 15 the value uses growth of 3% and a return on new capital of 12%, which is the company’s own stated target.²¹²

Bear. The AI racks are a bulge. Cloud and networking revenue grows 62% in the first year because the third quarter is already in the books, then falls 12% and 8% as the largest cloud buyers digest, and settles at 3%. The operating margin falls to 2.5%, where it was in 2021 to 2023. Revenue in year 5 is NT$11.2T and the share is worth NT$137.

Base. Cloud and networking revenue grows 68% in year 1, because the third quarter is NT$3.03T and the racks keep ramping, then 18%, 11%, 8% and 7%, fading to 3% by year 15.⁶¹² The rest of the company grows 8% in year 1, partly on prices, then 3% and 2.5%. The operating margin starts at 3.65%, a little below the first half, and eases to 3.3% as rivals and customers take some of the gain. Capital spending is 2.0% of revenue in the first years, as the company said it will rise at least 30% in 2026, and falls to 1.6%.¹² The share is worth NT$203. A long-run margin of 3.5% instead of 3.3% would give NT$214. I chose 3.3% because the margin was below 3.0% in eight of the last ten years E (estimate).⁷

Bull. The racks stay in short supply. Cloud and networking revenue grows 74%, 30%, 20% and 14% in the first four years, the company meets its aim of about half the market for the next rack generation and the margin climbs to 4.1% as in-house parts and optical switches add value.¹² Revenue in year 10 is NT$28.0T. The share is worth NT$300.

Weighted 25%, 50% and 25%, the cases give NT$210 a share against a price of NT$245.50.¹ If my base case happens, the price returns 9.3% a year; in the bear 6.8% and in the bull 11.8% C (calculated). Only the bull pays my 10.5% hurdle.

The next best use of the same dollar

The nearest comparison on this site is Nvidia, whose racks Hon Hai builds. Nvidia’s price is 5% above my base and Hon Hai’s is 21% above.¹⁰¹ TSMC, the other Taiwan company, is 18% above my base, and Apple, the likely largest customer, is 93% above.²⁰²¹ These are different models with different risks, and I do not rank them. On my numbers Nvidia’s price is closer to its value than Hon Hai’s is, and Nvidia keeps 65% of its sales as operating profit against Hon Hai’s 3.5%. The risk-free alternative pays 4.91% in dollars.²⁵ Hon Hai’s own dividend is 2.9%.

The margin of safety

My margin-of-safety price is NT$132, the base value less 35%. My usual cushion is 25%. I add ten points because the first year of owner cash is negative, because the margin input is a judgement within a point either way, and because one customer is 44% of sales. At NT$132 even my bear case, NT$137, would be worth 4% more than the price. The price of NT$245.50 is 86% above that level and 21% above the base.

Quality and price

These are two different questions. As a business, Hon Hai is large, reliable, financed by investment-grade lenders and the supplier customers trust with their biggest launches. It is also a business that keeps a few cents on the dollar and does not set its own price. As a purchase at NT$245.50, the price is above my base value and sits between the base and the bull. With growth stopped, the business would yield about 6.0% on it. The two do not have to agree: this looks like a decent business at a price that already counts on the best margin of the last ten years, held for a decade.

Multiplying the odds

The price needs three things at once. Cloud and networking revenue has to follow my base path or better, 18%, 11%, 8% and 7% in years 2 to 5 (I give it 60%). The operating margin has to settle near 4.0%, which it has reached once in ten years (I give it 20%). And no tariff, customer or Taiwan event can cost more than a tenth of the value over fifteen years (70%). Multiplied, that is 8% E (estimate). These are my guesses. They show why a price above the base is a bet on three things going right.

Where this could be in 5 and 10 years

The base case needs the AI racks to keep growing through the Vera Rubin generation, and the margin to ease only to 3.3%. At five years the base is worth NT$258 a share, with only NT$29 of owner cash paid before then, because the company spends on capacity and working capital first. At ten years it is NT$313, with NT$121 of owner cash paid on the way. The bear and bull differ by the cloud cycle and the margin: in the bear, cloud and networking revenue falls and the margin returns to 2.5%; in the bull revenue reaches NT$28.0T by year 10 and the margin reaches 4.1%. The way the base is most likely wrong is the margin. A half point lower in every year takes about NT$42 a share off every value in the table.

These are estimates of value, not forecasts of the share price. The price can sit above or below value for years. The 9.2% in the cases table is the rate that makes the base cash flows equal today’s price. The return in this table is lower, 5.9% for the base, because it marks the stock at my 10.5% rate in 2036 and leaves out today’s net cash and investments of NT$28 a share, which the price includes.

The price is 21% above the base. The business is a good one with a thin margin, and the price needs that margin to match its best year of the last ten and hold it.

VIIThe case for the price

What most people miss

First, growth takes cash out of this company. In the first half of 2026 profit was NT$127B and operating cash flow was an outflow of NT$69.1B; in 2023, when revenue fell 7%, it was an inflow of NT$445.6B.²¹⁶ Second, nearly all the profit growth sits in one Shanghai-listed subsidiary: FII earned NT$136.3B of the half’s NT$170.5B operating income, while the parent group’s fell 23%.² Third, the biggest customer is not driving it. Client E bought less in 2025 than in 2022, and everyone else bought 53% more.¹⁶⁴

The bull case

The case is made by the company. Chairman Young Liu said on 16 March 2026 that "2026 will still be a very positive year".¹⁷ On 12 August the rotating CEO, Michael Chiang, said the largest North American cloud buyers’ capital spending is expected to approach US$800 billion this year and soon cross US$1 trillion, that AI rack shipments are expected to grow "multiple times" in 2026, and that Hon Hai is still moving toward its target of a 50% share of racks.¹² He said orders for next year "will continue to move upward", and that the medium-term operating margin target stays above 3%.¹² The CFO added that the target for custom-chip (ASIC) servers is a market share above 40%, and that the company has no plan to raise cash through new shares.¹² I found no named outside investor whose published view on Hon Hai I could check against a filing, so I quote none.

Against reality

If the cloud and networking line reaches NT$7.1T over the next twelve months, grows 18% after that, and the operating margin stays between 3.3% and 3.65%, the share is worth my base of NT$203. The price is above it. To reach NT$245.50 the margin has to be 4.0% from year six on, or the cloud growth 2.0 times my base in every later year. Today’s numbers are close on revenue and short on margin: the third quarter was NT$3,027B, up 47%, and the June quarter’s operating margin was 3.75%.⁶² Cash is negative, and the base case keeps it negative for another year.

What a 10x would take

A tenfold rise from NT$245.50 is NT$2,455 a share and a market value of NT$34.6T, about US$1.06 trillion C (calculated). At today’s 16.1 times earnings it needs net income of NT$2.14T, ten times the last twelve months C (calculated). At 2025’s net margin of 2.3% that is revenue of NT$92T, 10 times today’s, or 26% a year for ten years. At the best net margin of the last ten years, 3.4%, and a multiple of 25, it needs net income of NT$1.38T and revenue of NT$41T, or 16% a year for ten years C (calculated).⁷ Revenue grew 7% a year from 2016 to 2025.⁷ NT$41T is about US$1.2 trillion and NT$92T is US$2.8 trillion, against revenue of US$797.5B at Amazon and US$713.2B at Walmart, the two largest on this site.²⁹³⁰³ I did not compile a base rate for companies this size. I think a tenfold move is not plausible within ten years.

Where I come down

Profit moved me more than revenue: operating profit grew 65% in the half on 35% more sales. Revenue alone did not, because a business that keeps 3.5% of it turns most of a gain into working capital. The price, NT$245.50, is 21% above my base of NT$203; it is what my base would be worth at a 9.3% discount rate, with none of the 1.3 points I add for Taiwan and the single customer. Two things would raise my values: an operating margin that stays above 3.7% through a quarter with the new rack generation shipping, and operating cash flow turning positive while revenue still grows.

VIIIRisks, and what would change my view

Risks

Margin. The company keeps about 3.5% of sales. A 0.5 point change is a 14% change in profit and about NT$42 a share in my base E (estimate). The new rack generation has higher component content and the company says it expects mix to weigh on margins in the second half.¹²

The AI cycle. Cloud and networking is 50.5% of revenue and the company’s growth case depends on a few large buyers continuing to spend. Their capital spending is the company’s own estimate of the market, about US$800 billion this year.¹² If it slows, revenue falls and the margin falls with it; my bear case puts the value at NT$137.

Working capital and credit. Receivables of NT$1,385B and inventory of NT$1,370B are 49% of total assets C (calculated).² A customer that fails or cancels leaves Hon Hai holding parts. The company sells part of its receivables to banks, and net cash at the end of 2025 was lifted by it.¹⁷

Customer and supplier concentration. One customer is 44% of sales and one supplier is 37% of purchases.⁴¹¹ The customers can adjust orders and delivery dates under their contracts.¹¹

Structure. Four-fifths of operating income comes from FII, which is listed in Shanghai and in which outside holders have about 16% of the profit.² The group’s tax rate has risen to 26% with the global minimum tax and the repatriation of profit.¹⁹ The Lordstown estate is suing Hon Hai and others for breach of contract and other claims; the annual report says the company will reject the plaintiff’s appeal.¹⁹

Electric vehicles and the side bets. The company gives no revenue for them. In March the chairman said the outsourcing it expected had come later than planned.¹⁷

Exchange rate. The company sells mainly in US dollars and reports in New Taiwan dollars. The NT dollar’s rise in 2025 weighed on gross margin.¹⁷

What would change my view

I would lower my values if the operating margin fell below 3.3% for two quarters with revenue still rising, if operating cash flow stayed negative into 2027, if the share of profit going to minority holders rose above 16%, or if the company raised equity. I would raise them if the operating margin held above 3.7% with the new racks shipping, or if the company showed a margin by product that proved AI racks earn more than the group average. The next evidence is the 12 November third-quarter call.¹⁸

Geopolitical risk, and what it means for the value

Hon Hai depends on three places and on one policy. Taiwan. Hon Hai buys tens of billions of dollars of chips a year, and on the second-quarter call it said next year’s rack volume depends on how much advanced packaging capacity the chip makers get.¹¹¹² I did not open a document that says where all of its chips are made; the best known supplier, TSMC, is in Taiwan.²⁰ Taiwan holds only 8% of Hon Hai’s long-lived assets, so a conflict would hit the supply of chips more than Hon Hai’s own plants.⁴ China. China is 4.6% of revenue by customer location but 42% of long-lived assets, NT$281B, and the company that earns four-fifths of the operating income is listed there.⁴² The United States. It is 45% of revenue and the source of the tariffs. Management says it is expanding plants in Texas, Wisconsin, Ohio, California, Mexico, Vietnam and Taiwan to produce near customers.¹²⁴ The annual report says tariffs and geopolitics are now long-term drivers of supply chains, and that the company’s global footprint lets it move work between countries.¹⁵

I price three scenarios by changing my base case and nothing else. Tariffs and a tougher China policy are the likeliest, because they are already in place; Washington decides, and I would not be surprised if they last. I take 0.5 point off the margin for four years. A cut-off of China-based output is less likely, perhaps one chance in ten over a decade by my guess; Beijing or Washington decides. A blockade or quarantine of Taiwan is the worst case. I have no better basis for its odds than a guess of about one in twenty over ten years, with a wide band, because nothing I opened gives odds and the risk has been in Taiwan’s market for decades without happening. The table shows what each does to my base value. A war that destroyed assets is outside the model.

What is in the price? Almost none of it. My discount rate has 1.3 points for Taiwan and the largest customer. At the plain CAPM rate of 9.2%, my base is worth NT$248, within 1% of the price E (estimate). So the market price is what my base would be if these risks were worth nothing. That is a statement about my model, not about what the market thinks.