Yuanjie Semiconductor: China's InP Laser Chip Champion and the AI Optical Wave

I. Introduction & Episode Roadmap (0:00 – 14:00)

On April 17, 2026, a rare milestone occurred on the Shanghai Stock Exchange. A company with fewer than 800 employees, headquartered not in Shanghai or Shenzhen but in Shaanxi province, overtook 贵州茅台 Kweichow Moutai to become the highest-priced stock in mainland China.¹ Moutai is a cultural institution — a baijiu producer with gross margins exceeding 90% and a brand predating the People's Republic. The company that displaced it manufactures a component roughly the size of a grain of coarse salt, sells it for about 22 yuan, and reported a net loss as recently as 2024.

That company is 陕西源杰半导体科技股份有限公司 Yuanjie Semiconductor Technology Co., Ltd. (SSE: 688498), a producer of indium phosphide laser chips — tiny light engines inside optical transceivers that convert electrical signals into photons. Modern AI training clusters rely heavily on these components. Yuanjie's shares reached an intraday record of RMB 1,766.88 before pulling back, giving the company a market capitalization of roughly RMB 140.5 billion by mid-May.² Less than four years after its initial public offering, its market value had expanded nearly thirtyfold.³

The financial results driving this valuation shift were swift and recent. In the first half of 2026, Yuanjie reported revenue of RMB 925 million, representing a 351.4% year-over-year increase, alongside net profit attributable to shareholders of RMB 607 million, up 1,212.2%.⁴ Data center laser chips accounted for RMB 774 million of that revenue — a 640.3% increase — while the telecom segment that previously comprised the entire business generated RMB 148 million.⁴ In six months, the company generated more than triple the profit recorded in all of 2025.

The bull case, stated plainly: Yuanjie is among a select group of global manufacturers capable of end-to-end indium phosphide laser chip production — spanning epitaxial crystal growth, wafer fabrication, chip processing, and testing — under one roof. It scaled continuous-wave light sources for silicon photonics just as hyperscalers ramped up purchases of 800G and 1.6T transceivers. Independent industry research commissioned for its Hong Kong listing ranked Yuanjie as the sixth-largest laser chip supplier globally by external sales in 2025, with a 3.1% market share, and the second-largest supplier of laser chips for silicon photonics-based high-speed interconnects, holding a 23.6% share.⁵ In a market led by American and Japanese incumbents, it ranked first among domestic Chinese suppliers across all three evaluated segments.

The bear case is equally concrete and detailed in the company's regulatory filings. Yuanjie's largest customer accounted for 53.4% of 2025 revenue, up from 16.4% in 2024, while its top five customers generated 71.8% of total sales.⁵ The integrated device manufacturer model that yields 71% gross margins on data center chips during a demand surge resulted in a net loss in 2024 when capacity utilization dropped to 68%.⁵ Furthermore, the company has not yet operated through a full downcycle in the AI market. In May 2026, amidst peak market valuation, the vice president of sales was taken into criminal detention.⁶

This story centers on a physics constraint that silicon cannot resolve, a materials scientist who spent a decade in California mastering crystal growth before establishing manufacturing in Xi'an, and the dynamics of a high-fixed-cost manufacturer navigating a sharp demand expansion in communications hardware. It serves as a case study for a central question facing technology investors: how much of a company's apparent competitive moat reflects durable differentiation versus a temporary supply shortage?

An examination of the company's historical claims — including its early cost-advantage narrative, the endorsement from Huawei, its entry into automotive LiDAR, and its customer diversification commitments — tests which strategies have survived contact with operational reality.

---

↑ Back to Top

II. The Physics & Industry Structure of Optical Laser Chips (14:00 – 32:00)

Start with a copper wire. For sixty years, moving bits between chips meant pushing electrons down a metal trace. That approach worked because distances were short and frequencies were low. Then AI happened. A modern training cluster is not a single computer; it is a warehouse full of accelerators that must operate as one cohesive machine, requiring every GPU to have a high-bandwidth, low-latency connection to every other GPU. Push a signal at 200 gigabits per second down a copper cable, however, and the physics turns hostile: attenuation increases with frequency, the cable heats up, and the signal degrades. Past a couple of meters, copper simply stops working.

Light does not suffer from these limitations. Photons traveling through glass experience modest loss over kilometers, carry vastly higher bandwidth, and remain entirely immune to electromagnetic interference. Consequently, the interconnect fabric of every AI data center relies on optics. At the front of every optical link sits a laser chip whose job is to convert electrical signals into a beam of light of precise wavelength, power, and stability — millions of times per second, for ten years, without drifting.

Why silicon can't do this

Silicon is a magnificent switch but a poor light emitter. Its band structure is indirect, meaning that when an excited electron in silicon drops to a lower energy state, it releases its energy overwhelmingly as heat rather than as a photon. As a result, efficient lasers cannot be manufactured from silicon. Instead, producers rely on compound semiconductors — materials assembled from two or more elements, primarily indium phosphide (InP) and gallium arsenide (GaAs). InP is the material of choice for the 1310-nanometer and 1550-nanometer wavelengths used in fiber optics.

That single physical constraint shapes the entire industry structure. In logic chips, progress stems from feature shrinkage executed by standardized lithography equipment. In laser chips, progress relies on growing crystals. The primary manufacturing process is metal-organic chemical vapor deposition (MOCVD), in which metal-organic gases decompose over a heated substrate to deposit atomic layers of semiconductor material, one at a time, controlling lattice spacing to a fraction of a nanometer. Slight deviations in chemical composition shift the emission wavelength, while subtle lattice strains propagate defects through the active region that cause premature failure in the field. Yuanjie's technology disclosures group its intellectual property into six technology platforms, led by "advanced epitaxial and wafer fabrication."⁵

The manufacturing process resembles metallurgy or winemaking rather than conventional semiconductor fabrication: output quality depends on proprietary recipes, specific reactor behaviors, and tacit engineering knowledge accumulated over thousands of production runs. That operational know-how cannot be easily documented or transferred.

Why the industry is IDM, and what that costs

This tight coupling explains why the fabless-foundry split that transformed logic manufacturing never materialized in compound lasers. In logic, design firms submit standardized files to foundries like TSMC. In compound lasers, the design is inextricably linked to the epitaxial recipe, waveguide geometry, facet coating, and burn-in protocol — each calibrated to specific manufacturing tools. As noted in Yuanjie's Hong Kong filing, given the scarcity of foundries with comprehensive process capabilities, most laser chip suppliers "should operate under an IDM model."⁵ Global market leaders — including Broadcom, Lumentum, Coherent, 三菱電機 Mitsubishi Electric, and 住友電工 Sumitomo Electric — all operate integrated device manufacturing facilities.

This model creates a cost structure with high fixed overhead, comparable to an airline's capital intensity. Depreciation, cleanroom utilities, equipment maintenance, and specialized engineering payroll remain fixed regardless of output. Yuanjie's regulatory disclosures detail this exposure: under a high-fixed-cost structure, "our unit production costs and overall profitability are highly sensitive to changes in capacity utilization," and the IDM model "may limit our flexibility to quickly reduce manufacturing-related costs in response to adverse market conditions, compared with a fabless model."⁵ This structural characteristic drives substantial margin expansion during demand surges but exposes the business to steep operational leverage during downturns.

The product ladder

Three primary product families define the market, forming a hierarchy from commodity components to scarce high-value devices.

FP and DFB lasers are directly modulated devices, where light output varies directly with fluctuations in drive current. They are simple, low-cost, and suitable for short-distance communications, such as fiber-to-the-home passive optical networks and 4G/5G wireless fronthaul. Yuanjie established its initial business in this segment, shipping 2.5G, 10G, and 25G DFB chips.

EMLs (electro-absorption modulated lasers) integrate a continuous-wave laser with an absorption modulator on a single chip. Rather than modulating the laser directly, an optical shutter section controls the light output, eliminating the wavelength "chirp" that restricts directly modulated lasers over longer distances. Consequently, EMLs are required for per-lane transmission speeds of 100G and above. A 400G optical transceiver using an EML architecture requires four 100G chips, whereas a 1.6T transceiver requires eight 200G EMLs.⁵ Dollar content per module increases with each network generation, enabling the market for these components to expand faster than overall transceiver unit volume.

CW lasers represent the newest and highest-growth segment. Silicon photonics uses an alternative architecture: modulators, waveguides, and multiplexers are fabricated in silicon at conventional CMOS foundries, powered by an external, unmodulated, continuous-wave light source. Because silicon photonics circuits exhibit optical insertion loss, that external CW light source must be high-power. A 400G silicon photonics module requires two 50mW or 70mW CW lasers, while a 1.6T module requires four chips operating at 70mW or 100mW.⁵ Next-generation near-package and co-packaged optics architectures push power requirements higher, toward 150mW, 300mW, and 400mW.

Industry research cited in Yuanjie's filing projects the continuous-wave laser market to expand at a 148.3% compound annual growth rate from 2024 to 2030, with high-power devices above 100 milliwatts reaching $6.6 billion, or 65.3% of the total market, by 2030.⁵ While these projections were prepared for capital-raising disclosures, the structural trend aligns with underlying architectural shifts: as silicon photonics adoption expands for integration and power efficiency, value shifts from discrete modulators to high-power light sources.

The economics hiding inside a laser chip

Understanding these physical and architectural dynamics clarifies how a component priced at roughly 22 yuan can generate a 71% gross margin during peak demand.

The primary margin driver is manufacturing yield and reliability qualification rather than sheer unit scale. A laser chip that functions only on a test bench has no commercial value; optical module manufacturers require components verified to maintain exact wavelength and power tolerances over a ten-year operational lifespan inside high-temperature data center environments. Establishing compliance requires extensive accelerated aging tests, lifetime modeling, and statistical failure analysis — capabilities Yuanjie explicitly outlined plans to expand in its regulatory filings.⁵ Consequently, while the marginal production cost of an additional compliant chip is low, the fixed cost of qualifying a reliable process is substantial, allowing established, qualified suppliers to capture premium margins while unaccredited competitors face high barriers to entry.

Furthermore, higher data rates drive non-linear growth in component value. Upgrading from 400G to 1.6T transceivers on an EML architecture doubles chip density from four to eight units per module while requiring higher per-lane speeds that command premium pricing. On a silicon photonics architecture, a 400G module uses two mid-power CW lasers, whereas a 1.6T module requires four higher-power units.⁵ Because dollar content per optical port increases with each speed generation, supplier revenue can outpace port volume growth — though missing a technology transition risks exclusion from the highest-margin market segments.

This technical and economic landscape forms the foundation of Yuanjie's expansion in China. Transitioning from legacy telecom components to high-speed data center laser chips required thirteen years of development, following a decade focused on lower-speed optical products.

↑ Back to Top

III. Founding Story & The Early Telecom Era (2013–2019) (32:00 – 48:00)

Consider the choice facing 张欣刚 Zhang Xingang in 2012. He held a bachelor's degree from 清华大学 Tsinghua University and a doctorate in materials science from the University of Southern California. He had spent more than a decade inside the American optical components industry, holding research and engineering roles at Luminent, LuminentOIC, and Source Photonics — companies that navigated the optical fiber expansion, the 2001 telecom crash, and the grinding consolidation that followed.⁵ By then a US citizen, he had built a comfortable career in Southern California mastering compound semiconductor fabrication.

Instead of staying, he moved to Shaanxi.

In January 2013, Zhang established Yuanjie Semiconductor as a limited liability company with registered capital of RMB 10 million — roughly $1.6 million at the time.⁵ Rather than setting up shop in Shanghai, Shenzhen, or the tech hubs of Suzhou where most returning semiconductor founders landed, he chose China's inland northwest. While the optoelectronics research base across Xi'an's universities and research institutes was deep, the commercial chip ecosystem there remained sparse.

The gap he was aiming at

The strategic logic was straightforward and, for its era, unfashionable. China had built an enormous optical module industry — companies assembling transceivers by the tens of millions — but relied almost entirely on foreign chip suppliers. The high-speed laser chips inside those modules came overwhelmingly from North American and Japanese vendors. Zhang recognized a critical domestic supply gap, though closing it would require substantial capital and years of patient development.

His response was to build the most challenging business model: a fully integrated device manufacturer (IDM) line. Rather than outsourcing epitaxy or licensing external recipes, Zhang insisted that if tacit process knowledge formed the primary competitive moat, the company had to own and operate its own reactors.

The early operational trajectory looks deceptively smooth in retrospect. In 2014, Yuanjie introduced its first 2.5G DFB lasers at 1490nm and 1550nm. Cumulative DFB shipments passed one million units in 2015, followed by 10G DFB lasers in 2016. The company achieved process breakthroughs for medium-to-high-speed chips in 2017, and cumulative DFB shipments passed ten million in 2018.⁵ Behind each milestone lay the unglamorous work of epitaxy engineering: reducing stubborn defect densities, managing thermal impedance that compromised performance at 70°C, stabilizing wafer-to-wafer yield variations, and establishing the process discipline required for commercial qualification.

Who backed him, and what that says

Yuanjie's shareholder register from those years reflects an unusual investor base. Disclosed in its Hong Kong listing prospectus, the promoter list spanned more than thirty names, including state-backed vehicles — a China Development Bank manufacturing transformation fund, a CDB science and technology venture arm, an SDIC technology-transfer fund, and 陕西先导光电集成科技投资 Shaanxi Xiandao Optoelectronic Integration Technology Investment, a regional optoelectronics investment vehicle — alongside 中信证券投资 CITIC Securities Investment and a tail of private partnerships.⁵

Yet the two largest individual shareholders after Zhang were neither institutional venture capitalists nor optoelectronics engineers. 秦为兴 Qin Weixing and his brother 秦彦生 Qin Yansheng between them held more A shares than the founder. Regulatory filings explain that Qin Weixing made an early-stage investment driven by his long-standing relationship with Zhang and confidence in the industry, with his brother co-investing on the same basis. Both were primarily engaged in manufacturing and selling automotive rubber components and focused on their own business operations, with Qin Weixing serving as a non-executive director without day-to-day management duties and Qin Yansheng holding no position at all.⁵

This structure illustrates a distinct feature of Yuanjie's governance. Early capital depended heavily on personal trust rather than institutional diligence. While the arrangement proved successful, it left Zhang's effective voting control tied to the continued alignment of two passive investors from an unrelated industry. Though formalized in a 2021 concert-party agreement, that voting coalition fundamentally rested on personal relationships rather than aligned commercial interests.

Commercial pull arrived through national policy. China's 宽带中国 Broadband China strategy and sequential fiber-to-the-home upgrade waves placed state telecom carriers into a decade-long rollout of passive optical networks, requiring a laser chip inside every home terminal. Yuanjie entered this supply chain by selling through module makers and network equipment vendors serving the state carriers. By 2019, it was launching 25G DFB chips in CWDM and LWDM variants aimed at 5G wireless fronthaul and, tentatively, at data centers.⁵

Historical falsification: did the early IDM model actually create a cost moat?

A core premise often advanced by market commentators is that Yuanjie's early integrated manufacturing model established a durable cost and yield advantage in telecom laser chips.

The record rejects that claim for low-speed products, supported by clean disclosure extracted during regulatory oversight. When Yuanjie's financial results collapsed in 2023, the Shanghai Stock Exchange issued an annual-report inquiry letter with seven lines of questioning. Among the extracted disclosures: the gross margin on the company's 2.5G product line fell from 48.49% to 12.18% in a single year.⁷ A 36-percentage-point margin collapse on a mature product is not a routine wobble; it reflects classic commoditization as domestic competitors reach adequate yields and compete strictly on price.

The broader telecom segment tells the same story more slowly. Even in the recovery year of 2024, when telecom volumes rose sharply, telecom gross margin fell to 14.3% from 26.7% in 2023 — an erosion the filing attributes directly to "pricing pressure under prevailing market conditions."⁵ Average selling prices for telecom chips dropped from RMB 4.7 per unit in 2023 to RMB 4.1 in 2024, recovering to RMB 5.8 in 2025 only because the company deliberately reallocated capacity toward higher-end parts.⁵

These results indicate that the IDM model did not generate pricing power in low-speed DFB chips. Instead, it provided operational capability — establishing the foundation to run a fab line and progressively master higher-speed devices. Meaningful pricing power emerged only at the top of the product ladder, where few suppliers could produce high-specification components at scale. The key indicator for evaluating whether this advantage persists is data center chip gross margin: if margins compress toward telecom levels as volumes expand, the underlying driver was temporary market scarcity rather than a structural process moat.

That strategic distinction remained latent until 2020, when an inquiry arrived from Shenzhen.

↑ Back to Top

IV. The Huawei Catalyst, IPO, & The Telecom Peak (2020–2022) (48:00 – 66:00)

By September 2020, 华为 Huawei was fourteen months into operating under U.S. Entity List restrictions and had concluded that its only durable response was to cultivate a domestic supply chain. Its venture capital arm, 哈勃科技创业投资 Hubble Technology Venture Capital, had been established the previous year with an unconventional mandate: acquire early minority stakes in the upstream Chinese component companies Huawei needed for its survival.

Hubble acquired a 4.36% stake in Yuanjie ahead of its listing, a position that diluted to 3.27% after the IPO.⁸ By the time Yuanjie went public, Hubble had already generated a series of listings from a portfolio barely three years old.⁹

What the endorsement was actually worth

It is easy to overstate the impact of this investment. Hubble's capital injection was modest in absolute terms and conferred no corporate control. Instead, it provided Yuanjie with two far more valuable assets: a qualification signal and a demand signal. In an industry where customers spend twelve to twenty-four months testing a new chip vendor before committing to volume orders, an equity investment and product adoption by China's premier communications equipment maker served as a shortcut through the trust barrier.

The company's financial disclosures illustrate both the immediate benefit and the ultimate limitation of that relationship. A customer described in regulatory filings as a Guangdong-headquartered ICT company serving telecom and datacom markets — matching the profile of a major domestic equipment vendor — was Yuanjie's largest client in 2023 at RMB 27.5 million, or 19.0% of revenue. By 2024, sales to that customer fell to RMB 24.1 million, representing 9.6% of revenue. By 2025, that customer dropped out of the top five entirely.⁵

This trajectory highlights a key misconception in the market: the narrative of a Huawei-anchored competitive moat. A strategic shareholder relationship did not insulate Yuanjie from the broader telecom downturn, did not drive its subsequent expansion into the AI cycle, and did not prevent a wholesale rotation of its customer base. The relationship provided critical validation at a time when credibility was scarce, but as with most timing-based advantages, its strategic impact naturally expired.

Scaling into the listing

Yuanjie entered its public listing on strong financial footing. Revenue reached RMB 283 million in 2022, with a gross margin of 61.90% and net profit of roughly RMB 100 million.⁷¹⁰ For a hardware component supplier, those software-like margins reflected a specific, temporary condition: China's domestic 10G PON upgrade cycle was at peak intensity, and Yuanjie was among the few domestic suppliers qualified into it.

Yuanjie listed on the 科创板 STAR Market on December 21, 2022 under stock code 688498, issuing 15 million shares — exactly 25% of post-listing capital — at RMB 100.66 per share.⁵¹¹ The offer price implied 69.26 times earnings, and the deal was heavily oversubscribed: against a plan to raise RMB 980 million, gross proceeds came to RMB 1.51 billion, representing an over-raise of more than RMB 500 million.¹¹ The four announced uses of proceeds were a 10G/25G optical chip production line, a 50G optical chip industrialization project, an R&D center, and working capital.¹¹ The listing valued the company at just over RMB 7.6 billion.¹²

What the company actually was in 2022

Examining the company's core operations at the time of its listing clarifies how market expectations evolved, as retrospective accounts often project its later AI positioning onto an earlier stage of development.

The earliest revenue segmentation provided in Yuanjie's Hong Kong listing disclosures covers 2023. In that year, telecommunications laser chips generated 92.1% of revenue. Data center laser chips accounted for 3.2% — RMB 4.6 million across 225,000 units.⁵ The remainder came from project-based technical services. At the time of its domestic IPO, Yuanjie operated primarily as a supplier of directly modulated lasers for Chinese fiber-to-the-home and 5G wireless fronthaul networks, while its data center business existed mainly as sample shipments and a strategic roadmap.

This operational baseline is critical for interpreting Yuanjie's initial public valuation. A multiple of nearly seventy times earnings was not paid for the business as it existed; it was paid for an option on the business becoming something else. Options of that nature carry substantial mispricing risk in both directions — a vulnerability that became apparent over the subsequent twelve months before the market repriced it again.

The incentive machinery

The company's equity incentive structure provides insight into its talent retention strategy within an industry where specialized human capital is the primary competitive asset.

Yuanjie adopted a share option plan in July 2021, layering restricted share schemes on top in 2024 and 2025. The initial option plan issued 599,000, 623,070, and 486,056 new A shares on successive March exercises in 2023, 2024, and 2025 before being fully exercised or lapsed. By the time of the Hong Kong filing, 884,030 restricted shares had been granted to 326 employees, including three directors and three senior managers, with a further 126,900 shares reserved.⁵

The key metric is the operational cost of this retention framework. Share-based payment expenses ran at roughly RMB 3.5 million annually in both 2023 and 2024. In 2025, that expense rose to RMB 32.2 million, widening the gap between the company's non-IFRS adjusted net profit of RMB 223.1 million and its reported net profit of RMB 190.9 million.⁵ Investors should treat that non-IFRS add-back with standard analytical skepticism: share-based compensation is a tangible cost of retaining epitaxy engineers in a market where competing firms bid for the same small talent pool, and it scales with equity valuation rather than top-line revenue.

Two features of the capital structure warrant attention. First, corporate control. Zhang holds his position not through a majority equity stake but through a concert-party agreement signed in January 2021 with his sister 张欣颖 Zhang Xinying and two brothers, 秦为兴 Qin Weixing and 秦彦生 Qin Yansheng, under which they align their votes with his. Together with an employee incentive vehicle for which Zhang serves as general partner, this coalition controlled approximately 27.85% of voting rights as of the Hong Kong filing.⁵ This structure grants effective control at a low ownership level — a common arrangement in domestic listings that provides minority shareholders limited operational leverage.

Second, the balance sheet the IPO created. Yuanjie ended 2023 with RMB 1.33 billion of cash and total liabilities equal to 5.4% of assets.⁵ That fortress balance sheet was about to become the essential buffer standing between the company and a severe cyclical contraction.

Because the 10G PON deployment cycle that generated those 62% gross margins was, at the moment of listing, already coming to an end.

↑ Back to Top

V. The 2023 Downturn: The Dark Side of IDM Operating Leverage (66:00 – 82:00)

The IPO bell rang on December 21, 2022. New orders stopped arriving in roughly the same quarter.

There is a particular cruelty to the optical component business cycle. Yuanjie sells to module makers, who sell to equipment vendors, who sell to state telecom carriers. When 中国移动 China Mobile, 中国电信 China Telecom and 中国联通 China Unicom completed their initial gigabit broadband targets and slowed procurement, the demand signal reached Yuanjie amplified by three layers of inventory correction. Downstream customers that had over-ordered during earlier component shortages abruptly halted purchases to burn through existing stock.

In 2023, revenue fell 48.96% to RMB 144 million, while net profit dropped 80.58% to RMB 19.5 million. Optical chip sales specifically contracted by 51.08% to RMB 138 million, driven by a 43.83% decline in the core telecom segment to RMB 133 million.¹⁰ On a domestic reporting basis, gross margin compressed from 61.90% to 40.40%.⁷

What the fixed-cost machine did

This sharp contraction highlighted the structural operational risks of integrated device manufacturing.

A semiconductor fabrication facility carries heavy fixed overhead regardless of production volume: equipment depreciation continues, cleanroom environmental controls operate continuously, and specialized epitaxy engineers must be retained to preserve proprietary process knowledge. In 2023, Yuanjie's capacity utilization dropped to 74.2%, yet fixed manufacturing expenses — primarily depreciation and facility overhead — still accounted for 56.1% of cost of sales.⁵ Compounding the margin pressure, the company booked RMB 15.7 million in inventory impairments, representing 15.3% of cost of sales, to write down unsold chips.⁵

Under international financial reporting standards, gross margin fell to 28.7% in 2023 and 23.4% in 2024, culminating in a net loss of RMB 6.1 million in 2024 — even as revenue rebounded 74.6%.⁵ Despite top-line growth of nearly three-quarters, profitability deteriorated as R&D spending increased to RMB 54.5 million, selling expenses more than doubled, and capacity utilization dropped further to 68.1% while production lines were taken offline and reconfigured for continuous-wave laser production.⁵

The company relied on its balance sheet to bridge the shortfall. Operating cash flow registered a negative RMB 17.4 million in 2023 and a modest positive RMB 19.0 million in 2024, while cash reserves decreased from RMB 1.41 billion at the beginning of 2023 to RMB 1.08 billion by late 2024.⁵ Without the capital raised during its IPO, Yuanjie would have faced external financing needs during a severe cyclical downturn.

The regulator's seven questions

On June 19, 2024, the Shanghai Stock Exchange issued an annual-report inquiry letter to Yuanjie covering seven operational and financial areas — a standard regulatory mechanism that elicited details omitted from routine reporting.⁷

The resulting disclosures revealed significant operational strain. Sales to Yuanjie's top five customers had fallen from RMB 161 million to RMB 77 million. Accounts receivable reached RMB 114 million — equivalent to 79% of full-year revenue — alongside an aging collection profile, which management attributed to financial distress among its customer base. Inventory rose to RMB 158 million, or 109.72% of annual revenue, as inventory turnover slowed from 1.82 times in 2021 to 0.71 times in 2023. Additionally, regulators queried the placement of RMB 30 million in IPO proceeds into a private investment fund, requesting clarification on the investment's purpose and potential related-party connections.⁷

While individual disclosures were not fatal to the business, collectively they indicated severe working-capital deterioration. High receivables and inventory ratios provided early operational signals that component shipments had outpaced downstream carrier demand.

What management did with the worst year in the company's history

Despite the sharp contraction in revenue, management avoided defensive cost-cutting. Rather than reducing R&D or deferring capital expenditures to support short-term profitability, Yuanjie expanded technical investments. R&D spending rose from RMB 30.9 million in 2023 to RMB 54.5 million in 2024 and RMB 80.8 million in 2025 — representing a 161% increase across a period that included an operating loss.⁵ Capital expenditures similarly accelerated, with investing cash outflows rising from RMB 37.8 million in 2023 to RMB 231.7 million in 2024.⁵ Furthermore, company filings clarify that the low 68.1% capacity utilization in 2024 was partially deliberate: fabrication lines were temporarily taken offline and reconfigured to manufacture continuous-wave lasers for silicon photonics applications.⁵

By allocating capital from its IPO proceeds toward unproven high-speed optical products during a severe market slump, management prioritized long-term technology development over short-term earnings. That structural pivot directly enabled the company's rapid scaling during the subsequent 2025–2026 AI expansion.

However, this aggressive investment strategy presents longer-term capital allocation risks. Early capital deployment was vindicated when a RMB 232 million investment was backed by a RMB 1.4 billion cash reserve. Applying the same capital commitment model to far larger outlays — such as the RMB 4.268 billion commitment announced in August 2026 — represents a significantly higher risk profile relative to annual corporate revenue.

Historical falsification: the IDM "defensive cash flow" claim

The 2023–2024 downturn refutes the thesis that vertical integration provides Yuanjie with structural margin stability, a dynamic explicitly acknowledged in the company's risk disclosures.⁵ Operating an IDM model acts as a double-edged leverage amplifier, expanding margins during demand surges while depressing profitability when capacity utilization falls. A more sustainable rationale for the IDM model rests on supply reliability and process iteration speed rather than margin defense. That refined advantage would face challenge if future capacity expansions replicate the 2024 pattern of top-line growth accompanied by stagnant or negative net earnings.

Historical falsification: the automotive LiDAR optionality

In February 2023, amidst declining telecom orders, Yuanjie announced to investors that its product portfolio encompassed 1550-nanometer automotive LiDAR laser chips.¹³ Financial analysts positioned the technology as a second high-margin growth driver.

However, commercial adoption failed to materialize at scale. In 2022, the year of the product announcement, related revenue reached just RMB 1.26 million, categorized under ancillary business income.¹³ By the time Yuanjie submitted its Hong Kong listing application proof in March 2026, its reported revenue structure was divided into just two core segments — data center laser chips and telecommunications laser chips — alongside a minor technical services category that generated RMB 1.7 million in 2025.⁵ Automotive LiDAR was omitted entirely as a standalone reporting line.

This trajectory invalidates early growth assumptions surrounding LiDAR diversification. While Yuanjie demonstrated early technical capability and delivered sample shipments, product validation did not translate into meaningful commercial revenue. This outcome underscores a critical operational reality for compound semiconductor manufacturers: qualifying a prototype does not guarantee commercial market entry. Consequently, Yuanjie's track record warrants caution when evaluating future product roadmap claims — including higher-power co-packaged optics light sources, which remain technological milestones rather than established revenue streams.

Ultimately, Yuanjie's recovery was driven not by expansion into adjacent markets, but by the rapid resurgence of its core optical business in a data center market expanding faster than industry forecasts.

↑ Back to Top

VI. The AI Pivot: 100G EML & High-Power CW Lasers (2024–2026) (82:00 – 102:00)

Yuanjie's strategic pivot began in 2020, when corporate disclosures record that the company "introduced high-power CW lasers designed for SiPh applications."⁵ At the time, silicon photonics remained a contested architecture with a history of delayed commercial deployment, making product development in the segment an uncertain commitment for a manufacturer with RMB 100 million in revenue. In 2021, the company completed electro-absorption modulated laser (EML) development, addressing the long-reach modulation constraints that had previously limited its access to high-speed data centers.⁵ In 2023, it introduced 100G PAM4 EML chips.⁵

Demand subsequently expanded as AI cluster deployments accelerated.

The mechanics of the surge

The financial mechanics of this expansion are evident in the company's unit economics.

Data center chip sales grew from 225,000 units in 2023 to 2.3 million in 2024, reaching 17.7 million in 2025 — a nearly eightyfold expansion over two years.⁵ Average selling prices remained firm across the period, moving from RMB 20.5 to RMB 21.0 and then RMB 22.2 per chip.⁵ A massive volume expansion accompanied by stable-to-rising unit prices indicates a supply-constrained market environment rather than price-discounted market-share acquisition.

Data center laser chip gross margins expanded accordingly, rising from 25.2% in 2023 to 62.1% in 2024 and 71.3% in 2025.⁵ Capacity utilization climbed from 68.1% to 90.1% as designed capacity nearly doubled.⁵ Positive operating leverage enabled 2025 total revenue of RMB 601 million — a 138.5% increase — to generate net profit of RMB 191 million, reversing the net loss recorded in 2024.⁵¹⁴ Data center sales reached RMB 393 million, up 719.1%, accounting for over half of total corporate revenue for the first time.¹⁵

In 2024, Yuanjie achieved volume production of 70mW CW lasers, shipping over one million units.⁵ By 2026, the company was supplying 50mW, 70mW, and 100mW silicon photonics light sources alongside high-speed DFB and EML components, positioning it in regulatory filings as one of few global suppliers capable of delivering continuous-wave lasers at a ten-million-unit scale.⁵

Financial growth continued into 2026. On July 21, ahead of its formal interim report, Yuanjie pre-announced first-half net profit of RMB 600 million to RMB 650 million — a year-over-year increase of 1,196.91% to 1,304.98% — attributing the surge to expanding data center revenues and a favorable product mix that lifted blended gross margins.¹⁹ The interim financial report published on August 28 confirmed results within that projected range, matching management's preliminary guidance.

First-half revenue of RMB 925 million exceeded total full-year revenue for 2025. Second-quarter revenue reached RMB 570 million, yielding RMB 428 million in net profit.⁴ Operating cash flow rose to RMB 488 million, up 729.7% year-over-year, indicating that earnings converted into cash flow rather than accumulating in accounts receivable as occurred during the 2023 downturn.⁴ Data center products represented approximately 84% of total revenue during the period.⁴

What management actually said, and how it should be weighed

At the annual shareholders' meeting on April 23, 2026, Zhang characterized the industry environment by stating, "AI demand growth is exponential, but capacity expansion is linear."¹⁶ He explained that production capacity coming online in 2026 reflected capital allocation decisions made in 2024 based on conservative demand estimates, adding that management expected elevated demand to persist through 2027 and 2028. Board secretary 程硕 Cheng Shuo noted that new equipment arriving throughout 2026 would expand output "quarter by quarter."¹⁶

Two aspects of management's commentary provide critical analytical context. First, Zhang stated that profitability was driven primarily by product scarcity and manufacturing yields rather than raw material costs, noting that raw materials — including indium, despite recent price appreciation — represented only 10% to 20% of total production costs.¹⁶ This structure indicates that current gross margins reflect a temporary scarcity premium rather than a structural raw-material cost advantage.

Second, regarding pricing strategy, Zhang indicated that existing customers would not face price increases, though new customers might be subject to higher rates.¹⁶ Strategically, this approach prioritizes long-term customer relationship stability over immediate pricing realization. However, it also implies that unit price stability relies partly on commercial policy choices, leaving margins vulnerable should key customers seek price concessions in future renegotiations.

Capital allocation: the Xixian bet

On August 10, 2026, the board approved a RMB 4.268 billion investment plan to construct a semiconductor industrial park in Xi'an's 西咸新区 Xixian New Area. The project encompasses laser chip fabrication lines, manufacturing facilities, and supporting infrastructure across approximately 126 mu of land under a 50-year industrial lease, with a planned 24-month construction timeline funded through internal resources and self-raised capital.¹⁷ At the time of announcement, shareholder approval was pending, and required land-use, environmental, and construction permits remained outstanding.¹⁷

This project represents the company's largest capital commitment to date. The RMB 4.268 billion budget is approximately seven times 2025 total revenue and substantially exceeds the company's cash reserves, which stood at RMB 796 million at year-end 2025.⁵ Management is initiating this expansion during an unprecedented demand surge based on recent hyperscaler procurement volumes. However, while previous capacity additions in 2024 proved insufficient for subsequent demand, committing capital at this scale introduces structural risk if long-term market growth falls short of current projections.

From a bullish perspective, securing manufacturing capacity for 2028 requires committing capital during a two-year construction lead time to avoid losing market share during supply shortages. Conversely, the bear perspective views this as a classic semiconductor industry risk: treating a temporary supply deficit as a structural condition and bringing heavy fixed capacity online just as industry-wide supply catches up. As demonstrated during the 2024 downturn, underutilized fabrication assets can quickly depress corporate profitability through fixed depreciation overhead.

Going to Hong Kong

On March 25, 2026, Yuanjie filed a listing application proof with the Hong Kong Stock Exchange for a proposed H-share offering, appointing Guotai Junan Capital and GF Capital (Hong Kong) as joint sponsors.⁵ At the time of submission, the company's mainland-listed A shares had appreciated 582% over the preceding twelve months, briefly ranking Yuanjie as the third-highest-priced stock in the domestic market before it reached the top position in April.²⁰ Disclosed uses of proceeds included purchasing fabrication equipment for epitaxy, lithography, etching, polishing, dicing, and testing; acquiring land and constructing manufacturing and R&D facilities; and expanding research capabilities.⁵ Pricing details and offer size were omitted from the preliminary draft.

Beyond capital formation, the proposed dual listing reflects broader international expansion plans. The filing disclosed a Singapore subsidiary established in June 2023 and a U.S. subsidiary, YST Semiconductor Technology Corporation, incorporated in March 2024, alongside a U.S. production facility noted as being "in the initial stage of construction" and not yet operational.⁵ Establishing offshore capital access, international manufacturing capacity, and an overseas shareholder base aligns with serving a global AI supply chain primarily driven by North American cloud operators, though operating across geopolitical boundaries introduces ongoing regulatory and trade policy considerations.

↑ Back to Top

VII. Competitive Landscape & Deep Dive on Core Business (102:00 – 118:00)

Strip away the narrative, and the competitive landscape in optical laser chips resembles a small room with only a handful of key players.

Independent research commissioned for Yuanjie's Hong Kong listing estimated the global external-sales laser chip market at RMB 19.35 billion in 2025 — slightly under US$3 billion.⁵ The top four suppliers captured 57.9% of that total. Ranked by 2025 external revenue, the industry leaders comprised: a U.S. photonics company founded in 2015 and listed on Nasdaq, serving AI, cloud, telecom, industrial, and 3D sensing markets, at 16.7%; a U.S. semiconductor and infrastructure software conglomerate founded in 1961, at 14.5%; a Japanese diversified electrical manufacturer founded in 1921, at 13.4%; and a Japanese materials and electronics manufacturer founded in 1897, at 13.3%. Additionally, a vertically integrated U.S. photonics platform founded in 1971 held 3.7%. Yuanjie ranked sixth globally with a 3.1% market share, placing first among domestic Chinese suppliers.⁵ While regulatory filings anonymize these competitors, their corporate profiles correspond directly to well-known Western and Japanese incumbents.

Two other Chinese suppliers made the global top ten — a private Guangdong firm founded in 2004 offering components through to board-level solutions at 2.9%, and two Hubei-based optical chip specialists founded in 2018 and 2016 holding 1.5% and 0.9%, respectively.⁵ Other domestic peers, including 光迅科技 Accelink in integrated telecom components, Source Photonics in vertically integrated transceivers, and 长光华芯 Everbright Photonics in high-power semiconductor lasers, compete across adjacent market segments.

Where the position is actually strong

That aggregate ranking understates Yuanjie's position in the highest-value market segment. In laser chips for high-speed optical interconnects — encompassing 100G-and-above EMLs and continuous-wave lasers operating at 50 milliwatts or higher, which represented an RMB 8.73 billion market in 2025 — Yuanjie ranked sixth globally with a 4.3% share, trailing the same four global incumbents plus one additional competitor.⁵

However, in the specific market for silicon photonics light sources, valued at RMB 1.59 billion, Yuanjie ranked second globally with a 23.6% share, trailing only the Japanese materials group at 31.7% while outranking both major American competitors.⁵

That single metric encapsulates the core commercial thesis. Yuanjie is not merely a general-purpose laser chip maker operating in China; it is a specialist that committed early to silicon photonics light sources, constructing dedicated fabrication lines in 2020 when the architecture was still commercially unproven, and securing a top-two global position in the fastest-growing niche within optical interconnects.

How it wins

Several distinct operational mechanisms support this market position:

Scarcity and scale. Regulatory disclosures explain the 2025 customer expansion by noting that Yuanjie's primary customer was among the first to deploy silicon photonics at scale, and that Yuanjie's continuous-wave capability and available fab capacity "positioned us as a key supplier to meet Customer F's growing demand during the global CW laser supply shortage."⁵ That candid assessment indicates that the surge was driven primarily by an industry-wide supply deficit where Yuanjie possessed available inventory and production capability.

Process iteration speed. Operating an integrated device manufacturing line allows custom engineering requests — such as modified epitaxial layer structures or tailored facet coatings — to yield test samples within weeks rather than quarters. Legacy Western incumbents serving standardized global product lines are structurally slower to configure bespoke process runs for individual module assemblers.

Cost structure. Fab construction, cleanroom utilities, and engineering compensation in China remain materially lower than in the United States or Japan. With raw materials representing only 10% to 20% of total production costs, Yuanjie's cost base benefits directly from these localized operational efficiencies.¹⁶

Supply chain security. For Chinese optical module makers supplying both domestic and international markets, qualifying a domestic chip source eliminates export-license dependencies. That structural preference offers a durable advantage in China, though it conversely complicates direct sales efforts to Western end-customers.

How it loses

Conversely, several structural vulnerabilities threaten this market position:

200G per lane yield hurdles. Although Yuanjie lists 200G PAM4 EML chips in both single-ended and differential-drive configurations in its technical literature, commercial product literature does not equate to volume qualification.⁵ High-speed 1.6T transceivers require eight 200G EML chips per module, a technology generation where established global competitors possess longer operational records. If Yuanjie's manufacturing yields for 200G components lag, its strong position in current 100G architectures risks obsolescence as module makers transition to higher speeds.

Architectural shifts. Should co-packaged optics fundamentally alter light-source packaging and specifications, or if thin-film lithium niobate modulators and heterogeneously integrated on-chip lasers capture market share, the addressable market for discrete external light sources could contract relative to current projections.

Customer vertical integration. Yuanjie's regulatory filings explicitly highlight the risk of customer disintermediation, noting that competition will intensify "as certain optical transceiver manufacturers increasingly develop in-house chip capabilities" to build full-chain systems.⁵ Downstream module makers possess the capital and incentive to internalize chip production, capturing the 71% gross margins currently earned by component suppliers.

Incumbent expansion. The four leading global incumbents retain substantially larger balance sheets and deeper R&D budgets. A coordinated capacity expansion or aggressive pricing response from established competitors represents the standard market mechanism by which temporary scarcity margins are eroded.

The domestic substitution question, examined properly

In domestic financial research, Yuanjie's growth trajectory is frequently framed around "domestic substitution" — the premise that Chinese optical module manufacturers will systematically replace Western laser chips with local alternatives, guaranteeing multi-year market share gains.

That thesis warrants rigorous examination, as empirical evidence supports two opposing conclusions.

In favor of the thesis, Chinese module manufacturers that qualify domestic chip suppliers eliminate supply-chain export-license risks that could disrupt production — an operational buffer worth securing even at price parity. Yuanjie's filings note that the company maintains a diversified supplier network across China, Europe, and Japan while proactively building domestic supplier relationships,⁵ mirroring the supply-chain hedging strategy implemented one level up its own value chain.

Against the thesis, domestic substitution functions as a procurement policy dictated by buyers, rather than an irreversible structural shift. Module makers that qualified Yuanjie during peak supply shortages retain their legacy Western vendors; as supply constraints ease, buyers gain the flexibility to dual-source and leverage price competition. Furthermore, substitution logic explains initial vendor qualification but does not justify the retention of 71% gross margins over time. Established global incumbents still control 65% of the market for high-speed 100G-and-above EML chips,⁵ driven primarily by technical capability and yield performance rather than buyer inertia.

Consequently, domestic substitution provides a reliable tailwind for shipment volumes but an unreliable mechanism for preserving profit margins. Investors relying on the substitution narrative must monitor realized unit pricing rather than gross volume growth alone.

Yuanjie's competitive advantage remains concentrated within a specific, high-growth product segment. That structural concentration leads directly to the uncomfortable arithmetic of who is actually buying.

---

↑ Back to Top

VIII. Playbook: Business & Investing Lessons (118:00 – 130:00)

1. Hamilton Helmer's 7 Powers

Process Power — present, and the strongest of the set, but narrower than usually claimed. Thirteen years of epitaxy iteration produce process knowledge that cannot be easily replicated. The evidence of its existence is clear: Yuanjie scaled continuous-wave lasers to a ten-million-unit volume when few competitors could match that capacity, driving data center gross margins to 71.3% without price reductions. The evidence of its limits is equally clear: that same manufacturing capability yielded just a 12.18% gross margin in 2.5G DFB lasers.⁷ Process power in this industry is node-specific and decays as individual product generations commoditize. It must be re-earned with every technical transition, making the ongoing 200G EML qualification timeline far more decisive for long-term value than past success in 100G.

Cornered Resource — weak, and weakening. The specialist indium phosphide engineering talent pool in China remains thin, and Yuanjie has secured a notable share of it — 535 of its 716 employees worked in manufacturing roles at the end of 2025.⁵ However, its intellectual property position is modest for a company claiming technical leadership, comprising 36 registered domestic patents, 11 pending applications, and four overseas patents.⁵ Defensibility relies on proprietary trade secrets and team cohesion rather than an expansive patent portfolio. Expanding domestic competitors pose a direct threat to both talent retention and process exclusivity.

Counter-Positioning — moderate, and mostly geographic. Operating as a nimble, localized supplier to Chinese optical module manufacturers establishes a distinct position that Western incumbents cannot replicate without altering their global service models. This is not a model Western incumbents cannot adopt; rather, it is one they are reluctant to pursue given their established cost structures.

Switching Costs — moderate and asymmetric. Qualification timelines spanning twelve to twenty-four months create substantial operational friction. However, key buyers have strong commercial incentives to dual-source. During supply shortages, this friction operates in reverse: customers actively qualify alternative vendors specifically to mitigate single-source vulnerability.

Scale Economies, Network Economies, Branding — absent. Generating RMB 601 million in 2025 revenue against global incumbents operating at several times that scale places Yuanjie at a structural scale disadvantage. Laser chip manufacturing features no network effects, and brand recognition carries little weight with procurement engineers evaluating technical qualification reports.

2. Porter's Five Forces

Buyer power — high, and getting higher. Generating 53.4% of total revenue from a single customer transforms that buyer from a routine client into a central counterparty upon whom the entire financial trajectory depends.⁵ Management framed this reality at the annual shareholders' meeting by noting that AI infrastructure demand may concentrate around a small group of large operators, making deep single-account integration preferable to pursuing a fragmented client base.¹⁶ While commercially logical, that perspective acknowledges structural buyer power rather than refuting it.

Supplier power — moderate and rising. Purchases from the company's top five suppliers accounted for 48.1% of total procurement in 2025.⁵ Indium phosphide substrates and metal-organic precursors are sourced from a concentrated group of global chemical and materials suppliers, alongside rising market prices for raw indium. While raw materials represent only 10% to 20% of production costs today, MOCVD epitaxy reactor manufacturing remains a critical equipment bottleneck concentrated among a few suppliers in Europe and the United States.

Substitutes — high over the long term. Heterogeneously integrated lasers, thin-film lithium niobate modulators, and architectural shifts toward co-packaged optics all threaten to reduce or alter demand for discrete external light sources over time.

New entrants — low to moderate. Entry barriers remain formidable: mastering end-to-end manufacturing processes, navigating multi-year qualification cycles, and committing capital at scale — illustrated by Yuanjie's RMB 4.268 billion industrial park plan — deter casual entrants.¹⁷ Nevertheless, regulatory filings explicitly highlight that new domestic competitors are actively seeking market entry.⁵

Rivalry — intense and structurally asymmetric. Yuanjie competes against established global entities possessing deeper capital reserves, longer operating histories, and broader product portfolios in an industry where the top four suppliers control 57.9% of global market share.⁵

3. Historical falsification: the customer diversification claim

The thesis that Yuanjie has built a diversified customer base across end markets is directly contradicted by its regulatory disclosures.

Top-five customer concentration expanded from 55.6% of revenue in 2023 to 59.1% in 2024, reaching 71.8% in 2025. Concentration in the company's single largest customer rose from 19.0% to 16.4%, before surging to 53.4% in 2025.⁵ Furthermore, four of the top five customers from 2023 were replaced or demoted by 2025. Customer concentration did not merely persist through the AI expansion — the surge in AI demand directly caused it.

Regulatory filings describe this dominant buyer as a Shandong-headquartered, Shenzhen-listed optical interconnect transceiver manufacturer expanding into intelligent automotive systems, with a commercial relationship dating to 2018.⁵ That profile corresponds to 中际旭创 Zhongji Innolight, though the filing omits the company name, requiring readers to treat the identification as an informed analytical inference.

Management's assertion that customer concentration will ease as overall fabrication capacity expands¹⁶ represents a testable prediction rather than an established outcome. That projection runs counter to an industry structure where a small group of transceiver manufacturers commands the vast majority of market demand. The core conclusion supported by empirical evidence is clear: Yuanjie operates as a concentrated component supplier to a concentrated downstream market, making its revenue stability dependent on the procurement policy of a single primary buyer. The key metric to evaluate management's diversification narrative is the single-customer revenue share disclosed in future reporting periods. Unless that share falls below 40% alongside sustained top-line expansion, claims of customer diversification remain unsupported by operational performance.

↑ Back to Top

IX. Analysis & Bear vs. Bull Stress Test (130:00 – 142:00)

Imagine a skeptical investor with no position, reading the 2026 interim report for the first time. What would they attack?

The governance question they would ask first

On May 14, 2026 — three weeks after the stock became the most expensive in China — Yuanjie announced that vice president 陈文君 Chen Wenjun had been placed in criminal detention by public security authorities on suspicion of criminal offences and was unable to perform his duties.⁶¹⁸ Chen's responsibilities covered product sales and marketing, and he held 308,600 shares. The board convened the same day and removed him from his executive role, effective immediately, while stating that the company had not received official notice from police and that production and operations remained normal.¹⁸

As of September 2026, the alleged conduct has not been publicly specified. That absence of clarity is precisely why it matters. An investor cannot yet distinguish between a personal matter unrelated to corporate operations and an issue touching the sales function of a company whose revenue is dominated by a single customer relationship. While the board's response was procedurally swift, the lack of disclosed detail remains an unresolved question. It sits alongside the 2024 exchange inquiry as a second instance in three years where Yuanjie's corporate disclosures were expanded under external pressure rather than volunteered.

The concentration and cyclicality attack

The skeptic's core argument is straightforward: Yuanjie's current high profitability reflects a temporary supply shortage. When industry capacity catches up — driven by Yuanjie's own RMB 4.268 billion industrial park, expanded incumbent output, and new domestic entrants — data center gross margins of 71.3% will naturally normalize. If margin compression occurs just as depreciation on RMB 4.268 billion in new assets begins to hit the financial statements, the company could experience a rerun of 2024 at ten times the scale: rising revenue alongside stagnant or declining net profits.

The counter-argument holds that AI interconnect demand follows an unprecedented growth trajectory, that chip content per transceiver increases with each speed generation, and that a two-year construction lag means capacity arriving in 2028 will enter a market several times larger than today's. Both positions are consistent with available evidence. In truth, neither market participants nor management can predict the outcome with certainty — a reality Zhang acknowledged when noting that the 2024 capacity expansion was based on demand forecasts that proved far too conservative.¹⁶

The supply chain and geopolitics attack

Yuanjie's fabrication facilities depend on imported capital equipment — including MOCVD reactors, lithography, etching, and testing tools — sourced from Europe, the United States, and Japan. The company's risk disclosures explicitly acknowledge that its products or manufacturing inputs could become subject to export controls or licensing requirements in jurisdictions including the U.S. and the European Union.⁵ Management maintains a supplier network spanning China, Europe, and Japan while actively developing domestic alternatives.⁵

This geopolitical exposure operates in both directions, highlighted by the U.S. production facility currently under construction.⁵ A Chinese semiconductor producer building manufacturing capacity in the United States while supplying the American AI hardware chain sits directly at the intersection of volatile trade policies. Rather than a distant tail risk, this dynamic represents an active variable in the company's capital allocation strategy.

The related-party and disclosure attack

An activist investor would also highlight customer-supplier overlaps. Three counterparties maintain dual relationships with Yuanjie — purchasing laser chips while simultaneously providing contract packaging services. One generated RMB 21.1 million in 2025 sales while accounting for RMB 5.7 million of Yuanjie's purchases; another generated RMB 11.9 million in sales against RMB 4.0 million in purchases.⁵ Management explains that these entities possess specialized packaging capabilities that Yuanjie outsources to, reflecting standard industry practice.⁵ While transaction amounts are modest and disclosure is adequate, circular commercial arrangements within a rapidly scaling revenue stream warrant ongoing scrutiny, particularly given how quickly customer balance sheets came under strain during the 2023 downturn.

Skeptics would also point to the RMB 30 million private fund investment that drew regulatory inquiry in 2024⁷, while questioning the composition of the RMB 404 million in investing cash outflows recorded in 2025 beyond direct capital expenditures.⁵

Myth versus reality

Four consensus narratives about Yuanjie circulate widely enough to evaluate against empirical disclosures.

"Yuanjie broke a foreign monopoly in high-speed laser chips." Partially true, but heavily qualified. Yuanjie holds a 3.1% share of the global laser chip market and 4.3% of the high-speed segment.⁵ Four Western and Japanese incumbents still control 57.9% and 85.5% of those respective markets. The monopoly that has truly been disrupted is in silicon photonics light sources, where Yuanjie holds a 23.6% market share and ranks second globally — a major achievement within a specific product category, rather than across the broader semiconductor sector.

"The Huawei investment gave Yuanjie a protected position." Unsubstantiated. Hubble held a 4.36% stake prior to the IPO, and sales to the customer matching that profile contracted from 19.0% of revenue in 2023 to below the top-five disclosure threshold by 2025.⁵⁸ The recent AI expansion was driven by an entirely different customer base.

"The IDM model means stable margins." Refuted by the company's 2024 net loss and its own risk disclosures, which state that an integrated manufacturing model restricts cost flexibility during industry downturns.⁵

"Gross margin above 70% proves pricing power." Unsupported. Management attributes high gross margins to temporary market scarcity and manufacturing yields rather than structural cost advantages, with raw materials representing only 10% to 20% of production costs.¹⁶ True pricing power allows a firm to raise prices without sacrificing volume; scarcity rent allows a firm to hold prices firm during supply deficits. Disclosed evidence supports the latter, while management's commitment not to raise prices on existing customers reflects a supplier prioritizing long-term account stability over short-term margin maximization.

The reality surviving all four tests is more nuanced: Yuanjie made an early, targeted architectural bet on silicon photonics in 2020 and executed well enough to become an essential supplier during a severe global supply shortage.

The bull's strongest ground

The bullish case relies on operational arithmetic rather than sentiment. A nearly 79-fold expansion in data center chip volumes alongside stable unit pricing cannot be manufactured through accounting adjustments. Operating cash flow of RMB 488 million in the first half of 2026 — up 729.7% year-over-year — confirms that reported earnings reflect genuine cash generation.⁴ Securing a second-place global market position in silicon photonics light sources alongside major global competitors represents the tangible payoff from capital decisions made in 2020.⁵ Furthermore, the company entered this growth cycle with minimal debt, maintaining total liabilities at 9.8% of assets at year-end 2025.⁵

The most defensible synthesis is clear. Yuanjie has demonstrated specialized technical capabilities in the fastest-growing segment of optical interconnects and converted that position into cash flow faster than most peer hardware manufacturers. However, it has not yet proven that its elevated margins represent more than temporary scarcity rents, that its customer concentration will naturally abate, or that its capital allocation framework can withstand a prolonged downcycle while managing heavy new facility depreciation.

The three KPIs that matter

Beyond short-term sentiment, three operational metrics will determine the company's long-term trajectory:

1. Largest-customer revenue share. Disclosed in periodic filings, this metric serves simultaneously as a measure of concentration risk, a test of diversification progress, and an early indicator of upstream dual-sourcing decisions. A declining share driven by revenue growth in secondary accounts validates management's strategy; a decline caused by reduced purchasing from the primary customer signals a structural break in the core investment thesis.

2. Data center gross margin. This serves as the primary gauge of supply scarcity. So long as data center chip margins remain near 70%, supply remains tight and pricing holds. Sustained margin compression toward historical telecom levels near 25% — even amidst expanding unit volumes — indicates that industry capacity has caught up, triggering a downward re-rating of the earnings base.

3. Capacity utilization. Disclosed annually within the Hong Kong reporting framework, utilization metrics will reveal whether new capital investments are being absorbed efficiently. With RMB 4.268 billion in new assets slated for deployment, low utilization will manifest here before impacting the income statement through depreciation. The 68.1% capacity utilization recorded in 2024 serves as the benchmark for operational underperformance.

Two secondary metrics warrant close monitoring: the qualification and commercial shipping timeline for 200G-per-lane EML chips, which determines whether Yuanjie participates in the 1.6T transceiver deployment cycle; and inventory and accounts receivable relative to revenue, which provided early warning of financial stress prior to the 2023 downturn.

↑ Back to Top

X. Epilogue & Summary (142:00 – 148:00)

There is a version of this story that gets told in Chinese financial media: a patriotic returnee builds a chip company in the inland northwest, breaks a foreign monopoly, and is vindicated when his stock overtakes the national liquor champion.

The factual record is more nuanced than the myth — and considerably more instructive.

What actually occurred is that a materials scientist with deep process expertise spent a decade building manufacturing capabilities in a product category — low-speed telecom lasers — that offered almost no pricing power, watching that legacy business commoditize until its 2.5G product line earned just a 12% gross margin. In 2020, well before commercial deployment was assured, he committed the company's scarce capital to continuous-wave light sources for silicon photonics, an architecture that had spent two decades as an unproven emerging technology. In 2023, the original telecom business contracted by half, the company booked inventory write-downs, drew a seven-part regulatory inquiry from its exchange, and avoided severe financial distress primarily because of cash reserves accumulated during its oversubscribed initial public offering. In 2024, the company recorded a net loss even as top-line revenue rebounded 75%.

Then the strategic bet paid off. That outcome stemmed not from an impenetrable competitive moat or a strategic shareholder relationship, but from a severe global shortage of high-power continuous-wave lasers colliding with a manufacturer that had spent four years preparing to produce them at scale.

For founders, the experience highlights a key imperative in compound semiconductors: technical capability is merely the price of admission, while excess returns depend on anticipating architectural transitions and having qualified manufacturing capacity ready when demand arrives. For investors, the lesson is more demanding. Yuanjie's current high profitability reflects a temporary supply-demand imbalance that management explicitly acknowledges. The RMB 4.268 billion investment now committed to the Xixian industrial park will determine whether the company builds a durable global franchise or serves as a case study in over-capitalizing a cyclical peak. That trajectory will not be decided by corporate guidance, but by three key metrics tracked over time: single-customer concentration, data center gross margins, and manufacturing capacity utilization.

Optical interconnects will continue to advance to higher speeds. Whether Yuanjie's returns can withstand the next industry capacity wave remains the central, unresolved question.

---

References

1. 源杰科技登顶"股王"！A股十大高价股洗牌 — 新浪财经, 2026-04-19 ↩

2. A股股王突发，源杰科技副总经理被刑拘 — 金融界, 2026-05-14 ↩

3. 源杰科技拟42.68亿建半导体产业园 四年不到市值飙升28倍筹划赴港上市 — ZAKER, 2026-08 ↩

4. 源杰科技：上半年归母净利润6.07亿元 同比增超12倍 — 搜狐财经, 2026-08-28 ↩↩↩↩↩↩

5. Application Proof of Yuanjie Semiconductor Technology Co., Ltd. — The Stock Exchange of Hong Kong, 2026-03-25 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

6. "股王"源杰科技：副总经理陈文君被刑拘 — 中国新闻网, 2026-05-14 ↩↩

7. 源杰科技去年业绩断崖式下滑，交易所连发七问，公司这样回应 — 界面新闻, 2024-06-19 ↩↩↩↩↩↩↩

8. 源杰科技IPO：华为哈勃投资入股，主力产品单价下滑 — 新浪财经, 2022-09-08 ↩↩

9. 做投资3年 华为拿下第10个IPO — C114通信网 ↩

10. 源杰科技2023年营收1.44亿元，同比下降48.96% — C114通信网, 2024 ↩↩

11. 今日开启申购！源杰科技超募5亿发行价超百元 — OFweek光通讯网, 2022-12 ↩↩↩

12. 源杰科技登陆科创板 最新市值超76亿 — C114通信网, 2022-12-21 ↩

13. 源杰科技：在车载激光雷达领域，公司的主要产品涵盖1550波段车载激光雷达激光器芯片等产品 — 每日经济新闻, 2023-02-08 ↩↩

14. 源杰科技2025年营收6.01亿元，同比增长138.50% — C114通信网, 2026-03 ↩

15. 源杰科技2025年营收同比增长138.5% 数据中心业务占比提升 — 证券时报, 2026-03-24 ↩

16. 直击"新股王"源杰科技股东会：AI需求"指数级增长" 设备到场后产能将逐季释放 — 财联社, 2026-04-23 ↩↩↩↩↩↩↩↩↩

17. 42.68亿元！源杰科技，有大动作 — 新浪财经, 2026-08-12 ↩↩↩

18. 源杰科技副总经理陈文君被刑拘 — 法治日报, 2026-05-15 ↩↩

19. 源杰科技：上半年净利润同比预增1196.91%至1304.98% — 证券时报, 2026-07-21 ↩

20. 源杰科技赴港IPO强化研发及产能 股价一年涨582%成A股第三高价股 — 新浪财经, 2026-03-13 ↩