YOFC: The Fiber Behind the AI Boom

I. Introduction & Episode Roadmap

On the morning of June 22, 2026, a Hong Kong-listed stock that almost no Western generalist investor had ever traded printed a number that should not have been possible. 长飞光纤 Yangtze Optical Fibre and Cable — YOFC, ticker 6869 in Hong Kong and 601869 in Shanghai — touched HK$304.479. Ten months earlier, in the second half of 2025, the same shares had changed hands at HK$19.872. That is a gain of 1,432%.¹

Nothing about the company had changed in any way a pharmaceutical analyst or a software investor would recognize as a catalyst. YOFC had not discovered a molecule. It had not shipped a chatbot. It had not been acquired. What happened was far more elemental: the world ran short of a strand of glass roughly 125 microns across — about the thickness of a human hair — and YOFC happened to be the largest and most vertically integrated maker of that glass on the planet.

The mechanism is almost absurdly physical. Artificial intelligence training clusters are not built like ordinary data centers. A conventional data center port needs two fiber cores; an AI cluster port needs eight to sixteen, because GPUs must talk to each other constantly and at enormous bandwidth rather than talking mostly to the outside world.² Industry analysis has put the multiple even higher at the rack level — one widely cited estimate holds that an AI-optimized rack consumes roughly 36 times the fiber of a standard server rack.³ Multiply that by the global capital-expenditure wave of 2025 and 2026, and demand for a mature, boring, commoditized industrial product went vertical against a supply base that had spent the prior five years apologizing to shareholders for having built too much.

Prices did what prices do. Mainstream G.652.D single-mode fiber — the workhorse product, the two-by-four of telecommunications — rose more than 400% from its mid-2025 level, from under RMB 20 per core-kilometer to above RMB 80.¹⁴ Bend-insensitive G.657.A2 fiber went from roughly RMB 32 to roughly RMB 240 per core-kilometer inside a year, a 650% move.⁵ Ultra-low-loss G.654.E, the premium long-haul and data-center-interconnect grade, roughly doubled to the RMB 240–260 range.¹ For a business whose economics had been defined for half a decade by price deflation, this was less a recovery than a regime change.

And then, just as suddenly, the market lost its nerve. From the June peak, YOFC's Hong Kong shares fell 47.52% in thirteen trading sessions, erasing roughly RMB 128 billion of market value.¹ The Shanghai line fell further — from roughly RMB 600 in late June to RMB 262.66 by July 30, a 56% drawdown — and it did so after the company had pre-announced first-half net profit up somewhere between 711% and 914% year over year.² Record earnings, collapsing stock. That disconnect is the puzzle this story exists to unpack.

Here is the tension. On one reading, YOFC is a genuine chokepoint supplier to the physical layer of the AI build-out: the only company in the world that has industrialized all three major optical-preform manufacturing processes, holder of a reported 80–90% share of China's highest-grade fiber, with a second, higher-margin business in optical interconnect components growing at nearly 50% a year.⁵⁶ On the other reading, this is the same industry that boomed on a Chinese broadband policy in 2017 and 2018 and then cratered so violently that fiber prices fell by two-thirds in twenty-four months.⁷ The demand driver has changed from government policy to hyperscaler capital expenditure. The supply response mechanism — everyone with a balance sheet builds capacity into a shortage — has not.

What follows traces the whole arc: a 1988 joint venture in Wuhan between the Chinese post-and-telecommunications ministry and Royal Dutch Philips; the decision to master preform chemistry rather than settle for drawing fiber someone else's rod produced; a dual Hong Kong and Shanghai listing that layered a retail shareholder base on top of a state-controlled ownership structure; the near-death pricing experience of 2019 and 2020; the deliberate globalization that now supplies more than 40% of revenue; the AI supply shock; the contested bet on hollow-core fiber; and finally the honest question that the stock's own violence in 2026 suggests the market has not answered — whether a structurally better business and a cyclically extreme price are being confused for one another.

↑ Back to Top

II. Origins: From State Joint Venture to Global Preform Leader

Start with the name, because the name is the history. 长飞 — "Chang Fei" — is a compression of two words: 长江, the Yangtze River, and 飞利浦, the Chinese rendering of Philips.⁸ The company was, quite literally, named for its two parents: a Chinese river and a Dutch electronics conglomerate.

The story properly begins in 1984, when China's State Development Planning Commission approved the import of foreign optical-fiber communication technology — an unglamorous line item in a national plan that would eventually underpin the largest telecommunications network ever built.⁸ Four years later, in May 1988, Yangtze Optical Fibre and Cable was established in Wuhan as a joint venture between Royal Dutch Philips Electronics, Wuhan Optical Communication Technology (the entity that became 烽火通信 FiberHome), and Wuhan Trust and Investment.⁸ The Chinese partners sat inside the orbit of the Ministry of Posts and Telecommunications; the technology came from Eindhoven.

The transfer of that technology was not abstract. YOFC's first cohort of five engineers was sent to Philips facilities for three months of training, and the company's own institutional history records that all five scored above 90 on their final examinations — a detail preserved for four decades precisely because, at the time, it was the entire basis of the enterprise.⁸ Trial production began in 1991; formal operations started in 1992.⁸

What happened next is the part worth dwelling on, because it establishes a behavioral pattern that recurs through the company's entire history. By 1993 — one year into real operations — YOFC's multimode fiber output exceeded that of Philips' own comparable production by 30%.⁸ Then came the more consequential decision. YOFC was offered a partnership arrangement that would have confined it to multimode fiber, the lower-value product. It refused, and in 1997 developed single-mode fiber to what it described as top international standard on its own.⁸ A licensee that declines to stay a licensee is making a specific bet: that the value in this industry accrues upstream, to whoever controls the process chemistry, and that renting someone else's chemistry is a slow death.

The ownership structure churned through the 1990s in a way that reads as a compressed history of Chinese industrial policy. Philips' holdings were partly transferred to Draka Holding N.V. and Chinese partners in 1994; the Chinese side reacquired the Philips stake around the turn of the millennium; and by 2005 Draka had consolidated a 37.5% position alongside Chinese shareholders.⁹ The company was folded into the optical-communications arm of what became 中国华信邮电科技 China Huaxin Post and Telecommunication Technology.⁹ Then in 2011, Italy's Prysmian acquired Draka Holding outright — and inherited, almost as a byproduct of a European cable consolidation, roughly 28.63% of the Chinese company that would one day become its most direct competitor in the highest-value part of the market.⁹

Around all of this sat the demand engine that made the whole thing work: China's telecom build-out, run in sequence rather than in parallel. Fixed-line first, then mobile, then broadband, then 4G, then fiber to the home. For roughly two decades, a Chinese fiber manufacturer's principal job was to keep up. Volume growth was a given; the only question was whether you owned the value-dense part of the chain or the commodity part.

By the mid-2010s YOFC had its answer. It had become one of the two or three largest optical-preform manufacturers in the world by capacity, and the scale of the physical artifact had changed almost beyond recognition. The company's early preforms were about one meter long and yielded roughly 400 core-kilometers of fiber; by the modern era, a three-meter rod 200 millimeters in diameter yielded about 7,000 core-kilometers.⁸ That is a seventeen-fold improvement in output per rod, and it is the single most important number in understanding YOFC's cost position — because in an industry where the input glass is cheap and the process is everything, yield per furnace-hour is the business.

Which raises the question that the next section has to answer plainly, because it is the load-bearing wall of the entire investment case: why does a glass rod matter so much more than the fiber drawn from it?

↑ Back to Top

III. The Moat: Why Preforms, Not Fiber, Is the Business

Picture a clear glass cylinder, roughly the length of an adult and about as thick as a dinner plate, hanging vertically inside a tower. At the bottom of the tower is a furnace running at around 2,000°C. The tip of the rod softens, and a machine below draws it downward into a filament thinner than a hair, running at highway speeds, while sensors hold the diameter constant to within a fraction of a micron. A single rod becomes thousands of kilometers of fiber. That rod is the 光棒 — the optical preform.

Everything about the economics of this industry follows from one fact: the preform is where the value is. Industry analysis in China puts the preform at roughly 70% of the cost of the finished fiber.¹⁰ It is also, by a wide margin, the hardest thing to make. Drawing fiber from a preform is a precision manufacturing task, but it is a task you can buy equipment for. Cabling fiber — bundling it, jacketing it, armoring it — is closer to industrial assembly. Making a preform is chemistry: depositing ultra-pure silica and dopants layer by layer, in vapor phase, with impurity tolerances measured in parts per billion, because a single stray transition-metal atom becomes a light-absorbing defect stretched across kilometers of finished cable.

There are three principal industrial routes to doing this, and the acronyms matter less than the shape of the problem. VAD, vapor axial deposition, grows the rod from the end, like building a candle downward. OVD, outside vapor deposition, sprays soot onto the outside of a rotating rod, like building up layers of paint. PCVD, plasma chemical vapor deposition, deposits material on the inside of a glass tube using a plasma, which gives extremely fine control over the refractive-index profile — the invisible internal geometry that determines how light behaves as it travels. Each route has different strengths in cost, scale, and precision. Historically, mastering even one at industrial scale was a decades-long undertaking, and most manufacturers licensed their process from someone who had already done it.

YOFC's structural claim is that it has industrialized all three. Chinese state media has described it as the only company globally to have mastered PCVD, OVD, and VAD, with a 100% preform self-sufficiency rate.⁵¹⁰ That claim deserves scrutiny rather than applause — it is repeated frequently in Chinese coverage and originates substantially with the company — but the underlying capability is corroborated by an observable fact: YOFC does not buy preforms. Several of its domestic competitors historically did, or built their capability on licensed process technology. In a shortage, the difference between a company that makes its own rods and a company that buys them is not a rounding error in gross margin. It is the difference between capturing the price spike and paying it.

This is where it is worth being precise about the type of advantage in play, because "moat" is used too loosely in this industry. In Hamilton Helmer's taxonomy, YOFC's edge is closest to Process Power — an advantage embedded in organizational routines and accumulated know-how that a competitor cannot buy off the shelf and cannot replicate quickly even with unlimited capital, because the learning is tacit and the iteration cycles are slow. There is a supporting layer of Switching Costs: fiber going into a carrier backbone or a hyperscaler's interconnect must pass qualification cycles that run for quarters, sometimes years, and once a grade is qualified into a network architecture, requalifying a second source is a project nobody undertakes casually. What YOFC does not have is Network Economies, Branding, or Counter-Positioning. Nobody buys fiber because of the logo on the drum.

The honest read on concentration supports "strong incumbent," not "monopoly." As of 2021, China's total preform capacity exceeded 20,000 tons, with YOFC the largest at roughly 6,000 tons — about 30% — and 亨通光电 Hengtong Optic-Electric second at around 3,300 tons, or 17%.¹¹ Industry CR3 was around 60% and CR5 around 80% — what Chinese industry researchers classify as a moderately concentrated oligopoly.¹¹ By 2023, the top five held roughly 78% of the domestic preform market, with YOFC still above 30% and followed by Hengtong, 中天科技 Zhongtian Technology, FiberHome, and 富通集团 Futong Group.¹¹ Those are the numbers of a leader in a field of credible rivals, not a company that can dictate terms indefinitely.

There is a second-order point buried in that structure that matters enormously for the later sections. A moderately concentrated oligopoly with four or five well-capitalized players and a demand shock is the textbook setup for a capacity race. Every one of those competitors can read the same price charts. Every one of them has access to Chinese industrial credit. And the barrier that keeps new entrants out — the 18-to-24-month construction and qualification cycle for a new preform line — is also the barrier that guarantees the supply response arrives all at once, roughly two years after the shortage begins.¹²

Hold that thought. Before the AI shock, YOFC had to become a public company — and the shape of who owned it turned out to matter as much as what it made.

↑ Back to Top

IV. Going Public: A+H Listing and Who Actually Owns YOFC

In December 2014, YOFC listed H shares on the Hong Kong Stock Exchange under the code 6869, becoming the first and only Hong Kong-listed company specialising in optical fibre preforms, fibre and cable.¹³⁸ For a Wuhan manufacturer whose customers were three state-owned telecom carriers, tapping international capital markets was less about the money than about the discipline: Hong Kong listing rules, an English-language reporting cadence, and an investor base that would ask about return on capital rather than about production targets.

The second listing took three and a half more years and mattered more domestically. On July 20, 2018, YOFC completed an A-share IPO on the Shanghai Stock Exchange at RMB 26.71 per share, making it the first A+H dual-listed company both in China's fiber-optic cable industry and in Hubei province.¹⁴ The stock hit its daily limit on debut, up 43.99%.¹⁴ The timing, in hindsight, is almost cruelly instructive: YOFC listed onshore at the precise apex of the previous fiber cycle, weeks before the domestic price collapse began. Investors who bought at the IPO and held would spend the next several years learning what "cyclical" means.

The more consequential question for anyone underwriting this business today is not when it listed but who controls it. China Huaxin Post and Telecommunication Technology has been the largest shareholder, holding roughly 23.7% — and in 2019 that stake's ultimate parentage was transferred without consideration to 保利集团 China Poly Group, a central state-owned conglomerate best known outside the industry for real estate and defense trading.[^15]¹⁵ The second-largest holder has been 武汉长江通信产业集团 Wuhan Yangtze Communications Industry Group, itself a listed vehicle, which held 118,837,010 domestic shares — 14.35% — as of March 31, 2026.¹⁶ Draka Comteq, the Prysmian subsidiary, held roughly 9.9% until it did not.[^15]

That last thread resolved on September 16, 2025, and the resolution is one of the great "so close" trades in recent capital markets history. Prysmian announced the placement of its entire remaining holding — 37,595,257 H shares, about 5.0% of total share capital — at HK$47.10 per share, raising approximately HK$1.8 billion, or EUR 193 million, through J.P. Morgan as sole bookrunner.¹⁷ Draka thereby divested its entire stake, ending a relationship that traced back to the Philips joint venture of 1994. Nine months later the same shares traded above HK$304.¹ Prysmian sold a strategic legacy position into the last quiet moment before the largest re-rating in the company's history, and converted a partner into a pure competitor at the worst possible price. It is a useful reminder that industry insiders with thirty years of proximity to a business are not necessarily better at timing its cycle than anyone else.

The ownership picture that emerges should be read carefully, because it drives capital-allocation expectations. YOFC is effectively a central-SOE-controlled enterprise. Poly Group sits at the top of the control chain; a second state-linked listed vehicle holds a substantial minority; and layered on top is a large, fast-moving Chinese retail shareholder base that has made the A-share one of the most heavily traded names on the Shanghai exchange. This is not a founder-led growth company, and it should not be modeled as one. Decisions about whether to plow a windfall into new preform capacity or return it to shareholders are made inside an institutional framework where national industrial self-sufficiency is a legitimate objective in its own right — not a euphemism, an actual mandate.

There is a live example of the ownership dynamic that skeptical investors should note. Wuhan Yangtze Communications' own stock became a retail proxy for YOFC exposure during the rally, and YOFC's market capitalization crossing RMB 170 billion in February 2026 was covered in the Chinese financial press explicitly through the lens of the wealth it created for its parent shareholder.¹⁸ A significant shareholder above the 5% threshold completed a disclosed sell-down programme, with tranches executed in the RMB 380–448 range during the 2026 run-up.²¹⁹ Insider distribution into a parabolic move is not proof of anything by itself, but it is information, and it sits uncomfortably beside a narrative of permanent re-rating.

One further capital-allocation data point, small in size and revealing in pattern: on July 30, 2026, YOFC announced it would pay EUR 12 million — about RMB 92.61 million — to acquire Draka Comteq's remaining 25% of YOFC Shanghai, converting the joint venture into a wholly owned subsidiary.²⁰ YOFC Shanghai generated RMB 719 million of revenue in 2025 against net profit of just RMB 4.85 million, so this is a cleanup, not a transformation.²⁰ But it fits a consistent behavior: when the foreign partner is leaving, take full ownership of the strategic asset quietly and cheaply.

To understand why management and investors alike should be cautious with a windfall, though, you have to go back to the last time this industry printed money — and to what happened immediately afterward.

↑ Back to Top

V. The First Boom-Bust: China's Broadband Supercycle Turns to Bust (2013–2020)

In 2013, China's State Council issued the 宽带中国 "Broadband China" strategy, and a national demand curve was written into policy. Fiber to the home would be pushed into hundreds of millions of Chinese households on a defined timetable. For the fiber industry, this was the closest thing to a guaranteed order book that a manufacturing business ever receives: the buyers were three state carriers, the volumes were set by plan, and the plan was national policy.

Everyone behaved exactly as economic theory predicts. Standard fiber prices climbed toward RMB 60 per core-kilometer at the 2018 peak, gross margins fattened across the industry, and every producer with access to credit — which, in the Chinese industrial system, meant every producer — put steel in the ground.⁷ Preform capacity, fiber-drawing towers, and cabling lines all expanded simultaneously to chase a demand curve that policy had made look permanent.

It was not permanent. FTTH penetration approached saturation. The 4G build-out finished. And in 2019, Chinese optical cable demand fell by double digits year over year — reversing a ten-year trend in a single reporting period.²¹ The centralized procurement tenders run by 中国移动 China Mobile, 中国电信 China Telecom, and 中国联通 China Unicom, which had functioned for a decade as an escalator, abruptly became a ratchet in the other direction. Prices fell from roughly RMB 60 per core-kilometer to above RMB 30, and then to under RMB 20.⁷

The 2020 China Mobile tender is the historical low-water mark, and it deserves to be named precisely because it is the counterfactual to everything happening in 2026. In that award, YOFC's most competitive bid was RMB 39.23 per fiber-kilometer of standard loose-tube cable — approximately $5.61 per fiber-kilometer — from which CRU estimated an implied bare-fiber price of $2.80 to $3.00 per fiber-kilometer.²² Bare fiber, the physical medium of the global internet, was being sold at a price that would not cover a cup of coffee per kilometer.

The consequences ran straight through the income statements of the entire sector. Margins collapsed. Smaller and mid-tier producers posted outright losses. Even the leaders — the companies with the best cost positions and the most integrated supply chains — saw gross margins compress toward single digits on the core fiber business. YOFC's own trajectory tells the story without commentary: revenue of RMB 13.353 billion in 2023 fell to RMB 12.197 billion in 2024, declines of 3.45% and 8.65% respectively, while 2024 net profit dropped 47.91% to RMB 676 million.²³ The company's first-half 2024 earnings announcement documented a sharp downturn.²⁴ Five and a half years after the peak, the industry was still working off the hangover.

What makes this history analytically indispensable is the structural identity between then and now. In both cases: a demand shock arrives from outside the industry (state broadband policy then, hyperscaler AI capital expenditure now). In both cases: the demand shock is large enough that existing capacity cannot serve it, so prices spike. In both cases: prices spiking causes every participant to expand, with a construction-and-qualification lag of roughly two years. In both cases: the demand driver has a natural terminus — households that already have fiber do not get fiber again, and data centers that are already built do not need to be rewired annually.

The differences are real and worth stating fairly. Broadband China was a one-time penetration event with a mathematically hard ceiling — a country has a finite number of households. AI infrastructure demand is not obviously bounded in the same arithmetic way, and the product mix has shifted toward higher-specification fiber where the technical barriers are meaningfully higher. But the mechanism of the bust — a fast-follower supply response to a temporary price signal — does not care whether the demand was policy-driven or capex-driven. It cares only that the price signal was loud enough to be heard by everyone at once.

That lesson was learned expensively. The next chapter is about what YOFC did with it.

↑ Back to Top

VI. Recovery, Diversification, and Going Global (2021–2024)

The most revealing thing about YOFC's post-bust strategy is that it was not primarily a cost-cutting strategy. It was a customer-mix strategy — and it started, importantly, before anyone had heard of a large language model.

The domestic recovery came first and came slowly. Through 2021 and 2022, smaller producers exited or consolidated, the 5G and F5G build-outs provided a second demand leg, and the 东数西算 "East Data, West Computing" initiative — China's plan to relocate compute-intensive data centers to the resource-rich west while keeping demand centers in the east — created genuine long-haul fiber demand of a kind FTTH never had. Carrier procurement volumes and prices both recovered modestly. Analysts at the time described a two-to-three-year upcycle. It is worth registering that both management and the sell-side were calling a shorter, shallower cycle than the one that actually arrived in 2025 and 2026 — a mark against anyone's ability to forecast this industry, including the people running it.

The more deliberate move was geographic. YOFC's internationalization began around 2014 and was scaled aggressively from roughly 2015 onward. Speaking to Developing Telecoms, Executive Director and President 庄丹 Zhuang Dan framed it in plain operational terms: "We've expanded overseas, establishing eight manufacturing bases across six countries."²⁵ Those countries were Indonesia, South Africa, Brazil, Poland, Mexico, and Germany, supported by more than fifty sales offices outside China and a workforce more than 20% of which was based internationally.²⁵ The Polish facility, roughly 80 kilometers from Warsaw, became the primary regional supplier for Europe, and its expansion project was completed in 2024.²⁶

The revenue arithmetic on that strategy is the part that should interest investors. International sales rose from under 8% of revenue in 2014 to roughly a third by the time of that interview, and Zhuang stated the goal of making overseas the largest single revenue contributor.²⁵ By full-year 2025, overseas revenue reached RMB 6.092 billion, up 47.81% year over year, and represented 42.7% of total revenue.⁶²⁷ At the 2026 World Artificial Intelligence Conference in July, Zhuang said the company expected more than half of 2026 revenue to come from overseas customers, across ten production bases in seven countries.²⁸

Read that as a risk-management decision rather than a growth story. A Chinese fiber manufacturer selling only to Chinese carriers has exactly three customers, all state-owned, all buying through centralized tenders designed to extract price. Building plants in six countries does not create pricing power on its own — a Polish cable plant faces Polish competition. What it does is decouple the revenue base from a single procurement regime, and provide a partial hedge against tariff and technology-security actions that increasingly target Chinese-origin telecom equipment. The 2025 numbers suggest the hedge is real: overseas grew nearly 48% in a year when the domestic core was still recovering.

The second and less-discussed leg of the diversification is arguably more important to the current story. YOFC has been building businesses outside fiber and cable — optical interconnect components and modules, and specialty materials — and by 2025 they had stopped being a footnote. Diversified businesses generated approximately RMB 5.9 billion in 2025, crossing 40% of total revenue for the first time.⁶²⁷ Within that, the optical interconnect components segment alone delivered RMB 3.144 billion, up 48.58% year over year, at a gross margin of 39.73% — materially above the 35.90% earned by the much larger optical transmission products segment, which grew only 6.09% to RMB 8.346 billion.⁶²⁹

That margin differential is the single most strategically interesting number in the pre-boom financials. Optical interconnect components — the connectors, modules, and assemblies that terminate fiber inside a data center — are closer to the customer, more design-intensive, and less exposed to the centralized-tender pricing regime that has historically capped fiber economics. A business that is growing at 49% with a higher gross margin than the legacy core, funded by the legacy core's cash flow, is the textbook definition of a good reinvestment. Whether it is durably differentiated is a separate question, and one the AI boom is currently making impossible to answer, because everything in this supply chain is growing.

Full-year 2025 pulled it together: revenue of RMB 14.252 billion, up 16.85%, and net profit attributable to shareholders of RMB 814 million, up 20.40%, with gross margin of 30.7% — an improvement of 3.4 percentage points over 2024.⁶²⁹ Respectable. Unremarkable. A recovering industrial company with an interesting sideline.

Six months later, that description had become obsolete.

↑ Back to Top

VII. The AI Supercycle: 2025–2026

There is a line from Chinese state media coverage of this boom that captures the shift better than any market-sizing chart: "Optical fiber used to be laid to households; now it is laid next to every GPU."⁵

That sentence does a lot of work. For thirty years, the fiber industry's job was to connect people — one strand to a home, a few strands to a cell tower, a big bundle down a highway between cities. The fiber ran long distances and carried traffic that was overwhelmingly asymmetric: the network sent far more data down to users than it received back.

An AI training cluster inverts every one of those assumptions.

The mechanism, in plain terms. When tens of thousands of GPUs train a single model, they are not serving users. They are talking to each other, constantly, in an all-to-all pattern, exchanging gradients and parameters at bandwidths that would have described an entire national backbone a decade ago. The distances are short — meters to a few kilometers, within and between buildings on a campus. But the density is extreme. A conventional data center port needs roughly two fiber cores; an AI cluster port needs eight to sixteen.² At the rack level, industry estimates put AI configurations at something like 36 times the fiber content of a standard server rack.³

So the industry has been asked, essentially overnight, to supply a different product profile: far more cores per unit of floor space, tighter loss budgets, higher-specification grades, and delivery on hyperscaler timelines rather than carrier timelines.

The demand data confirms the shift is not marginal. Data-center optical fiber demand in China grew 75.9% year over year in 2025 to 69.6 million core-kilometers, with 2026 projected to exceed 100 million.⁵ Globally, CRU data has been cited putting AI data centers below 5% of total fiber demand in 2024, rising toward 30% by 2027.³ Zhuang Dan's own framing at the 2026 World AI Conference was more aggressive still: datacom historically accounted for under 10% of global fiber demand and would exceed 50% by 2030.²⁸ Management projections should be discounted accordingly, but the direction is corroborated by independent forecasters.

Why supply could not respond. Because the industry had spent five years being punished for building capacity.

Chinese preform capacity utilization reached 84.3% in 2026 — effectively full, since maintenance and product changeovers consume the remainder.¹² CRU-sourced analysis has put the 2026 global fiber supply-demand gap at roughly 16.4%.¹² Lead times stretched to twenty weeks for large-volume buyers and up to a year for smaller ones.³ Nobody had spare preform capacity, because spare preform capacity had been the thing that nearly bankrupted the sector in 2020.

Then the prices moved, and they moved in the way prices move when a physically constrained input meets inelastic demand from customers whose alternative is leaving a billion dollars of GPUs idle.

YOFC's specific leverage. Three things determined how much of this the company captured.

First, self-sufficiency. With 100% of its preform needs met internally, YOFC did not have to buy rods at spiking prices to draw fiber.¹⁰ Every yuan of the preform price increase flowed to its own margin rather than to a supplier's.

Second, product mix. YOFC has been reported to hold 80–90% of the high-end G.654.E ultra-low-loss segment — the grade used for long-haul routes and data-center interconnect, and the most AI-exposed, highest-margin slice of the fiber line.⁵ The company has also stated it is China's only enterprise with formal mass-production certification for polarization-maintaining fiber, a specialty grade Zhuang described in July 2026 as facing demand growth of ten to twenty times over the following one to two years.²⁸ These are company-sourced claims and should be treated as such, but the G.654.E position in particular is widely reported and consistent with the company's technical history.

Third, integration. Owning the rod, the fiber, and the cable means capturing margin at each step in a period when every step repriced upward simultaneously.

The financial translation. The 2025 full-year numbers, published in March 2026, already looked good but not extraordinary. The 2026 numbers are a different species.

First-quarter 2026 revenue rose 27.7% to RMB 3.695 billion while operating costs rose just 3.5% — the arithmetic of a price shock passing through a fixed cost base.²³ Gross margin hit 41.51%, up 13.67 percentage points year over year and 5.85 points sequentially, and net profit jumped 226.4% to RMB 495 million.²³²⁷

Then the second quarter happened. On July 14, 2026, YOFC pre-announced first-half net profit of RMB 2.4–3.0 billion, growth of 711% to 914% against RMB 296 million in the prior-year period, with non-GAAP net profit up 1,349% to 1,784%.²⁷ The implied second quarter alone was RMB 1.9–2.5 billion — three to four times the first quarter's result, sequentially.²⁷

The audited half-year figures confirmed the scale. Revenue rose 53.6% to approximately RMB 9.81 billion, gross profit nearly tripled to roughly RMB 5.23 billion, and — the detail that matters most for reading management's intent — the board resumed dividend payments.³⁰

UBS responded by modeling EBIT margin expanding from 11.4% in 2025 to 32.9% by 2027, describing the company as having entered a profit upswing cycle, and noting that the mainstream G.652.D price move had spread beyond China into European and U.S. markets.³¹

What the evidence actually establishes. It is worth separating what these results prove from what they merely suggest.

They prove operating leverage. A 53.6% revenue increase producing a near-tripling of gross profit means the incremental revenue arrived at extremely high contribution margin — which is what happens when price rather than volume drives the top line and the cost base is largely fixed.

They prove the vertical integration thesis in the only way that counts: under stress. When preform was the binding constraint, the company that made its own preforms earned the scarcity rent.

They do not prove pricing power in the durable sense. Pricing power is the ability to raise price and keep it. What YOFC demonstrated in 2026 was the ability to receive a price set by a shortage it did not create. Those are different things, and the distinction is precisely what the bear case turns on.

They also do not settle the question of market share. A less-discussed data point cuts against the narrative of unassailable dominance: in China Mobile's 2025 fiber optic cable tender, YOFC's share fell from the 19–20% range to 13.6%, dropping it to third place from first.²³ In a 2026 special cable tender, it took 15.94%, ranking second.²³ The company has been explicit that it is prioritizing high-end products over volume share in commodity tenders — a defensible strategy, and also exactly what a company losing commodity share would say. Both readings fit the facts.

The stock, and the vertigo. Markets did not treat this as a normal earnings recovery. At the peak, YOFC's combined A+H market capitalization crossed RMB 300 billion, the first Hubei-headquartered company to reach that level, and the A-share line alone was valued at RMB 331.3 billion at RMB 400.18 per share in mid-July.²³²⁷ Trailing P/E briefly exceeded 400 times against a historical industry median closer to 28 times.¹ Even after the crash, the trailing multiple was still around 308 times.²

For context on how far outside normal this sat: the same analysis noted that YOFC's trailing multiple exceeded that of far more richly valued technology names in the Chinese market.¹

That is the valuation tension the rest of this story has to resolve. But before that, there is the technology bet that management has positioned as its answer to the question everyone should be asking — what happens when the shortage ends?

↑ Back to Top

VIII. The Next Bet: Hollow-Core Fiber and "AI-2030"

Here is a fact about optical fiber that sounds like it should not be true: light travels roughly a third slower inside glass than it does through air.

Every kilometer of conventional fiber is a kilometer of solid silica, and the light pulse carrying data has to push through it, slowed by the material and subtly distorted by nonlinear interactions with the glass itself. For a telephone call, irrelevant. For a high-frequency trading firm arbitraging two exchanges, or for a GPU cluster where a microsecond of added latency multiplied across millions of collective operations becomes real training time, it is the whole game.

Hollow-core fiber removes the glass from the path. The core is air — or near-vacuum — held in place by an intricate microstructured lattice of thin glass membranes that confine the light through interference rather than through the total internal reflection that governs ordinary fiber. As YOFC Senior Vice President Jan Bongaerts put it: "Light travels through a vacuum so you don't have any attenuation."³²

The engineering difficulty is immense. Those glass membranes are sub-micron structures that must be held to tolerance over kilometers of drawn fiber, and for two decades hollow-core fiber was a laboratory curiosity with attenuation orders of magnitude worse than standard fiber.

The AI-2030 programme. YOFC unveiled its AI-2030 initiative in June 2025, structured around three declared objectives: ultra-large bandwidth, ultra-low latency, and ultra-low loss.³²³³ Hollow-core fiber is the centerpiece.

The claims are specific, which is helpful, and almost entirely company-sourced, which is not. YOFC has reported a laboratory minimum attenuation of 0.040 dB/km — below the Rayleigh scattering floor that limits conventional silica fiber, and claimed as a world record — with commercial-scale production achieving consistent sub-0.1 dB/km performance.³² Company-reported testing indicated a 31% reduction in transmission latency and a 47% increase in propagation speed versus conventional fiber.³² The company introduced its HollowBand branded hollow-core product at MWC Barcelona in 2026.³⁴

The proof points that are not just claims. Two deployments give the programme more substance than a specification sheet would.

The first is a high-frequency trading link between the Shenzhen and Hong Kong exchanges achieving sub-one-millisecond round-trip latency.³² Financial trading infrastructure is an unusually credible early market: the customers measure latency obsessively, they will pay extraordinary premiums for microseconds, and they do not accept vendor claims on faith.

The second is a 110-kilometer corridor connecting data centers in Dongguan and Hong Kong, described as the world's longest commercial hollow-core deployment.³² Length matters here because hollow-core's historical failure mode was that losses and splice complexity compounded with distance. A 110-kilometer production link is a meaningfully harder engineering problem than a campus interconnect.

Zhuang Dan's own framing has been notably restrained relative to the marketing: "It is seldom to find companies that are able to design and manufacture hollow-core fibre. YOFC is one of the few."²⁵ Not "the only." Few.

Why this is contested rather than consensus. In mid-2026, the market delivered its verdict on how much of this to believe, and it was not flattering.

Part of the trigger was Corning's Glass Bridge technology, which Chinese analysts read as potentially reducing demand for premium fiber by eliminating manual calibration requirements — and therefore as a challenge to the scarcity premium embedded in high-end fiber pricing.¹ YOFC management characterized the actual impact as "very, very limited."² The market did not fully accept that.

The deeper concern is structural. Hollow-core fiber, if it works at scale and at cost, is genuinely disruptive — but disruptive to whom? A technology that reduces the amount of glass required per unit of data transmitted is not obviously good for a company whose earnings currently derive from a glass shortage. And a technology race in which several global players are running hard is not one where a lab record confers durable advantage; it confers a lead measured in quarters.

The honest assessment is this. Hollow-core fiber is real optionality — a credible technical position in a genuinely difficult field, with two deployments that are hard to fake. It is not a present-day earnings driver, it is not independently verified at the level of the headline attenuation claims, and it is not the reason anyone owns this stock today. Management has positioned it as the answer to "why we keep winning after the shortage normalizes." The market's mid-2026 response suggests investors want more evidence before accepting that framing.

Which brings the story to the people making these bets, and to whether their track record supports the benefit of the doubt.

↑ Back to Top

IX. Current Management, Incentives, and Capital Allocation

There is a specific kind of executive career that only exists inside Chinese state-linked industrial companies: the technically credentialed lifer who runs the same asset through multiple cycles, accumulating institutional knowledge that no external hire could replicate, while operating inside constraints that no Western CEO would accept.

庄丹 Zhuang Dan is that executive.

The operator. Zhuang serves as Executive Director and President, and his credentials are unusually academic for a manufacturing chief: doctoral-level education associated with Wuhan University and Zhongnan University of Economics and Law, with postdoctoral work at the Shanghai University of Finance and Economics. He has led YOFC through the 2019–2020 collapse and through the 2025–2026 boom — the same person, the same seat, two opposite worlds.

What makes him analytically interesting is not the biography but the consistency of the strategic narrative across those two worlds.

In the Developing Telecoms interview — conducted before the AI shock, when the domestic business was still grinding — Zhuang described internationalization not as an opportunity but as a structural necessity, and set the explicit goal of overseas becoming the largest revenue contributor.²⁵ By 2026, with the domestic business printing record margins, the message had not changed: he was still guiding to over half of revenue coming from overseas customers.²⁸

That consistency matters. A management team that changes its strategic story to match whatever is currently working is telling you it does not have a strategy. YOFC's internationalization was articulated years before it paid off, funded during a period of depressed earnings, and has not been abandoned now that the domestic business is more profitable than the overseas one. The overseas revenue trajectory — from under 8% in 2014 to 42.7% in 2025 — is the receipt.²⁵⁶

Where Zhuang is more open to challenge is on forecasting. His WAIC 2026 statement that polarization-maintaining fiber demand would grow ten to twenty times in one to two years is the kind of number that gets quoted in every bullish note and is essentially unfalsifiable in the short run.²⁸ Executives in a shortage have a strong incentive to describe the shortage as structural. Investors should weight the operating decisions more heavily than the forward-looking adjectives.

The chairman and the framework. 马杰 Ma Jie serves as chairman and legal representative, and the governance structure sits within the state-owned enterprise framework established when the controlling interest passed to Poly Group.¹⁵ Board decisions on major capital deployment are taken inside that framework. This is neither good nor bad in the abstract, but it does mean that "what would maximize per-share value" and "what the controlling shareholder wants" are two questions with potentially different answers.

The compensation friction. One detail has drawn explicit criticism in the Chinese financial press, and it is worth naming because it illuminates the hybrid nature of this company.

For 2024, directors and senior management collectively received RMB 27.6352 million, an average of RMB 2.7656 million per person.¹⁵ Zhuang Dan individually received RMB 5.6353 million — substantially more than the second-highest-paid executive, Vice President Wang Ruichun, at RMB 3.7358 million.¹⁵ Coverage in Sina Finance framed this pointedly: pay levels far exceeding comparable state-owned telecommunications enterprises such as FiberHome, at a company that is itself state-controlled.¹⁵

Two readings are available. The charitable one is that YOFC competes for talent against private-sector optical companies and pays accordingly, and that market-rate compensation inside an SOE framework is a governance positive because it aligns management with performance rather than with bureaucratic rank. The skeptical one is that a company with a state controlling shareholder and a captive retail investor base has weak checks on executive pay in either direction. Both readings are defensible. The fact that Chinese domestic media raised the issue at all is a reminder that this is contested territory locally, not just an outsider's concern.

Capital allocation: the actual record. This is where behavior beats rhetoric, and YOFC's 2025–2026 record is genuinely mixed rather than uniformly good or bad.

On the shareholder-return side, the company ran an A-share buyback programme in 2025. For full-year 2025 it proposed a cash dividend of RMB 2.95 per 10 shares, totaling roughly RMB 244.2 million — approximately 30% of attributable profit.²⁹ Then, with the first-half 2026 profit surge, it resumed dividend payments after a period without them.³⁰ Resuming a distribution in the same period that earnings inflect is the minimum acceptable behavior, but it is behavior rather than a promise, and it is the concrete evidence available.

On the reinvestment side, the picture is unresolved and is the single most important thing to watch. The company maintains R&D spending at 5% or more of revenue with, in its own framing, no upper limit.²⁹ That is a defensible policy for a process-technology business. What is not yet knowable is how much of the windfall goes into new preform capacity — the exact decision that, taken collectively across the industry in 2017 and 2018, produced the 2019–2020 collapse.

The course-correction test. The most useful diligence exercise available to an investor here costs nothing: read the FY2019 and FY2020 annual report letters against the FY2025 and H1 2026 disclosures and check whether management explains the bust in the same terms it now explains the boom.

If the 2019 explanation was "industry overcapacity and irrational competition" — an external framing that assigns blame outward — and the 2026 explanation is "our technology leadership and integrated platform" — an internal framing that claims credit inward — that asymmetry is itself an analytical finding. It would suggest a management team that treats downturns as weather and upturns as skill. If instead the 2019 disclosures acknowledged the company's own role in the capacity build, that materially raises the credibility of current capital-allocation discipline.

The company's own investor relations portal and announcements page host the primary documents required for that comparison.³⁵³⁶

The other half of the credibility question is external: what the competition is doing while YOFC decides.

↑ Back to Top

X. Competitive Landscape and Industry Structure

War-game this industry properly and the first thing you notice is that it has two entirely different competitive maps depending on which product you are looking at.

The global field. In commodity fiber and cable, the map is crowded and Chinese-dominated. Chinese manufacturers account for more than 60% of the global optical fiber and cable market, and roughly 70% of global preform and fiber capacity is physically located in China.⁵ The domestic rivals are formidable: Hengtong Optic-Electric, FiberHome, and Zhongtian Technology, with Futong Group rounding out the top five in preforms.¹¹

In high-specification fiber and in preform process technology, the map narrows dramatically. Corning has historically been the number one or two preform producer globally and remains the benchmark for process innovation outside China. Prysmian, YOFC's former part-owner, is now a pure competitor across cable and increasingly in high-value segments. フジクラ Fujikura and 住友電工 Sumitomo Electric hold strong positions in Japan, particularly in specialty fiber and connectivity.

The Fujikura data point from May 2026 is worth flagging as a sentiment marker: the Japanese fiber stock halved in a single week, and YOFC's Hong Kong line fell more than 8% in sympathy on May 26, 2026, more than 25% off its high at that point.³⁷ These stocks now trade as a correlated global AI-infrastructure complex, which means YOFC shareholders are exposed to sentiment shocks originating anywhere in the chain — including in companies they have never heard of.

Porter's five forces, read honestly.

Supplier power: low. The raw inputs — silicon tetrachloride, high-purity gases, quartz tubes — are not scarce in any fundamental sense. The scarcity in this industry is process know-how and qualified capacity, not materials. This is genuinely favorable and it is a structural feature, not a cyclical one.

Buyer power: high, and bifurcating. For three decades, the buyer side was China Mobile, China Telecom, and China Unicom running centralized tenders explicitly designed to grind price to the marginal producer's cost. The 2020 tender demonstrated exactly how effective that machinery is.²² The new buyer cohort — hyperscalers and AI infrastructure operators — behaves differently: less price-transparent, longer sales cycles, more willing to pay for specification and delivery certainty. Whether they will behave differently once supply loosens is unknown, and it is one of the genuine open questions in the thesis. Large technology buyers have historically been ruthless about commoditizing their supply chains the moment they can.

Threat of new entrants: moderate-to-low, but decidedly not zero. Preform manufacturing requires heavy capital, an 18-to-24-month build cycle, and multi-year customer qualification.¹² Those are real barriers. But they are barriers of time, not of impossibility — and in an industry earning windfall margins, time is exactly the thing capital is willing to wait through.

Threat of substitutes: rising, and newly relevant. This force was dormant for twenty years. It is not dormant now. Corning's Glass Bridge, hollow-core architectures, multi-core fiber, and co-packaged optics all represent routes by which the amount of conventional glass required per unit of transmitted data could fall.¹²³ Any of these could compress demand in specific applications without any change in AI capital expenditure at all.

Rivalry: violently variable. This is the force that has swung twice in a decade — brutal price competition in 2019 and 2020, unusually rational and tight pricing in 2025 and 2026 because AI demand outran everyone's planned capacity simultaneously. The current state is not the product of discipline. It is the product of a synchronized capacity shortfall, and it has historically not persisted.

Where YOFC's advantage is real, and where the evidence is thin.

The evidence bar should be set consistently, and applied in both directions.

Real and reasonably well-supported: full in-house mastery of all three preform processes with 100% self-sufficiency, corroborated by state media and consistent with a documented four-decade technology history that includes refusing to remain a multimode licensee in the 1990s.⁵⁸¹⁰ Real and widely reported: the 80–90% position in G.654.E.⁵ Real and independently visible in the financials: the optical interconnect components business at 39.73% gross margin, growing 48.58%, which is a genuinely different economic profile from commodity fiber.⁶²⁹ Real and geographically verifiable: eight to ten production bases across six to seven countries and 42.7% overseas revenue.²⁵²⁸⁶

Thinner: the hollow-core commercialization claims, which are substantially company-sourced and not independently audited.³² Thinner still: the implicit assumption that current pricing reflects structural advantage rather than a cyclical shortage that lifts every producer.

And there is a genuinely awkward counter-fact. If YOFC's competitive position were as dominant as the narrative implies, its share of China Mobile's flagship 2025 cable tender should not have fallen from around 19–20% to 13.6%, nor should it have finished third.²³ The company's explanation — that it is deliberately allocating scarce capacity to high-margin product rather than to commodity tenders — is economically rational and probably true. But it also means that in the largest, most transparent, most competitive procurement event in its home market, YOFC is not winning on price. It is winning elsewhere. That is a legitimate strategy and a legitimate vulnerability, and investors should hold both ideas at once.

Which sets up the argument that the market itself has been unable to settle.

↑ Back to Top

XI. Bull vs. Bear: Stress-Testing the AI Infrastructure Thesis

The most striking thing about YOFC in 2026 is that the bull and bear cases are both built from the same facts. Nobody disputes the earnings. Nobody disputes the technology. What is disputed is the duration.

The bull case, argued at full strength.

Start with the physical reality. The AI build-out is not a software phenomenon; it is a construction project on a civilizational scale, and it requires an enormous quantity of a specific industrial product that only a handful of companies can make to specification. YOFC is one of them, and it is the only one that controls its own preform supply completely.

The technology position is not rhetorical. Three preform processes industrialized, 100% self-sufficiency, a dominant share of the highest grade, and the only Chinese mass-production certification for polarization-maintaining fiber.⁵¹⁰²⁸ In a market where the binding constraint is qualified capacity rather than raw material, that combination is worth a great deal.

The second engine is already visible in the accounts. Diversified businesses — optical interconnect components and specialty materials — reached roughly RMB 5.9 billion in 2025 and crossed 40% of revenue for the first time, with the components segment growing 48.58% at a 39.73% gross margin.⁶²⁷²⁹ This is not a promise about the future; it is a reported segment with a better margin than the legacy core.

The geographic hedge is built and operating. Ten production bases across seven countries, overseas revenue at 42.7% in 2025 and guided above 50% for 2026, with more than 99% of the roughly 2,000 overseas employees being local nationals — which matters for regulatory acceptance in an era when Chinese-origin telecom infrastructure attracts scrutiny.²⁸⁶³²

And the balance sheet sits inside a state framework that provides financing stability through a capital-intensive upcycle — a genuine advantage in a business where the ability to fund a two-year capacity build through a downturn separates survivors from consolidation targets.

Layer on operating leverage that UBS models as taking EBIT margin from 11.4% to 32.9% between 2025 and 2027, and the bull case is not a stretch.³¹ It is arithmetic.

The bear case, argued at equal strength.

The bear case does not require anything to go wrong. It requires only that the industry behave the way it has always behaved.

The supply response is already announced, and it is large. A laser-equipment manufacturer committed RMB 2.52 billion to build 60 million fiber-kilometers of annual capacity — an increment equal to roughly 8–10% of estimated global 2025 demand from a single new entrant.¹ Corning announced a 50% increase in U.S. capacity.¹ Zhongtian Technology planned 15–20% capacity increases starting in the second quarter of 2026; Hengtong has been ramping facilities in Inner Mongolia; FiberHome acquired capacity to reach 2,200 tons per year.¹² Because the preform build cycle runs 18 to 24 months, expansions initiated in 2026 land in 2027 and 2028 — and they land together.¹² Sell-side analysis has modeled the global market swinging into modest surplus around 2028 as a direct consequence.¹²

The demand driver has a terminus. AI data centers under construction consume enormous fiber. AI data centers already built consume replacement fiber, which is a fraction of installation demand. Unless the capital expenditure cycle extends indefinitely — and no capital expenditure cycle in industrial history has — the demand curve flattens and then falls back toward a maintenance run rate.

Technology could reduce glass intensity. This is the risk that did not exist in the last cycle. Corning's Glass Bridge, multi-core fiber, and hollow-core architectures all point the same direction: more data through less glass.¹²³ YOFC is a participant in that transition rather than a victim of it, but a company whose windfall depends on glass scarcity is structurally short any technology that relieves glass scarcity.

The valuation prices in a company that has never existed. Morgan Stanley's analysis implied that the post-rally share price embedded roughly RMB 10 billion of forward twelve-month net profit against a historical average multiple near 17 times — against a company whose best-ever full year before 2026 produced RMB 814 million.²³⁶ Even the peak first-half run rate does not clear that bar on an annualized basis without further price escalation. Meanwhile the trailing multiple briefly exceeded 400 times, and even post-crash sat above 300 times, against an industry historical median near 28 times.¹²

And the market has already demonstrated the fragility. A 47.52% decline in thirteen sessions, and a 56% peak-to-trough drawdown in the A-share, occurred with no negative operating news whatsoever — indeed, alongside a profit pre-announcement of up to 914% growth.¹² A stock that can lose more than half its value while beating expectations is a stock whose price is being set by narrative rather than by cash flow.

The activist stress test.

There is no known activist campaign at YOFC, and the ownership structure makes one structurally implausible. But the questions an activist would ask are the right questions regardless of whether anyone is positioned to force answers.

What happens to the windfall? A state-controlled capital allocator facing a national priority around domestic AI infrastructure self-sufficiency has every institutional incentive to build capacity aggressively — which is precisely the behavior that destroyed industry returns after 2018. The 2025 buyback and the 2026 dividend resumption are the counter-evidence, and they are modest in scale relative to the cash the current pricing generates.²⁹³⁰

Why did insiders sell? A shareholder above the 5% threshold executed disclosed sell-down tranches in the RMB 380–448 range during the run-up.²¹⁹ Related-party dynamics deserve attention here too: the Draka Shanghai buyout was formally classified as a related-party transaction because Draka remained a holder of YOFC H shares at the time of the announcement.²⁰

Is the disclosure adequate? Investors are being asked to underwrite hollow-core fiber leadership on the basis of company-reported laboratory results.³² An independent third-party validation of the attenuation and latency claims would be the single highest-value disclosure improvement available.

Is the segment reporting keeping pace? As diversified businesses cross 40% of revenue, the granularity of segment disclosure becomes materially more important than it was when fiber was 90% of the company.⁶

The synthesis.

Here is where the evidence actually lands. YOFC has a stronger, better-documented technical position than the median Chinese industrial cyclical — the preform self-sufficiency is real, the high-grade share is real, and the interconnect components business has genuinely different economics. Those are durable characteristics.

But the earnings currently being reported are not a function of those durable characteristics. They are a function of a price level set by a supply shortage that the entire industry is now, visibly and simultaneously, racing to eliminate. The technology shift toward higher-specification fiber may well be permanent. The pricing of that fiber almost certainly is not.

Investors are being asked to distinguish "structurally better business" from "cyclically extreme pricing." Those two things are currently arriving in the same income statement, and the stock's own 2026 behavior — a fourteen-fold rally followed by a halving, with no operating disappointment in between — suggests the market has not settled which one it is buying either.

↑ Back to Top

XII. Risk Radar

The risks that matter for this business are not generic macro concerns. They are specific, mechanical, and mostly traceable to the industry's own structure.

Demand-cycle risk. AI data-center capital expenditure is itself cyclical and sentiment-driven, concentrated among a small number of hyperscale buyers whose spending decisions are made annually and can be deferred quickly. A digestion phase — not a collapse, merely a pause — would hit fiber orders within a quarter, because fiber is purchased against construction schedules rather than against installed base. The 2019 demand reversal showed how fast a build-out-driven order book can evaporate once the build-out matures.²¹

Capacity-response and oversupply risk. This is the primary risk and it is already in motion. The specific expansions are announced, funded, and dated: new entrants adding capacity equivalent to a high single-digit percentage of global demand, Corning expanding U.S. output by half, and the major Chinese producers all ramping simultaneously.¹¹² The 18-to-24-month lag means the market will not feel this until 2027 and 2028, at which point it will feel it all at once.¹² Chinese preform utilization at 84.3% is the number to watch degrade.¹²

Technology-route risk. Two distinct versions. First, that hollow-core fiber fails to commercialize at the pace or cost management projects, stranding R&D investment and removing the narrative bridge to the post-shortage era. Second, and more dangerous, that a competing architecture — Corning's Glass Bridge, multi-core fiber, co-packaged optics — reduces the glass intensity of AI infrastructure and compresses volumes independent of any demand slowdown.¹²³ The mid-2026 selloff was triggered substantially by exactly this concern, which tells you the market considers it live rather than theoretical.

Geopolitical and export risk. Chinese-origin specialty fiber, submarine cable technology, and telecom infrastructure sit inside an active US-China technology-security conversation. Overseas manufacturing in Poland, Mexico, Indonesia, South Africa, Brazil, and Germany is a partial hedge — production inside a customer's jurisdiction is harder to restrict than an export from Wuhan — but it does not eliminate the risk, since the underlying process technology and preform supply remain Chinese in origin.²⁵³²

Valuation and re-rating risk. This deserves to be stated plainly as a standalone risk because it is independent of operations. A trailing multiple that briefly exceeded 400 times cannot be sustained on any normalized earnings basis, and a return toward historical industry multiples near 28 times would produce a severe drawdown even if every operating forecast were met exactly.¹ The 2026 experience is the proof: the largest decline in the company's listed history occurred during its best-ever operating period.

Governance and capital-allocation risk. Control by a central state-owned conglomerate means capital allocation may reasonably prioritize industrial-policy objectives — domestic preform self-sufficiency, national AI infrastructure capacity, employment in Hubei — over per-share returns.¹⁵ These objectives are not illegitimate. They are simply not identical to shareholder value maximization, and an investor should price the difference rather than assume it away.

Concentration and disclosure asides. Two smaller items worth tracking. Customer concentration remains meaningful despite diversification, since the Chinese carriers still represent a large block of domestic volume through centralized tenders whose pricing mechanics are designed to extract producer surplus.²² And accounting judgment increases in a price-spike environment: inventory valuation, revenue recognition timing on long-lead contracts, and the capitalization treatment of R&D running at 5%-plus of revenue all become more consequential when margins swing by twenty percentage points in four quarters.²⁹ No irregularity has been reported; the point is that the disclosure quality bar rises with the volatility.

Not a material risk, despite frequent mention. Input-cost inflation. In a business where the preform is roughly 70% of fiber cost and the preform's value is overwhelmingly process rather than materials, raw-material price movements are a second-order concern.¹⁰ Energy costs matter more than silica costs, and neither is currently the binding constraint.

↑ Back to Top

XIII. Durable Lessons and What to Watch

The investing lesson. The durable insight from YOFC's forty-year history is that a genuine technical moat and fundamentally cyclical economics are not mutually exclusive — and that conflating them is the specific error this industry has punished twice in a decade.

YOFC's preform capability is real. It was real in 2018, when the stock listed in Shanghai at the top of a cycle. It was real in 2020, when bare fiber sold for under three dollars per fiber-kilometer and the company's margins were compressed to a fraction of today's.²² It was real in 2024, when net profit fell nearly 48%.²³ The moat did not disappear during the bust and it did not appear during the boom. What changed was price.

A best-in-class supplier inside a commodity-adjacent industry earns a premium over its competitors through the cycle. It does not earn a growth-stock multiple, because the cycle does not go away just because the supplier is good. The mistake is not owning a cyclical; it is paying a secular multiple for cyclical earnings at the point of maximum cyclical earnings.

Three things to track, and only three.

KPI 1: High-end fiber pricing and volume, specifically G.654.E. This is the purest read available on whether the AI shortage is easing. G.654.E is the grade where YOFC's share is highest, the margin is richest, and the AI exposure is most direct.⁵ Price is a shortage indicator; volume is a demand indicator. Watching them together separates the two questions that matter most: is demand still growing, and is supply catching up? A single soft quarter proves nothing — pricing in centralized-tender markets is lumpy. A sustained decline across two or three reporting periods, especially with volume flat or rising, would indicate the supply response has arrived.

KPI 2: Diversified and optical-interconnect segment revenue and gross margin as a share of total. This tests the most important structural question about the company: whether YOFC is genuinely becoming less dependent on commodity fiber, or whether the diversified business is simply riding the identical AI capital-expenditure wave with a different product label. The tell is margin behavior in divergence. If interconnect gross margin holds while fiber pricing normalizes, the diversification is real and the business has been permanently repositioned. If the two fall together, the diversification was correlation, not hedging.⁶²⁹

KPI 3: Overseas revenue mix and overseas capacity utilization. The geographic hedge is management's most consistently articulated multi-year strategy and the primary buffer against both China-specific telecom cycles and geopolitical restriction.²⁵²⁸ Revenue mix alone is insufficient — a rising overseas share during a global shortage tells you little. Utilization at the overseas plants is the harder test, because it reveals whether those facilities are genuinely embedded in local supply chains or are swing capacity that runs hot only when the world is short.

What would change the thesis.

On the negative side: a sustained multi-quarter decline in high-grade fiber pricing rather than a single soft print; a major hyperscaler visibly dual-sourcing away from YOFC on a flagship deployment; evidence that the AI-2030 hollow-core programme is falling behind Corning or the Japanese specialists; or a large debt-funded preform capacity announcement that signals the company has chosen to fight the next cycle on volume.

On the positive side: independent third-party validation of the hollow-core performance claims; interconnect-components margins holding above the legacy fiber business through a pricing normalization; overseas utilization staying high after the shortage clears; and a capital-allocation pattern in which the windfall is visibly returned rather than reinvested into the industry's next overcapacity episode.

↑ Back to Top

XIV. Epilogue

As of late August 2026, YOFC occupies an unusual position: a company delivering the best operating results in its thirty-eight-year history whose shareholders have just lived through the sharpest drawdown in its listed history, with no operating disappointment in between.

The half-year accounts were approved and released in August with revenue up more than half, gross profit close to triple the prior year, and a resumed dividend.³⁰ The Chinese financial press, which had spent February writing about the wealth the stock created for its parent shareholder and August writing about a share price that had gone from RMB 49 to RMB 312 in a year, had by mid-2026 also written the other story — the one about a 47.52% collapse in thirteen sessions and a fiber narrative colliding with technology-route doubt and announced capacity expansion.¹⁸³⁸¹

Both stories are accurate. That is the problem.

What is not in dispute is the underlying achievement. A joint venture assembled in Wuhan in 1988 around five engineers sent to the Netherlands for three months of training became, over four decades, the largest and most technically self-sufficient optical preform manufacturer in the world — and did it by repeatedly refusing to stay in the low-value part of the chain it was originally licensed into.⁸

What is in dispute is whether that achievement translates into the earnings power currently embedded in the share price. The company is arguing, through hollow-core fiber and through a diversified components business now above 40% of revenue, that it has found a way to keep winning after the shortage ends. The market spent the first half of 2026 believing that argument enthusiastically and the second half interrogating it.

So the question the next two years will answer is the one this industry has posed before and answered badly: has YOFC actually escaped the boom-bust pattern of its own history — or is the AI data-center build-out simply the newest tailwind for an industry that has proven, twice now, that it cannot resist building its way out of a shortage?

Yangtze Optical Fibre And Cable Joint Stock Limited Company

Table of Contents

YOFC: The Fiber Behind the AI Boom

I. Introduction & Episode Roadmap

II. Origins: From State Joint Venture to Global Preform Leader

III. The Moat: Why Preforms, Not Fiber, Is the Business

IV. Going Public: A+H Listing and Who Actually Owns YOFC

V. The First Boom-Bust: China's Broadband Supercycle Turns to Bust (2013–2020)

VI. Recovery, Diversification, and Going Global (2021–2024)

VII. The AI Supercycle: 2025–2026

VIII. The Next Bet: Hollow-Core Fiber and "AI-2030"

IX. Current Management, Incentives, and Capital Allocation

X. Competitive Landscape and Industry Structure

XI. Bull vs. Bear: Stress-Testing the AI Infrastructure Thesis

XII. Risk Radar

XIII. Durable Lessons and What to Watch

XIV. Epilogue

References