Hua Hong Semiconductor: The Specialty Foundry Champion of China

I. Introduction & Episode Roadmap

In the summer of 2026, a semiconductor executive stood on a call with global analysts and said something that would have sounded absurd for most of the previous decade. Orders, he told them, were running at 1.5 to 2 times the company's available capacity, across microcontrollers, memory, and power-management chips. The fabs were full. Prices were rising. The problem was no longer finding customers — it was finding wafers.¹

The executive was 白鹏 Bai Peng, chairman and president of a company that most Western investors have never heard of and that, until recently, made almost nothing anyone would call exciting. No 2-nanometre logic. No high-bandwidth memory. No graphics processors. Instead: power transistors for air conditioners and solar inverters, embedded flash memory for smart cards and car door modules, analog chips that step 12 volts down to 1.8 volts inside a phone charger. The silicon equivalent of plumbing.

That company is 华虹半导体 Hua Hong Semiconductor — since 2026 formally renamed 华虹宏力半导体 Hua Hong Grace Semiconductor Limited — dual-listed as 688347.SS in Shanghai and 1347.HK in Hong Kong.² In the year ended December 2025, it produced $2.40 billion of revenue at an 11.8% gross margin, having run its fabs at an average capacity utilization rate of 106.1% — a figure above 100% because the denominator is nameplate capacity, and the plants ran harder than nameplate.³ It is, by revenue, China's second-largest wafer foundry behind 中芯国际 SMIC, and the sixth-largest foundry on Earth with roughly 2.5% of a global market in which 台积电 TSMC holds about 72%.⁴

Hold those two numbers next to each other, because the entire investment question lives in the gap. TSMC earns gross margins near 60%. Hua Hong, at the top of its last cycle, earned 34%. At the bottom, in a single quarter, it earned 4%. Same industry. Same customers, sometimes. Radically different economics.

The core thesis, and the reason to interrogate it

The story Hua Hong tells about itself — and the story most sell-side notes repeat — is a counter-positioning story. While TSMC and SMIC burn tens of billions chasing extreme ultraviolet lithography down to 3nm and below, Hua Hong deliberately declined the race. It planted itself in what the Chinese industry calls 特色工艺 specialty process technology: mature geometries from 65nm out to 350nm and beyond, where the competitive variable is not transistor density but device physics — how much voltage a structure can hold, how many times a memory cell can be rewritten, how little current leaks when a switch is off.

The economics are genuinely different, and it is worth understanding why before deciding whether they are defensible. In leading-edge logic, the product is transistors per square millimetre, and the way a foundry wins is by buying the newest lithography tool before its competitor does. It is a capital arms race with a two-year clock. In specialty processes, the product is a behaviour — a transistor that survives 700 volts, a flash cell that holds its charge for twenty years at 125°C in an engine bay. Getting there is less about the tool and more about the recipe: the sequence of implants, anneals, and etches, tuned over years, that produces a device that behaves the same way on wafer number one and wafer number ten million. Old equipment, fully depreciated, can make those chips perfectly well. That is the source of the "legacy fab as cash cow" thesis.

It is also, as subsequent financial results show, only half true — and Hua Hong's own financial record over the last four years is the sharpest available evidence against the strong version of the claim.

The questions this story has to answer

Can a specialty foundry hold price when China is flooding the mature-node market? The bull case rests on process differentiation. The bear case observes that Chinese domestic capacity at 28nm–180nm has been expanding faster than anywhere in history, and that in the last downcycle Hua Hong's margins did not behave like those of a differentiated supplier. They behaved like those of a commodity one.

Did the leap from 8-inch to 12-inch wafers create value or destroy it? Hua Hong spent the better part of a decade and billions of dollars moving its specialty recipes onto 300mm wafers in 无锡 Wuxi. By the first quarter of 2026, 12-inch wafers generated 62.7% of revenue.⁵ But the same transition buried the income statement under depreciation, and consolidated return on equity sat at 0.9% for two consecutive years.

What does the new management actually change? Bai Peng arrived from Intel by way of a Chinese start-up foundry, took the president's chair on 1 January 2025, and by early 2026 held the chairmanship as well.⁶ He inherited a company at the bottom of its margin cycle and midway through the largest capital programme in its history.

And what happens now that Washington has noticed? In April 2026, the US Commerce Department told Lam Research, Applied Materials, and KLA to stop shipping certain tools to two Hua Hong sites, following reports that the group was pushing toward 7nm production with support from a Huawei-linked equipment maker.⁷ A company whose entire strategic identity was "we do not compete at the leading edge" is now being regulated as though it does.

The roadmap

The story runs from a 1996 five-year-plan project and a Japanese joint venture that lost money making commodity memory; through a 2011 merger of two subscale Shanghai fabs that never should have both existed; a modest 2014 Hong Kong listing priced near the bottom of its range; a $3 billion Shanghai STAR Market offering in 2023 that funded a second Wuxi mega-fab; the margin collapse that immediately followed; and finally the 2026 recovery, the acquisition of a sister company from its own parent, and the arrival of American export controls at the door of China's most deliberately unglamorous chipmaker.

It begins, as so much of the Chinese semiconductor industry does, with a number: 909.

---

↑ Back to Top

II. The Genesis: Project 909 & The NEC Partnership (1996–2003)

Chinese industrial projects are often named for the month they receive approval. Project 908, sanctioned in August 1990, became an early cautionary tale: construction dragged on so long that its manufacturing technology was obsolete before the fab opened. Project 909, approved in September 1996 under China's Ninth Five-Year Plan, was the state's second attempt, designed to fix those flaws with faster execution, deeper funding, and a foreign partner capable of running a commercial fab.[^8]

The government established 上海华虹微电子 Shanghai Hua Hong Microelectronics in 1996 as the vehicle for the project, backed by the Ministry of Electronics Industry and the Shanghai municipal government.[^8] The name 华虹—"China Rainbow"—reflected national ambition. It functioned less like a conventional business plan and more like a policy mission with a corporate balance sheet attached.

The Japanese partner

In 1997, Shanghai Hua Hong signed a joint venture with NEC Corporation to form 上海华虹NEC电子 Shanghai Hua Hong NEC Electronics, known thereafter as HHNEC. NEC brought what China could not buy domestically: 200mm wafer processing at 0.35-micron geometries, along with the operational knowledge required to manage a production line. Chinese engineers were sent to Japan for training, and the fab, in Shanghai's 张江高科技园区 Zhangjiang Hi-Tech Park, was scheduled to begin manufacturing in February 1999.⁸

The product choice defined the company's next decade. HHNEC focused its initial output on 64-megabit DRAM—commodity memory, one of the most cyclical and scale-dependent products in the semiconductor industry.⁸ The strategic rationale made sense at the time: memory offered a high-volume path to master fab operations. Because memory specifications were standardized, engineers could focus on learning process control and improving wafer yields without simultaneously managing custom design requests. NEC made DRAM and could transfer that expertise directly.

The commercial reality, however, was unforgiving. By 1999, the DRAM market had already consolidated into a fierce battle among a handful of global manufacturers operating at massive scale. A single 200mm fab in Shanghai processing a few thousand wafers per month was not a real competitor on unit economics; it was a rounding error burdened with heavy debt.

The reckoning

For a brief window, the timing looked inspired. HHNEC caught the tail of the 1999–2000 tech upswing and reported early profitability. But when the tech bust arrived, DRAM prices crashed through 2001 and 2002. A company with a national mission quickly learned what every memory maker eventually discovers: in a commodity downturn, the marginal producer with the smallest volume absorbs the deepest pain. While the listed entity's filings do not disclose the full extent of HHNEC's accumulated losses during the DRAM years, the strategic verdict was unambiguous from what followed.

In 2003, HHNEC abandoned commodity memory entirely and converted the fab into a pure-play foundry, manufacturing chips based on external customer designs rather than selling its own product.[^8] It was the first genuine inflection point in the company's history—a pivot forced by commercial reality rather than strategic choice.

Testing the claim: did state backing guarantee success?

A persistent narrative among commentators holds that Chinese state industrial policy is an irresistible force: identify a sector, appoint a champion, fund the enterprise, and win market share. Hua Hong's founding decade provides strong evidence against that oversimplification, drawn directly from the company's own operating history.

Project 909 had every advantage the state-led model promises: cabinet-level political sponsorship, a world-class foreign technology partner, municipal land, discounted power, and patient capital that private markets would not have provided. Yet it lost money for years because none of those advantages altered the core metric of commodity memory competition: cost per bit at scale. Samsung and what is now SK하이닉스 SK Hynix possessed that cost advantage; HHNEC did not. NEC, meanwhile, progressively pulled back from the venture as its own memory division faltered—exiting DRAM into the Elpida joint venture in 1999 and retreating from the segment for the rest of the decade.

The broader lesson remains relevant for evaluating current industry dynamics. State capital can fund a fab, but it cannot buy a structural cost position. Where Chinese state backing has produced durable businesses, it has done so by subsidizing the fixed costs of entering an industry where incumbents also relied on fixed-cost advantages—not by overcoming an ongoing variable-cost disadvantage. That distinction provides a key framework for evaluating every subsequent capital allocation decision in this story, including the major investments unfolding in Wuxi.

The 2003 pivot to foundry work marked the point where HHNEC stopped fighting market economics. As a foundry, HHNEC no longer had to compete on cost per bit against Korean memory giants; instead, it competed on service, yield, and process customization for Chinese chip designers. That was a fight the company could plausibly win.

It also created an immediate competitive dynamic, because a few kilometers away in the same Shanghai technology park, another state-backed entrant was building the exact same business.

↑ Back to Top

III. The Parallel Story: Grace Semiconductor & The 2011 Merger

The founding of 宏力半导体 Grace Semiconductor Manufacturing Corporation in 2000 read less like an industrial plan than a political novel. Its two principal figures were Winston Wong, son of Formosa Plastics founder Y.C. Wang and heir to one of Taiwan's prominent industrial fortunes, and 江绵恒 Jiang Mianheng, son of then-President 江泽民 Jiang Zemin and a physicist who had become a key technology investor in Shanghai. A Taiwanese tycoon's son and a Chinese president's son built a semiconductor company together in Shanghai at a moment when cross-strait relations were particularly delicate.⁹

In November 2000, Grace broke ground on a $1.6 billion, 8-inch sub-micron fab in the same Zhangjiang Hi-Tech Park in Pudong where HHNEC already operated, with construction financed by state-owned lenders including Bank of China and China Construction Bank, and production scheduled for 2002.⁹ Wong served as president; Jiang was his principal partner.⁹

The setup was striking: two state-adjacent foundries operating in the same technology park, in the same city, building the same wafer size, and chasing the same domestic Chinese chip designers—in an industry where the margin between profit and loss is determined by a few percentage points of capacity utilization.

The arithmetic of duplication

This structure highlights a pattern that recurs across Chinese industrial policy—in solar, in batteries, in electric vehicles, and in mature-node foundry capacity.

A wafer fab carries roughly 70% to 80% in fixed costs, driven by equipment depreciation, cleanroom utilities, and a specialized workforce that an operator cannot easily furlough. Because the marginal cost of producing an additional wafer is far below average cost, falling utilization creates a strong incentive for operators to cut prices to keep production lines full. When two subscale fabs running identical technology target the same accounts, they do not split the market; they compete margins down to the floor and eventually turn to their respective sponsors for support.

Through the 2000s, that dynamic played out in Zhangjiang. Both companies invested heavily, yet neither achieved the operational scale needed to generate attractive returns. By 2011, the combined annual sales of the two entities were projected at about $600 million with roughly $100 million in profit—a modest return across two separate fab organizations burdened by duplicated management, duplicated research and development, and duplicated overhead.¹⁰

December 2011: the merger that took a decade too long

In mid-September 2011, the parties signed a merger agreement, closing the deal by late December. HHNEC acquired a 51% stake in the combined entity, named 华虹宏力 Hua Hong Grace Semiconductor Manufacturing Corporation—a name the listed parent company would formally adopt for itself fifteen years later.¹⁰¹¹

The combined business held approximately 145,000 200mm wafers per month in capacity, with manufacturing capabilities spanning the 130nm to 90nm process nodes. The deal vaulted the group to roughly fifth place among global wafer foundries by revenue, ahead of TowerJazz and behind SMIC.¹⁰ It also inherited a joint interest in a 300mm fab venture, Shanghai Huali Microelectronics, which was then targeting 65nm and 55nm production at a modest 10,000 wafers per month.¹⁰ That entity would re-emerge at the center of corporate developments in 2026 with an 8-billion-yuan transaction attached.

The strategic rationale was articulated at the time with unusual clarity by Gartner analyst Samuel Wang, who observed that Chinese foundries should build their growth on domestic chip design houses rather than race TSMC, Samsung, and GlobalFoundries at advanced nodes.¹⁰ That observation effectively outlined Hua Hong's operational roadmap for the subsequent fifteen years.

Testing the claim: did the merger deliver clean synergies?

Corporate consolidation narratives routinely promise swift synergies, but the period from 2011 to 2014 delivered a more nuanced reality.

Integrating semiconductor fabs is far more complex than merging software operations. Each fab relies on its own process design kit—the proprietary suite of models, design rules, and verification tools used by chip designers. A customer whose chip is engineered for Fab 1's kit cannot simply shift production to Fab 2's line without re-characterizing and re-qualifying the silicon, a process that can take a year or longer for automotive and industrial components. Consequently, while the merger yielded immediate administrative overhead savings, manufacturing synergies took years to materialize.

The capital markets provided a clear evaluation of this progress in October 2014, when Hua Hong Semiconductor listed on the Hong Kong Stock Exchange. Led by Goldman Sachs, the offering was priced at HK$11.25 per share—near the bottom of its indicated range—raising HK$2.57 billion (roughly $330 million before the over-allotment option) across 228.7 million shares.¹² The modest proceeds and bottom-of-the-range pricing reflected investor skepticism, showing that while the market was willing to fund the company, it did not pay a premium for the consolidation story.

Ultimately, the merger did not generate an immediate step-change in profitability, and operational benefits emerged only gradually. However, it achieved a vital strategic objective: it halted price destruction between two state-backed Shanghai fabs and created an entity with sufficient scale to justify future capital expansion.

↑ Back to Top

IV. Strategic Pivot: Escaping Standard Logic to Master Specialty Process Technology

Every foundry that is not TSMC eventually faces the same meeting. Someone puts up a slide showing the capital cost of the next node, and someone else points out that at the company's revenue base, one lithography scanner costs more than a year of operating cash flow. The room then divides into people who want to fight anyway and people who want to find a different fight.

Hua Hong found a different fight. The choice was not announced as a grand strategic doctrine; it emerged from the accumulated logic of what the company could actually do well, and it hardened into an explicit corporate identity — "8-inch plus 12-inch" combined with "specialty IC plus power discrete," the phrasing the company still uses in its own results releases.³

What "specialty process" actually means

Strip away the acronyms and there are four product families, and each one is a different kind of engineering problem.

Embedded non-volatile memory is the ability to build a rewritable memory cell directly onto the same die as a small processor. That is what makes a microcontroller a microcontroller: a chip that both computes and remembers when the power goes off. It is what sits inside a bank card, a passport chip, an electricity meter, and the module that controls a car's power windows. The engineering difficulty is not the transistor — it is making a memory cell that will still hold its charge in twenty years, at high temperature, after a hundred thousand rewrites. Hua Hong runs these processes from around 0.13-micron down to 90nm and 55nm.

Power discretes are the muscles rather than the brain: single-purpose switches that turn large currents on and off. Insulated-gate bipolar transistors handle the very high voltages inside electric-vehicle inverters and solar installations. Super-junction and trench MOSFETs handle the medium-voltage work inside chargers, appliances and power supplies. The design challenge here is vertical — the current flows through the thickness of the wafer, not across its surface — which is why these devices care almost nothing about the lithographic linewidths that dominate logic scaling and enormously about implant depth, epitaxial layer quality and edge-termination geometry.

Analog and power management covers the chips that convert, regulate and distribute voltage, built on what the industry calls BCD platforms — bipolar, CMOS and DMOS transistors coexisting on a single die so that one chip can contain precision analog, digital control and high-voltage switching. Every phone charger, every server power stage, every laptop has several.

Logic, radio-frequency and image sensors round out the portfolio — the closest thing Hua Hong has to conventional foundry work.

The reason these processes behave differently from leading-edge logic is worth stating plainly, because it is the load-bearing beam of the bull case. In logic, the recipe is broadly public and the tool is the moat; whoever buys the newest scanner wins. In specialty, the tool is broadly available and the recipe is the moat. Getting a 700-volt super-junction MOSFET to yield consistently is the accumulated product of hundreds of process experiments, and it does not transfer with a purchase order. Equipment written down over seven years keeps producing revenue for fifteen or twenty. Fully depreciated 8-inch lines can, in a good market, throw off cash with almost no capital consumption.

That is the theory. Now the evidence.

Testing the claim: does specialty process confer durable pricing power?

The strong version of the thesis says specialty differentiation insulates Hua Hong from the commodity price cycle. The company's own income statement between 2022 and 2024 rejects that version outright.

In the third quarter of 2022, Hua Hong reported a gross margin of 37.2% on record quarterly revenue of $629.9 million, and finished that year with a full-year gross margin of 34.1%, net profit attributable to shareholders of $449.9 million, and a return on equity of 15.2%.[^14] Fifteen months later, in the fourth quarter of 2023, gross margin was 4.0%. Not 24%. Four.¹³

The mechanism is fully visible in the disclosures and it was not a mix problem. Capacity utilisation fell to 84.1% in that quarter — the fab was running at five-sixths of nameplate while carrying essentially all of its fixed cost.¹³ For the full year 2023, revenue fell to $2,286 million and gross margin to 21.3%; in 2024 revenue fell again to $2,004 million, down 12.3%, with gross margin of 10.2%, which the company attributed explicitly to lower average selling prices and higher depreciation.¹³ Return on equity went from 15.2% in 2022 to 6.3% in 2023 to 0.9% in 2024.[^14]¹³

Notice what that sequence says. In 2024, wafer shipments actually rose 10.8% and utilisation recovered to 99.5% — the fabs were essentially full — and the company still earned a 10.2% gross margin and near-zero return on equity.¹³ Full fabs and terrible returns simultaneously is the signature of price competition, not demand weakness. Chinese mature-node capacity from peers including 华润微 CR Micro, 士兰微 Silan Micro and 晶合集成 Nexchip was arriving at the same time, and the market cleared on price.

The honest conclusion is a narrowed one rather than a rejection. Specialty process technology is real, and it is why Hua Hong's customers came back rather than leaving permanently — the company retained volume throughout, which a genuinely undifferentiated supplier would not have. But process differentiation set the floor on volume, not the price. On price, Hua Hong behaved like a capital-intensive commodity producer with high fixed costs, because in a downcycle that is what it is. Investors should hold the specialty thesis as a claim about share and utilisation resilience, not as a claim about margin stability. The KPI that tests it is straightforward: whether gross margin in the next downcycle falls to single digits again, or holds meaningfully above the 2023–2024 trough.

The domestic substitution story, and what the numbers say about it

The most repeated bull narrative about Hua Hong is that it is the silicon behind China's electric-vehicle and green-energy build-out — the domestic foundry partner for the IGBTs in 比亚迪 BYD inverters and the microcontrollers in Chinese car electronics, riding 国产替代 domestic substitution as Chinese OEMs designed out Infineon and STMicroelectronics.

There is real substance here. Hua Hong describes its quality system as meeting automotive requirements, and its European revenue growth in recent quarters has been driven partly by a European customer's "China for China" localisation strategy.¹⁴ Automotive and industrial demand is genuinely a growth vector.

But the disclosure that matters most sits in the company's own quarterly release, and it complicates the story considerably. In the first quarter of 2026, revenue by end market broke out as consumer electronics $444.4 million, or 67.2% of the total; industrial and automotive $143.2 million, or 21.7%; communications 9.5%; and computing 1.6%.⁵ Two-thirds of Hua Hong's revenue comes from consumer electronics. Industrial and automotive is a fifth.

That reframes the investment case. Hua Hong is not primarily an EV supply-chain company that happens to make consumer parts; it is primarily a consumer-electronics foundry with a meaningful and growing industrial and automotive business attached. The distinction matters because consumer demand is shorter-cycle, more price-elastic and more exposed to the Chinese domestic consumption environment than automotive design-ins, which are stickier but slower. Anyone underwriting Hua Hong as a pure play on Chinese EV growth is underwriting a business that does not exist in the reported segment data.

The same release shows how much of the company still lives on genuinely old technology: 31.5% of first-quarter 2026 revenue came from nodes of 0.35 micron and above, and only 27.1% from 65nm and below.⁵ The legacy 8-inch base is not a rounding error. It is the largest single node bucket in the company.

Which brings us to the decision that was supposed to change all of that.

---

↑ Back to Top

V. The Wuxi 12-Inch Expansion & The $3B STAR Market Listing (2018–2023)

There is a physical constraint at the heart of this transition, best understood by comparing two silicon discs. An 8-inch wafer is 200 millimetres across, while a 12-inch wafer measures 300 millimetres. Because area scales with the square of the radius, the larger disc carries about 2.25 times as much usable surface. Because key processing steps—lithography, etching, ion implantation—are applied to the entire wafer at once, the cost per step increases far less than 2.25 times. That mathematical gap forms the economic foundation for manufacturing on larger wafers, driving format shifts across the semiconductor industry for half a century.

The catch is that a 12-inch production line is not merely a scaled-up 8-inch line. It requires different tools, automated handling, and—critically for a specialty foundry—redeveloped process recipes. A super-junction MOSFET recipe that yields reliably on a 200mm line cannot simply be ported. Moving to 300mm required Hua Hong to rebuild its core intellectual property on a new substrate, costing billions of dollars before customers had committed to buying the output.

Wuxi, and the mechanics of Chinese industrial finance

Hua Hong did not finance this shift alone. The structure it adopted serves as a blueprint for how Chinese semiconductor capacity has been built for a decade.

In 2017, the company established a joint venture, 华虹无锡 Hua Hong Wuxi, combining the Shanghai operating entity, the listed parent, the state-backed 国家集成电路产业投资基金 China Integrated Circuit Industry Investment Fund (known as the Big Fund), and a Wuxi municipal investment vehicle.[^17] The resulting facility—which the company described as the world's first 12-inch fab dedicated to power semiconductors—was processing roughly 65,000 wafers per month by the time of its 2023 domestic equity offering.³¹⁵

The corporate structure matters more than the policy rhetoric. By housing the fab inside a partially owned subsidiary rather than on the parent balance sheet, Hua Hong shared both the capital burden and the operational losses. This arrangement remains a defining feature of the group's reported financials.

In 2024, the consolidated group recorded a net loss of $140.4 million. Yet net profit attributable to shareholders of the parent remained positive at $58.1 million, because $198.5 million in losses were allocated to non-controlling interests—the state and municipal co-investors in the Wuxi ventures.¹³ A similar pattern occurred in 2025: a group net loss of $110.8 million yielded $54.9 million in profit attributable to parent shareholders, after $165.7 million in losses were absorbed by minority partners.³

That accounting distinction cuts both ways. It shows that the headline net profit figure displayed on stock screeners does not represent the underlying profitability of the entire enterprise, as consolidated operations were loss-making for two consecutive years. Conversely, it demonstrates that minority state partners absorbed the bulk of the ramp-up losses. While this provides a structural advantage in capital cost, it cautions against valuing the company solely on parent-attributable earnings.

August 2023: the largest cheque in the company's history

By 2022, Hua Hong sought to construct a second 12-inch line in Wuxi, requiring more capital than the Hong Kong market was likely to provide. It turned to Shanghai's STAR Market, a domestic board where mainland institutional and retail investors have paid higher valuations for semiconductor issuers than international markets.

On 7 August 2023, Hua Hong listed A-shares in Shanghai under the ticker 688347. The company sold 407.75 million shares at RMB 52 each, raising RMB 21.2 billion—roughly $2.9 billion—exceeding its original target of RMB 18 billion. Mainland retail investors oversubscribed their allocation 142 times, while the Big Fund took RMB 3 billion. The shares opened at RMB 58.88, rising 13% on debut. The transaction marked the second-largest initial public offering globally in 2023 and the third-largest in STAR Market history.¹⁵¹⁶

The contrast with the company's 2014 Hong Kong offering was sharp. Nine years earlier, Hua Hong had priced near the bottom of its range to raise approximately $330 million. In 2023, it raised nearly nine times as much at a substantial premium to its Hong Kong-listed H-shares.¹⁵ While the underlying wafer manufacturing business had expanded, the investor base and geopolitical context had transformed: following the October 2022 US export controls, domestic mature-node capacity was viewed by mainland investors as strategic national infrastructure.

Roughly 70% of the proceeds were designated for a second 12-inch line in Wuxi, with the remainder allocated to upgrading 8-inch fabs, specialty process research, and working capital.¹⁵ That second line—Hua Hong Manufacturing (Wuxi), known as Fab 9—was announced in January 2023 as a $6.7 billion project financed with $4.02 billion in equity and $2.68 billion in debt. Hua Hong and its subsidiary retained a 51% stake, the Big Fund contributed $1.165 billion, and a Wuxi municipal entity provided the balance, targeting a monthly capacity of 83,000 wafers across 65nm, 55nm, and 40nm process nodes.¹⁷

Testing the claim: did greenfield 12-inch expansion earn its capital?

Management maintained that increased scale alongside state co-investment would drive earnings accretion. However, the financial record through the initial phase of the expansion leaves the strategic outcome unproven.

The timing proved challenging. Hua Hong committed to the $6.7 billion greenfield investment in January 2023 and priced its A-share offering in August 2023, only for gross margin to bottom at 4.0% four months later.¹³ The company then spent 2024 and 2025 absorbing heavy depreciation from new capacity during an industry price downturn, resulting in a consolidated return on equity of 0.9% in both years.¹³³ Depreciation on 12-inch assets continued to rise, with management guiding to approximately $210 million in 12-inch depreciation for the second half of 2026 alone, compared to about $55 million for the entire 8-inch fleet.¹ This disparity—new capacity generating four times the depreciation of the legacy 8-inch footprint—weighed heavily on reported returns.

Capital spending remained elevated. Capital expenditures reached $924.9 million in the first quarter of 2026, with $886.1 million directed to 12-inch facilities, against quarterly revenue of $660.9 million.⁵ A business spending more on capital additions than it generates in revenue is reinvesting aggressively rather than harvesting cash.

Comparing Hua Hong to mature-node peers like 世界先进 Vanguard International Semiconductor—which expanded primarily by purchasing fully depreciated secondhand fabs—highlights contrasting strategic choices. Buying existing fabs secures depreciated tools with lower capital risk, whereas greenfield construction establishes modern facilities capable of supporting long-term node transitions. Whether Hua Hong's approach succeeds depends on whether domestic demand for specialty processes expands structurally or remains cyclical. The greenfield strategy front-loaded fixed costs, leaving shareholders with low returns while waiting for the asset base to mature.

The primary offset remains the co-investment model: state-linked minority partners absorbed the majority of the early losses, shielding the parent company from the full earnings impact of the expansion.

The eventual test of the strategy will be reflected in operational metrics. If 12-inch revenue continues to grow and gross margins recover toward the mid-20% range while capacity utilization stays above 100%, the expansion will have generated economic value. If margins remain constrained in the mid-teens as Fab 9 depreciation fully hits the income statement, the company will have replaced a smaller, higher-margin 8-inch business with a larger, lower-return 12-inch footprint.

↑ Back to Top

VI. Current Management & Modern Operations: Bai Peng's New Era

Consider the job description: a state-controlled Chinese foundry, navigating the trough of its sharpest margin contraction as a public company while mid-way through a multi-billion-dollar capacity expansion, requires a chief executive. The conventional appointment would be a party-track industrial administrator. Instead, on the final day of 2024, the board announced that a 62-year-old physicist who spent the majority of his career at Intel would assume the presidency on 1 January 2025.⁶

Who Bai Peng is

The biography disclosed in regulatory filings is unusually detailed. Bai studied at Peking University, earned a bachelor's degree in physics from the University of Bucharest in 1985, and completed a doctorate in physics at Rensselaer Polytechnic Institute in 1991.⁶

At Intel, Bai advanced through a sequence of technical and operational roles: process integration engineer, process integration manager, director of yield engineering, director of research and development, vice president, and ultimately corporate vice president.⁶ Yield engineering is the core operational discipline that turns silicon recipes into profitable production by isolating yield-killing defects and tightening process variability. It is an execution role focused directly on manufacturing efficiency.

Bai left Intel to become chief executive of Chinese start-up foundry 榮芯半導體 Rong Semiconductor in September 2022, before joining Hua Hong.⁶ His appointment came with a three-year service contract, and regulatory filings recorded that he held no shares in the company at the time of the announcement.⁶

Minimal management equity ownership is standard for a Chinese state-owned enterprise, but it means traditional equity-based alignment with public shareholders is absent. Executive incentives depend on state performance mandates and board remuneration committees rather than share price performance.

Governance initially featured split leadership: 唐均君 Tang Junjun transitioned from president to chairman on 31 December 2024, while Bai assumed day-to-day operational control.⁶ Tang later departed the board, and by the 12 February 2026 earnings release, Bai served as both chairman and president.³ A July 2026 board notification confirmed his combined titles.² While consolidating the chairman and chief executive roles reduces independent board oversight—a frequent concern for international institutional investors—its practical impact is tempered by direct state shareholder oversight.

What the earnings calls reveal

Management credibility is best evaluated by consistency over time as market conditions evolve. Hua Hong's investor communications across this transition provide a clear baseline.

In his debut on the fourth-quarter 2024 earnings call in February 2025, Bai offered a direct assessment. He characterized 2024 as a challenging year, acknowledged soft demand for mid-to-high-end power devices, and cited intense market competition rather than pointing solely to macroeconomic headwinds.¹³

Throughout 2025, executive commentary focused on steady operational recovery through product mix optimization, cost controls, and fab efficiency. On the fourth-quarter 2025 call, Bai reported that the initial phase of Fab 9 had exceeded construction timelines and that the acquisition of the Shanghai 12-inch facility remained on track.³ By the first quarter of 2026, management commentary highlighted artificial intelligence demand while acknowledging persistent global supply chain uncertainties—a cautious reference to tightening trade and export restrictions.⁵

The second-quarter 2026 call highlighted a contrast between optimistic prepared remarks and detailed disclosures during analyst questioning. Chief Financial Officer Daniel Wang attributed quarterly revenue growth roughly 60% to higher average selling prices and 40% to increased wafer volume, explicitly identifying market tightness—rather than a structural shift toward higher-value products—as the primary driver of pricing gains.¹ This distinction clarifies that recent margin expansion stems from a cyclical supply imbalance rather than a permanent product re-rating.

When questioned on production flexibility across product categories, management noted that while lines for logic, image sensors, and display drivers offer significant interchangeability, converting lines for embedded memory and BCD analog platforms remains technically constrained.¹ Regarding long-term demand visibility, management expressed confidence through 2027 but explicitly declined to project conditions into 2028.¹

Management's refusal to project beyond 2027 places a firm boundary on the current upcycle. Hypotheses viewing the 2026 price recovery as a permanent structural re-rating of specialty foundry economics run counter to management's own limited forward visibility.

Operational priorities under Bai reflect a measured shift toward product mix and margin optimization. Full-year 2025 results highlighted embedded non-volatile memory and power management platforms as primary growth drivers rather than legacy discrete components.³ While disclosures indicate slower revenue growth in power discretes alongside faster expansion in analog and memory, the company has not formally disavowed price competition in legacy markets.

↑ Back to Top

VII. Segment Deep Dive, Unit Economics, & The 2023–2024 Capacity Gut-Check

By the middle of 2026, Hua Hong looked like a different company from the one that reported a 4% gross margin at the end of 2023 — and the change is more instructive than the headline recovery suggests, because almost none of it came from the products the bull case was built on.

Where the revenue actually comes from

Full-year 2025 revenue was $2,402.1 million, up 19.9%, with the fourth quarter setting an all-time high of $659.9 million.³ Momentum continued into 2026: $660.9 million in the first quarter, then $717.5 million in the second — another record, up 26.8% year on year and 8.6% sequentially, with gross margin of 16.5% and net profit attributable to shareholders of $38.6 million against $7.95 million a year earlier.⁵¹⁸

The platform mix in the second quarter of 2026 is where the story lives. Embedded non-volatile memory generated $200.1 million, up 41.8%. Standalone non-volatile memory — essentially NOR flash — generated $63.3 million, up 149.3%. Analog and power management contributed $183.1 million, up 13%. Logic and radio-frequency added $83.2 million, up 21.3%. And power discrete, the platform at the centre of the electric-vehicle narrative, produced $182.3 million, up just 9.4%.¹

Sit with that spread. The fastest-growing product line in the company, by a very wide margin, is standalone flash memory — a product with no meaningful specialty differentiation at all, whose price is currently being lifted by the global memory shortage that AI infrastructure demand has created. The second fastest is embedded memory. The slowest is the power discrete franchise that the domestic-substitution thesis is built on.

The honest interpretation is that Hua Hong's 2026 recovery has been driven substantially by a cyclical memory-adjacent windfall and by general mature-node tightness, rather than by the specialty moat asserting itself. Management effectively conceded this on the second-quarter call when it attributed roughly 60% of sequential growth to price and described the price increases as market-tightness-driven rather than mix-driven.¹ That is not a criticism of management — it is a description of what kind of upcycle this is, and it should govern how durable an investor assumes the margin recovery to be.

Geographically, the business remains overwhelmingly domestic: China accounted for $563.7 million, or 78.6% of second-quarter revenue. North America grew 77% to $93.8 million on AI-server-related power products, and Europe grew 90.1% to $28 million.¹ The international growth rates are eye-catching but from small bases, and they are the part of the revenue most exposed to any tightening of US restrictions.

Capacity: what is disclosed, and what is not

Hua Hong does not publish fab-by-fab capacity. What it does publish is a group total expressed in 8-inch equivalents: monthly capacity was 489,000 8-inch-equivalent wafers at the end of the first quarter of 2026, against 413,000 a year earlier — an 18% increase in twelve months.⁵ Wafer shipments rose in step, to 1,453,000 8-inch equivalents in the quarter and 1,540,000 in the second quarter.⁵¹⁸

That growth rate is the operational heart of the business right now. Adding roughly a fifth to capacity in a year, and filling it — utilisation was 99.7% in the first quarter and 102.8% in the second — is the reason revenue is compounding at north of 20% while average selling prices remain far below their 2022 peak.⁵¹⁸

The mix shift to larger wafers is likewise visible and rapid. Twelve-inch wafers generated 61.7% of fourth-quarter 2025 revenue, 62.7% in the first quarter of 2026, and 62.1% in the second quarter, on $445.7 million of 12-inch revenue against $271.8 million from 8-inch.³⁵¹⁸ Two years earlier, in the fourth quarter of 2023, 12-inch had been 44.9%.¹³ The company has genuinely become a 300mm business.

Unit economics, honestly derived

Hua Hong does not disclose average selling prices per wafer, and any specific per-wafer ASP figure circulating in research notes is an estimate rather than a company disclosure. What can be derived from the published numbers is a blended figure: roughly $717.5 million of second-quarter revenue against 1,540,000 8-inch-equivalent wafers works out to about $466 of revenue per 8-inch-equivalent wafer, including a small non-wafer services component.¹⁸¹

That single number carries the whole margin problem. At a 16.5% gross margin, roughly $389 of every $466 is cost of goods sold, of which depreciation is a large and rising share. Management's own guidance for the second half of 2026 — approximately $210 million of depreciation on 12-inch facilities and $55 million on 8-inch — implies annualised depreciation running at well over $500 million against roughly $2.9 billion of annualised revenue.¹ Depreciation alone is consuming something on the order of eighteen cents of every revenue dollar.

This is the clearest possible statement of why the "fully depreciated legacy fab as cash cow" story does not describe today's Hua Hong. It describes the 8-inch business, which is small, stable and profitable. The company as a whole is a heavy new-capacity business in the middle of its depreciation curve. Those are opposite economic characters, and they are consolidated into one income statement.

The margin trajectory, and what it proves

The arc is worth stating once, cleanly, because it is the single most important quantitative fact about this company. Gross margin peaked at 37.2% in the third quarter of 2022 and 34.1% for that full year.[^14] It fell to 21.3% for 2023, bottomed at 4.0% in the fourth quarter of that year, recovered to 10.2% for 2024 and 11.8% for 2025, and reached 16.5% in the second quarter of 2026, with guidance of 16–18% for the third quarter.¹³³¹

Four years into the cycle, with fabs running above nameplate, an unprecedented AI-driven demand surge, prices rising, and 12-inch capacity at scale, the company is earning less than half the gross margin it earned in 2022. That gap is the market's verdict on what happened to specialty mature-node pricing in China, and it will not close simply because utilisation is high — it will close only if pricing normalises further or if the depreciation burden is out-grown.

The related-party deal that changes the arithmetic

Which brings us to the most consequential capital allocation decision of the Bai era, and one the outline could not have anticipated: in December 2025 Hua Hong agreed to buy 华力微电子 Shanghai Huali Microelectronics — the 300mm venture it had held a small stake in since 2010 — from its own parent.¹⁹

The mechanics deserve scrutiny. The appraised value was RMB 8,480 million and the final consideration RMB 8,267,902,153, settled entirely in newly issued A-shares — 190,768,392 of them at RMB 43.34 each — to four vendors: 华虹集团 Huahong Group (63.54%), the Shanghai IC Fund (15.72%), Big Fund II (10.25%) and the Guotou IC Fund (7.98%). Hua Hong already owned 2.5%.¹⁹ The deal adds roughly 38,000 wafers per month of 65nm and 40nm capacity and, as of the second-quarter 2026 call, had received final regulatory approval with closing expected within about a month and consolidation from the third quarter.²⁰¹

Three features of this transaction should interest a sceptical investor.

First, the pricing of the shares. Under PRC rules the issue price was set at not less than 80% of the 120-day average trading price, producing RMB 43.34 — a 44.79% discount to the RMB 78.50 closing price on the date of the acquisition agreement, and a 59.82% discount to the RMB 107.87 close on the date the supplemental agreement was signed.¹⁹ Issuing equity to your controlling shareholder at roughly 40 cents on the market dollar is dilutive to everyone else. The consideration shares represent about 9.89% of enlarged share capital.¹⁹

Second, the choice not to use cash. The company held $4.53 billion of cash and equivalents at 30 June 2026.¹ The board's stated reasoning was that capital expenditure needs and long-term strategic commitments made share issuance preferable.¹⁹ That is a defensible position for a company spending nearly a billion dollars a quarter on fabs — but it is also a decision that transferred value to the parent at a depressed reference price rather than paying market value in cash.

Third — and this is the part that argues for management, not against it — the asset is genuinely good, and there is downside protection. Huali lost RMB 372.3 million in 2023, then earned RMB 521.5 million in 2024 and RMB 514.6 million in just the eight months to August 2025.¹⁹ At the consideration price that is a mid-teens multiple of 2024 earnings and roughly ten times the annualised 2025 run-rate. Crucially, the announcement explains that Huali's equipment has just completed its seven-year accounting depreciation cycle, which is why its book value is low relative to appraisal.¹⁹ Hua Hong is buying, at a moment when its own income statement is buried under new-fab depreciation, a 300mm fab whose depreciation is largely behind it. Huahong Group also agreed to compensate for any impairment of the acquired assets over three financial years, in proportion to its 65.2% share of the consideration.¹⁹

The transaction also required a whitewash waiver — meaning the parent's resulting stake would otherwise have triggered a mandatory general offer under the Hong Kong Takeovers Code — and was classified as a special deal, with the circular to shareholders delayed three times between September and November 2025 before terms were finalised.¹⁹ Alongside it, the company proposed a further non-public issuance of up to RMB 7,556 million of A-shares, not exceeding 30% of enlarged share capital.¹⁹

Netting it out: this is a value-accretive purchase of a well-timed asset, executed through a governance structure that a minority shareholder is entitled to dislike. Both statements are true, and an investor should hold both.

---

↑ Back to Top

VIII. Strategic Moat Analysis: 7 Powers & Porter's 5 Forces

Frameworks are only useful if you let the evidence overrule them. Applied honestly to Hua Hong, Hamilton Helmer's seven powers and Porter's five forces produce a picture that is neither the fortress the bull case describes nor the commodity the bear case assumes — it is a business with one real advantage, several partial ones, and a structural exposure that no amount of process expertise addresses.

Hamilton Helmer's 7 Powers, tested against the record

Process power. This is the strongest claim, and the evidence broadly supports it. Hua Hong's specialty recipes — deep-trench super junction, high-voltage BCD analog, embedded flash qualified to automotive standards, and 12-inch IGBT manufacturing — represent accumulated know-how that cannot be bought. The strongest supporting evidence is behavioural rather than rhetorical: through the 2023–2024 collapse, Hua Hong lost price but did not lose volume. Wafer shipments rose 10.8% in 2024 and 18.5% in 2025 while prices fell.¹³³ Customers stayed. That is what process power looks like when it is real but is not, on its own, sufficient to hold price.

Scale economies. Partial, and improving. The Wuxi 12-inch cluster gives Hua Hong the largest dedicated specialty-process 300mm footprint in mainland China, and the Huali acquisition adds another 300mm site. But scale is relative, and the relevant comparison is unflattering: at roughly 2.5% of global foundry revenue against TSMC's 72%, Hua Hong's R&D and overhead are spread across a base an order of magnitude smaller than the industry leader's.⁴ Within Chinese specialty foundry, the scale advantage is real. Globally, it is not.

Cornered resource. Weak, and frequently overstated. Preferential access to Big Fund equity and municipal land is a genuine funding-cost advantage — OECD-derived analysis has put Chinese state subsidies at roughly 22% of Hua Hong's revenues over the 2014–2018 period, against 3% or less for TSMC, Intel and Samsung.²¹ But subsidised capital is available to every Chinese foundry, including the competitors adding the capacity that crushed pricing. A resource your rivals also have is not cornered; it is an industry condition. And the 2024 disclosure that other income fell partly on decreased government subsidies is a reminder that this input is discretionary and can be withdrawn.¹³

Switching costs. Moderate, and genuinely underrated in automotive and industrial applications. Once a customer's chip is characterised against a specific foundry's process design kit and qualified for automotive use, moving it means re-silicon, re-qualification and re-certification — a process measured in quarters, sometimes years. This is why volumes held during the price war. But note the constraint the mix data imposes: two-thirds of revenue is consumer electronics, where switching costs are far lower and design cycles far shorter.⁵ The moat is strongest in the smallest fifth of the business.

Counter-positioning. Historically genuine, currently under threat, and this is the framework entry that has changed most. For fifteen years, declining the leading-edge race was Hua Hong's defining strategic choice, and it was the right one: it avoided a capital contest the company could not fund. But in April 2026 the US Commerce Department sent "is-informed" letters to Lam Research, Applied Materials and KLA restricting shipments to two Hua Hong sites — Fab 6 in Shanghai, which supports 28nm and 22nm, and a Fab 8a understood to be under construction — on reporting that Huali had begun developing a 7nm process at Fab 6 with support from Huawei-backed equipment maker SiCarrier.⁷ Whatever the operational reality of those reports, the strategic consequence is that Hua Hong is now being treated by regulators as an advanced-node actor. The company has, in effect, lost the protective ambiguity that came from being boring.

Branding and network economies are not meaningful in wafer foundry and do not apply here.

Porter's five forces, with the competitive geography

Threat of new entrants: low to moderate, and lower than it looks. A greenfield 12-inch specialty fab costs billions — Hua Hong's own Fab 9 was budgeted at $6.7 billion, and management has indicated a further Fab 9C programme involving roughly $6 billion of spending over several years.¹⁷¹ The catch is that in China the entrant does not need a normal return; several of the recent entrants are themselves state-backed. Capital cost is a barrier only where capital is priced.

Bargaining power of suppliers: high, and rising. This is the most acute force. Hua Hong depends on tools from Applied Materials, Lam Research, KLA, Tokyo Electron and ASML, and on Chinese suppliers including 北方华创 Naura where domestic substitution is available. Export controls give the supplier side, or rather the supplier side's government, effective veto power over specific expansion plans. Management flagged tightening global equipment supply on the second-quarter 2026 call as a business risk in its own right.¹

Bargaining power of buyers: moderate to high. Hua Hong's customers are Chinese fabless designers who can and do dual-source across SMIC, CR Micro, Nexchip and others. The 2023–2024 price collapse was buyers exercising exactly that power. What limits it is qualification friction in the industrial and automotive tranche.

Threat of substitutes: low. There is no digital replacement for a power transistor or a microcontroller. The nearest thing to a substitute is wide-bandgap material — silicon carbide and gallium nitride displacing silicon in high-voltage applications — which is a genuine long-term consideration for the power discrete franchise, though the same physics that make it a threat also make it an opportunity for a foundry that develops those processes.

Rivalry: high, and the defining force in this industry. Chinese mature-node capacity is projected to rise from about 31% of global mature-node production in 2023 to roughly 39% by 2027.²¹ TrendForce's assessment of Hua Hong's first quarter of 2026 was blunt: modest wafer shipment growth largely offset by ASP declines, leaving revenue broadly stable.⁴ That is a description of a market where volume growth is being competed away.

Myth versus reality

Three consensus statements about Hua Hong deserve correction.

Myth: Hua Hong is a play on Chinese electric vehicles. Reality: industrial and automotive is 21.7% of revenue; consumer electronics is 67.2%.⁵

Myth: legacy fabs make Hua Hong a cash cow. Reality: capital expenditure exceeded revenue in the first quarter of 2026, and depreciation on new 12-inch capacity is running at roughly four times that of the entire 8-inch fleet.⁵¹

Myth: state backing insulates the business. Reality: consolidated return on equity has been 0.9% for two consecutive years, and the group has reported net losses at the consolidated level in both.¹³³

None of these corrections make Hua Hong a bad business. They make it a different business from the one commonly described — a scaling 300mm specialty foundry in the investment phase, not a harvesting legacy operator.

---

↑ Back to Top

IX. Bear vs. Bull Stress Test & Current Risk Radar

Two experienced investors examining the exact same set of Hua Hong filings can arrive at diametrically opposed conclusions while relying on the identical underlying facts. That tension is precisely what makes the company a complex, multi-layered story rather than a simple investment case.

The bear case: the mature-node glut, and the depreciation trap

The short seller's thesis rests on three main arguments, beginning with the most formidable.

China is building more mature-node capacity than the market can absorb. With China's share of global mature-node production projected to expand from roughly 31% to 39% between 2023 and 2027, the marginal supplier driving price discovery is state-subsidised and unburdened by standard cost-of-capital requirements.²¹ Hua Hong's recent operating history provides the case study: during the last industry contraction, its gross margin fell to 4%. The 2026 margin recovery reflects a cyclical memory shortage and AI-driven tightness layered over that structural oversupply, a dynamic management implicitly acknowledged by citing price increases over product mix improvements and declining to project visibility into 2028.¹

The depreciation burden is a matter of arithmetic. Every dollar spent on Fab 9 capital expenditure adds fixed costs that must be recovered in a pricing environment the company does not dictate. Consolidated return on equity sat at 0.9% for two consecutive years, while quarterly capital expenditure continued to exceed revenue and a further multi-billion-dollar Fab 9C investment plan loomed ahead—diluting returns on the existing asset base before earlier investments mature.³⁵¹

Governance frictions and capital structure complexity. A critical inspection of corporate actions reveals multiple red flags: a related-party acquisition from the controlling shareholder funded with new shares priced at roughly a 60% discount to the market price at signing; a whitewash waiver exempting the parent from a mandatory takeover offer; a special deal requiring independent shareholder approval; a circular delayed three separate times; a proposed secondary offering of up to 30% of enlarged share capital; and a combined chairman and president holding no company shares.¹⁹⁶ Furthermore, reported parent-attributable net profit significantly exaggerates group profitability by shifting initial fab operational losses to state-backed minority partners.³ While common among Chinese state-owned enterprises, these mechanisms systematically dilute minority equity interests over time.

The central question for skeptics remains straightforward: if this represents the earnings capacity of the business during a capacity shortage with full fabs and rising prices, what does it earn during a standard cyclical downturn?

The bull case: substitution, 300mm cost position, and the memory adjacency

Domestic substitution is a policy mandate with multi-year momentum. Chinese industrial and automotive OEMs are actively transitioning toward domestic silicon suppliers due to cost advantages and policy directives, leaving Hua Hong as one of only two domestic foundries capable of producing at scale. The industrial and automotive segment grew 19.4% year-on-year in the first quarter of 2026 despite remaining a minority of total revenue, representing the product family with the highest qualification hurdles and customer stickiness.⁵

The 300mm cost curve is beginning to materialize. While bears view Fab 9 strictly through the lens of depreciation, bulls evaluate it as a structural cost advantage. Once a 12-inch specialty manufacturing line achieves full utilization and mature operational yields, its unit cost per surface area of silicon drops below what legacy 8-inch lines can deliver. This positions Hua Hong as a cost leader in specialty devices globally, competing effectively against both domestic peers and legacy 8-inch capacity operated by Infineon, STMicroelectronics, and Vanguard. Twelve-inch production now accounts for over 62% of total revenue.¹⁸

The Huali Microelectronics acquisition provides immediate accretion. Incorporating an established 300mm facility that generated approximately RMB 515 million in net profit across eight months—and whose manufacturing equipment has fully crossed its seven-year depreciation schedule—directly counteracts the parent entity's high depreciation overhead, regardless of valuation debates around the share issuance terms.¹⁹

AI infrastructure creates meaningful indirect demand. Although Hua Hong does not produce leading-edge AI logic chips, expanding AI infrastructure absorbs global leading-edge and memory capacity, tightening overall foundry supply while boosting direct demand for server power-management components. This trend is visible in North American revenue expanding 77% year-on-year and overall fab orders running at 1.5 to 2 times available capacity.¹

The calibrated conclusion

When weighed together, the empirical evidence points to a narrower conclusion than either extreme allows. Hua Hong's proprietary process technology and customer qualification friction successfully protect wafer volumes and market share during industry downturns. However, that differentiation does not grant pricing power or protect gross margins when broader capacity floods the market. The 12-inch expansion represents a valid long-term unit-cost strategy whose ultimate financial returns remain unproven, while its downside—a heavy fixed-cost burden during demand lulls—has already been demonstrated. Meanwhile, the operational recovery underway in 2026 stems primarily from cyclical market tightness rather than a permanent structural shift in product mix.

Current risk radar

Export restrictions represent the primary operational hazard. The April 2026 "is-informed" letters issued by the US Commerce Department target specific manufacturing sites directly, allowing regulators to restrict equipment deliveries to individual expansion projects without formal rule-making procedures.⁷ Concurrently, proposed bipartisan MATCH Act legislation seeks to add Hua Hong to official restricted entity lists.⁷ Because US equipment vendors depend heavily on the Chinese market—with Lam Research generating roughly 35% of quarterly revenue from China and Applied Materials approximately 30%—enforcement actions have historically been targeted rather than comprehensive.⁷ Consequently, trade restrictions act primarily as a drag on future expansion timelines rather than an immediate shock to existing production.

Trade policy adds a distinct layer of commercial risk. The US implemented 25% Section 232 tariffs on specific categories of advanced semiconductors effective 15 January 2026, using tariff scope as a broader trade negotiation tool.²² Although Hua Hong's direct sales to US customers remain small, its domestic clients supply global end products subject to expanding tariff parameters.

Depreciation and capital efficiency remain persistent financial risks. Fixed manufacturing costs continue to expand faster than the underlying pricing power of mature-node silicon can reliably support across full industry cycles.

End-market concentration in consumer electronics creates cyclical vulnerability. Consumer applications generate two-thirds of total revenue, leaving the company exposed to short-cycle consumer spending. Recent strength in microcontrollers and NOR flash reflects a broader memory cycle that historically lacks long-term stability.⁵¹

Operational execution risks are compounding. Managing the ramp of Fab 9A and 9B, initiating the Fab 9C buildout, integrating Huali's facilities, and executing a major secondary equity offering represent four major initiatives proceeding simultaneously under an executive leadership team that took charge at the start of 2025.

↑ Back to Top

X. Epilogue & Key Takeaways for Long-Term Investors

Thirty years separate the approval of Project 909 from the late summer of 2026, and the through-line diverges sharply from what its founding documents envisioned. China set out in 1996 to build a domestic memory industry. Instead, through market failure and state consolidation, it created something less glamorous and more durable: a specialty foundry that makes the unheralded chips essential to modern hardware, for customers outside the tech headlines, in production facilities designed to operate for decades.

What the playbook actually teaches

Specialisation is a defence against capital destruction, not against price competition. The conventional lesson drawn from Hua Hong — that bypassing the leading-edge node race was a masterstroke — is only half right. Declining to compete with TSMC at 3nm saved the company from an arms race it could not fund. Yet avoiding an advanced logic contest offered no immunity from mature-node commodity pressures. When domestic peers expanded mature capacity simultaneously, the fixed-cost dynamics that make memory volatile operated with equal severity in specialty processes, and gross margins compressed accordingly. Specialisation preserved Hua Hong's production volumes and customer retention through a severe downturn, but it did not shield its underlying unit economics.

State capital changes who bears operational losses, not whether those losses occur. Hua Hong's capital structure — combining Big Fund equity, municipal co-investment, minority-owned fab subsidiaries, and a domestic A-share listing at premium valuations — lowered its effective cost of capital and enabled counter-cyclical capacity expansion. However, when that expansion coincided with an industry price war, hundreds of millions of dollars in operational losses were absorbed by non-controlling state partners rather than hitting the parent company's reported line. For investors, this arrangement provides a clear funding-cost advantage alongside an accounting nuance: parent-attributable net profit has consistently presented a more favorable picture than the consolidated performance of the broader enterprise.

Cyclicality remains the dominant variable, making entry valuation critical. Between 2022 and 2026, Hua Hong's gross margin swung from peak levels near 37% down to 4% before recovering into the mid-teens, despite operating a virtually unchanged technology portfolio. In a manufacturing business where fixed overhead accounts for 70% to 80% of operating costs, ultimate investment returns depend far more on capacity utilization and pricing timing than on product positioning alone.

Related-party consolidation can build scale while diluting equity holders. The acquisition of Shanghai Huali Microelectronics demonstrates why corporate governance requires careful scrutiny. Structurally, the transaction brought a 300-millimetre fab past its peak depreciation cycle into a company carrying heavy capital overhead, backed by three years of impairment compensation from the seller. Mechanically, however, the issuance of new A-shares to the controlling shareholder at a steep discount to prevailing market prices, paired with a whitewash waiver and a substantial follow-on equity proposal, shifted value toward the parent entity. Both aspects define the transaction, requiring investors to weigh operational benefits against governance costs.

What has to be true from here

The bull case rests on one verified operational condition and two unproven assumptions.

Currently demonstrated: robust underlying volume demand. Fabs operating above nameplate capacity, a 20% annual expansion in wafer output fully absorbed by the market, customer orders running ahead of available production, and consecutive quarters of record revenue indicate a business with strong demand coverage.⁵¹

Unproven: pricing durability once market tightness normalises. Management attributed recent revenue gains primarily to supply tightness rather than a structural mix shift toward higher-margin products, while explicitly declining to project market conditions beyond 2027. This limited visibility highlights the cyclical nature of the current margin recovery.

Unproven: capital efficiency across the 12-inch footprint. Consolidated return on equity remaining below 1% for two consecutive years, alongside quarterly capital expenditures exceeding total revenue and multi-billion-dollar future expansion phases planned, indicates that 300mm investments have yet to deliver attractive return metrics.

Beyond operating execution sits an external risk: American export restrictions targeting two key facilities over reported advanced-node activities. Regardless of operational specifics, Hua Hong's regulatory status has shifted, creating potential equipment procurement hurdles that directly affect its long-term expansion roadmap.

The KPIs that matter

Three metrics, published quarterly, will resolve most of the debate.

Consolidated gross margin. As the primary financial variable, consolidated gross margin synthesises pricing power, capacity utilization, product mix, and depreciation into a single metric. The key test is where margins settle once current supply tightness eases and Fab 9 depreciation fully hits the income statement. Sustaining margins in the 20% range through a normal cycle would validate the specialty process model, whereas a drop back to single digits during a downturn would confirm commodity pricing behavior.

Twelve-inch revenue contribution alongside blended revenue per wafer. The transition to 300mm wafers represents Hua Hong's core scale strategy, while blended revenue per 8-inch-equivalent wafer reflects net pricing power. Expanding 12-inch volume accompanied by rising revenue per wafer indicates successful operational execution, whereas growing 12-inch share with stagnant or declining ASPs signals that volume is being bought at the expense of returns.

Capacity utilization. Given the high fixed-cost base, utilization acts as the primary leading indicator for profitability. Readings below 85% at year-end 2023 signaled the subsequent margin trough, while above-nameplate utilization throughout 2025 and 2026 preceded the financial recovery.

The last word

Hua Hong is neither an impregnable specialty fortress nor a state-subsidised zombie enterprise. It is an established semiconductor manufacturer with defined process capabilities, a modernizing 300mm asset base, and a resilient customer roster that maintained volumes through an intense price war — backed by a financial record that clearly delineates the boundaries of its market power.

The "China Rainbow" name originated in 1996 as a state policy goal for industrial self-reliance. Three decades later, the company has established substantial manufacturing capacity. Whether that operational scale translates into sustainable return on equity remains a question that upcoming gross margin performance and international regulatory decisions will determine.

---

References

1. Earnings call transcript: Hua Hong Grace posts record Q2 2026 revenue — Investing.com, 2026-08-13 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

2. Hua Hong Grace Semiconductor Limited — Notification of Board Meeting — HKEXnews, 2026-07-30 ↩↩

3. Hua Hong Semiconductor Limited — News Release: Reports 2025 Fourth Quarter Results — HKEXnews, 2026-02-12 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

4. Strong AI Demand and Early Consumer Electronics Inventory Build Drive Top 10 Foundries to 3.7% QoQ Revenue Growth in 1Q26 — TrendForce, 2026-06-12 ↩↩↩

5. Hua Hong Semiconductor Limited — News Release: Reports 2026 First Quarter Results — HKEXnews, 2026-05-14 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

6. Hua Hong Semiconductor Limited — Appointment of Executive Director and Change of President — HKEXnews, 2025-01-01 ↩↩↩↩↩↩↩↩

7. U.S. Reportedly Moves to Block Chip Tool Shipments to China No.2 Foundry Hua Hong as It Seeks 7nm After SMIC — TrendForce, 2026-04-29 ↩↩↩↩↩

8. Shanghai Hua Hong NEC to Produce 64Mb Mass DRAM — Tech Monitor, 1998-07 ↩↩

9. China's Shanghai Grace Semiconductor breaks ground on new fab, report says — EE Times ↩↩↩

10. Chinese foundries Hua Hong, Grace merge (finally) — Semiconductor Digest, 2011-12 ↩↩↩↩↩

11. HHNEC and Grace Semiconductor Complete Merger to Form HHGrace — EE Times, 2011-12-28 ↩

12. Hua Hong prices IPO towards bottom — FinanceAsia, 2014-10 ↩

13. Hua Hong Semiconductor Limited — News Release: Reports 2024 Fourth Quarter Results — HKEXnews, 2025-02-13 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩

14. Hua Hong Grace Semiconductor Ltd (HHUSF) Q2 2026 Earnings Call Highlights — Yahoo Finance, 2026-08 ↩

15. Hua Hong Semiconductor, world's second-largest IPO this year, makes strong start on Shanghai's Star Market — South China Morning Post via Yahoo Finance, 2023-08-07 ↩↩↩↩

16. China Chipmaker Hua Hong Raises $3 Billion in Shanghai STAR Market IPO — Reuters, 2023-08-07 ↩

17. Hua Hong Semiconductor gets state backing for US$6.7 billion wafer fab as China continues its pivot to mature chip nodes — South China Morning Post, 2023-01 ↩↩

18. Hua Hong Grace Posts Record Q2 as AI Demand and Huali Deal Lift Outlook — The Globe and Mail, 2026-08-13 ↩↩↩↩↩↩

19. Hua Hong Semiconductor Limited — Update Announcement on Major and Connected Transaction, Whitewash Waiver, Non-Public Issuance of RMB Shares and Special Deal — HKEXnews, 2025-12-31 ↩↩↩↩↩↩↩↩↩↩↩↩

20. China's Chip Consolidation Heats Up: Hua Hong Acquires 97.5% of Huali, Following SMIC's Earlier Move — TrendForce, 2026-01-02 ↩

21. Comments to the Bureau of Industry and Security Regarding Its Section 232 Investigation of Semiconductor Imports — ITIF, 2025-05-07 ↩↩↩

22. A Month in Semiconductor Policy: Section 232 Measures, BIS Rule, and Taiwan Deal Signal Strategic Push — Global Policy Watch, 2026-02 ↩