Accelink Technologies: AI Datacom Growth Racing Against a Stretched Valuation

The story it told the capital market at the time was also clear: a systematic opportunity for domestic substitution in China's optoelectronic components, rather than a new category with a single explosive product. The major foreign competitors listed in the prospectus were JDSU, Bookham, and Avanex, while products were mainly subsystems and passive optical components. The document also mentioned that the company's products had gradually gained recognition from major customers such as Alcatel-Lucent, Ericsson, Nokia-Siemens, and Nortel. In other words, Accelink's core capability before listing was to place domestic components into the global mainstream communications chain, not to build a single blockbuster product.

After listing, this main line did not change. The 2025 annual report states directly that since listing, the company's principal business has had "no change" and remains the R&D, production, and sale of optoelectronic components, modules, and subsystems. What changed is the environment. More than a decade ago, it faced carrier network expansion and import substitution. Today, it faces AI data centers, DCI, 800G, and 1.6T. The business base has not changed; the growth engine has.

Main-Line Changes Across Development Stages

If Accelink's past path is compressed into four stages, the first was the component-platform stage under a research-institute system. At that time, the company's task was to first build foundational capabilities in components, passive products, and subsystems, rather than to run the fastest, so that domestic products could secure a position in the optical communications supply chain. Early competitors were mostly established overseas vendors. That decided the company's initial path: broad categories, stable certification, first entry and then expansion. It could not start as a highly focused, high-gross-margin single premium line.

The second stage was the scaling stage in the carrier and access-network era. The company gradually expanded its business into transmission, access, wireless fronthaul, and other directions, forming the prototype of its later platform operations. Looking back today, the most important legacy of this stage was the customer structure, stable delivery capability, and product coverage depth, rather than any single product. It meant that in every later industry upgrade, Accelink did not have to find a position from zero. It could add next-generation products onto the existing system.

The third stage was the move from telecom dominance toward "telecom + datacom in parallel." The 2025 annual report data already shows the result of this shift: transmission business revenue was CNY 3.412 billion, up 9.97% year on year, while data and access business revenue reached CNY 8.463 billion, up 65.89% year on year, and had already become the absolute main engine. This change means Accelink is no longer merely a reflection of telecom investment. It has begun to increasingly reflect data-center capex and AI cluster interconnect demand. As the revenue structure changed, the capital market's valuation label changed with it.

The fourth stage is the current "AI datacom repricing" stage. In its annual report, the company disclosed that data-center high-speed optical modules already cover 100G to 1.6T. Its 800G products include VR8, DR8, 2xFR4, 2xLR4, LRO, and LPO, while 1.6T products already cover DR8. Under Omdia's measure, the company's rolling four-quarter global optical components share rose to 5.9%, and its datacom optical components rose to fourth globally. Both technology and share gave the market reasons for repricing. After that, the first-quarter report confirmed that growth was continuing. The share price rose sharply in the short term, and the narrative changed from "steady platform" to "AI challenger."

Longitudinal Financial Review

From the statements, Accelink's latest fundamental improvement is not an illusion. In 2023, the company recorded revenue of CNY 6.061 billion and net profit attributable to shareholders of CNY 619 million. In 2024, revenue was CNY 8.272 billion and net profit was CNY 661 million. In 2025, these jumped to revenue of CNY 11.929 billion and net profit of CNY 946 million. Revenue nearly doubled over two years, but the slope of profit improvement was still clearly slower than that of more pure AI datacom companies such as Zhongji Innolight and Eoptolink. This is exactly the difference in growth quality and explosive elasticity between a platform company and a purer AI datacom company. The former is steadier; the latter is more forceful.

The cash-flow signal is more worth watching. In 2024, the company's net operating cash flow was CNY -641 million, completely diverging from CNY 661 million of net profit. By 2025, net operating cash flow rebounded to CNY 1.628 billion, not only turning positive but also substantially exceeding net profit. Aggregating 2023-2025, operating cash flow/net profit was about 0.91x. That is not especially beautiful, but it also does not describe a company whose accounting profit cannot be converted into cash. The real issue is that Accelink's profit recovery came with heavier working-capital occupation and expansion cadence. Therefore, one cannot look only at the rise in net profit; inventory, prepayments, and capital expenditure must also be watched.

In 2025, the company's R&D expense was CNY 1.039 billion, up 45.75% year on year. Capital-expenditure cash for "cash paid to acquire and construct fixed assets, intangible assets, and other long-term assets" was CNY 825 million, higher than CNY 707 million in 2024. By Q1 2026, inventory further rose to CNY 7.478 billion, up 30.14% from the start of the year. Prepayments rose 219.85%, construction in progress rose 37.47%, and the company also pointed to an increase in prepaid equipment payments. All of this shows that the company is putting real money into production lines, equipment, and inventory preparation, rather than passively receiving AI dividends. For a manufacturing company, this is both the prelude to growth and the first risk point that will show if demand falls short of expectations.

Turning Points in Share Price and Valuation History

For a long time, Accelink looked more like a "carrier-chain optical components platform," and its valuation rarely enjoyed the extreme elasticity of a pure AI datacom leader. The true turning point came after the 2025 annual report and the Q1 2026 report. The annual report proved that datacom and access had replaced transmission as the main axis of growth. The first-quarter report proved this was not a one-off spike. The company then continued in May communications to point the next 2-3 years of growth drivers toward high-speed optical modules and high-end optical components. The capital market therefore rediscounted its future cash flows and examined it through a lens closer to that used for AI optical module companies.

But the share price moved much faster than profit. As of 2026-06-12, the Eastmoney Choice page showed the company's closing price at CNY 204.97, corresponding to PE (TTM) of 159.59x, static PE of 174.72x, price-to-book of 15.99x, and price-to-sales of 13.25x. This is already a valuation level clearly carrying theme reinforcement and identity rewrite, not a "cyclical manufacturing growth stock valuation" in the usual sense.

Business Model and Moat

How the Revenue Machine Works

Accelink's revenue machine today is built around "three lines running together": transmission, access, and datacom. The annual report shows that in 2025, transmission product revenue was CNY 3.412 billion, accounting for 28.6%; data and access revenue was CNY 8.463 billion, accounting for 70.94%. From product descriptions, the transmission business covers 100G and 400G long-haul modules, amplifiers, WSS, coherent components, and other products. The access business covers XGSPON, GPON, 10G PON, 25G/50G/100G fronthaul modules. The data communications business covers high-speed optical modules and AOC used in data centers and AI intelligent-computing centers. In other words, Accelink tries to put most important links in optical communications into its own product catalog, rather than betting on one lane.

The advantage of this model is that when the industry rotates, it is less likely to be flattened by a single cycle. In 2025, transmission business grew only 9.97%, but data and access grew 65.89%, allowing the company as a whole to lift revenue by 44.2%. The company also uses direct sales across all customers, domestic and overseas, so there is no sluggish signal caused by layer upon layer of distribution. The drawback is also clear: a business can be broad enough, but that often means the hottest segment is not necessarily the deepest one. When AI is at its hottest, the market is more willing to give greater elasticity to companies such as Zhongji Innolight and Eoptolink, whose profit is almost entirely tied to AI cloud capex.

Cost Structure and Operating Leverage

Accelink's cost structure makes it a typical manufacturing operating-leverage company. In 2025, the gross margin of communications equipment manufacturing was 23.31%, up 1.22 percentage points from the prior year. Selling expense was CNY 299 million, administrative expense was CNY 210 million, and R&D expense was CNY 1.039 billion, implying an R&D expense ratio of about 8.7%. This means profit first requires R&D, people, equipment, production lines, and certification to be loaded up front. Margins improve only when orders truly scale, unlike an asset-light software model where profit drops through once revenue rises.

This is also why I do not look only at net profit in valuation. In 2025, the company generated CNY 1.628 billion of operating cash flow, but capital expenditure consumed CNY 825 million. On the most conservative measure, owner earnings were only a little over CNY 800 million. In other words, on the surface, the company's 2025 EPS had already reached around CNY 1.2, but the cash that could truly settle freely for shareholders was far less ample than accounting profit suggests. As long as the company is still expanding production, expanding customers, and improving 800G/1.6T delivery capability, this state of "profit improving but free cash flow not necessarily moving in sync" will continue.

Industry Position, Cyclicality, and the Real Moat

On industry momentum, Accelink itself has already given a very clear background in its annual report. Citing Omdia data, the company said that as of Q3 2025, global optical components market rolling four-quarter revenue reached USD 24.3 billion, up more than 50% year on year. Among this, the data communications optical components market was USD 12.9 billion, accounting for 68% of the global total; telecom optical components were about USD 4.8 billion, and access optical components were about USD 1.2 billion. The company also cited Omdia's forecast for 2030: the global optical components market is expected to reach USD 90.9 billion, with a CAGR of about 13% over the next six years. This is a typical track where the capital-expenditure cycle, inventory cycle, and technology-iteration cycle are coupled. On the upswing, the most beneficial variable is speed upgrades driven by cloud vendors and AI clusters. On the downswing, the most fragile variables are high-speed optical module ASP, inventory levels, and expansion cadence.

In my view, Accelink's real moat has three parts. The first is product breadth and platform capability. It covers transmission, access, datacom, a group of passive components, amplifiers, and optical waveguide integrated products, rather than merely shipping one 800G DR8. This makes it easier to become a "one-stop supplier" in front of carriers, equipment vendors, and some integrated customers. The second is certification and delivery reliability. The optical module business looks on the surface like a specification-sheet competition, but in reality it competes on long-term certification, yield, volume delivery, and failure-rate control. Once these capabilities enter a major customer system, they create relatively strong stickiness. The third is the state-owned system and telecom-chain accumulation, which is especially important in the carrier and domestic equipment-vendor markets.

But there are also some "moats" that the market has amplified too much, and I do not think they have been proven. The most typical is "platform breadth equals pricing power in AI datacom." In the overseas hyperscaler high-speed module market, the real moat is often faster product cadence, larger delivery volume, deeper customer certification, and stronger coordination with upstream chips, packaging, and silicon photonics. Accelink is certainly chasing, but from the 2025 geographic revenue structure, overseas revenue was still only 26.69%, materially below the market's imagination of a "core global AI optical module player." In other words, its moat holds in an integrated optical communications platform. In the hottest small part of pure AI datacom, it has not yet been fully validated.

Management and Governance

By Q1 2026, among Accelink's top ten shareholders, FiberHome Technologies Group held 36.13%, and China Information and Communication Technologies Group held 2.10%; the two are parties acting in concert. In the public communication record from May 2026, attendees included chairman Huang Xuanze, director and general manager Hu Qianggao, and CFO and board secretary Xiang Ming. The meaning of this governance structure is direct: the company has a strong industrial background and system resources, especially natural advantages in the carrier chain and domestic industrial collaboration. But it is not a company led by a founder who can place a high-speed bet on one single track. State-owned platform governance usually means steadier execution and more balanced cadence. It also means that when the capital market most prefers a single-line explosive narrative, valuation usually needs a higher evidence threshold.

Horizontal Competitor Analysis

A Numbers Table First Shows the Gap

The key meaning of the numbers in the table is the profit-scale gap, not whose share price is higher. Accelink's 2025 profit has already improved significantly, but compared with Zhongji Innolight and Eoptolink, the profit scale is still one order of magnitude smaller. If the market prices it as an "AI optical module platform," it must answer one question: over the next few years, how much will this profit gap actually narrow?

What Each Company Has Become

Zhongji Innolight has become the most standard "pure high-end datacom leader." Its financial profile is almost exactly the type the market loves to see: 2025 revenue of CNY 38.240 billion, net profit of CNY 10.797 billion, and operating cash flow of CNY 10.896 billion. Its products cover 100G, 200G, 400G, 800G, and 1.6T, and it clearly serves cloud-computing data centers, data communications, 5G wireless, telecom transmission, and fixed-network access. Why do customers choose it? Because it has eaten the hottest demand most deeply, and scale and delivery capability have already become self-reinforcing. Compared with it, Accelink's advantage is more balance; its weakness is clearly insufficient profit elasticity in pure AI datacom.

Eoptolink has become another version: more aggressive, more frontier-oriented, and more complete in form factors. Its 2025 revenue was CNY 24.842 billion, net profit was CNY 9.532 billion, and operating cash flow was CNY 7.701 billion. Public disclosures are very clear: the company not only covers 400G, 800G, and 1.6T, but has also laid out LPO, LRO, XPO, NPO, CPO, and launched OCS equipped with self-developed MEMS chips. More importantly, in investor communications it has clearly stated that 1.6T product orders have increased significantly compared with last year and are expected to grow rapidly quarter by quarter in 2026. At the same time, the company believes CPO will not become mainstream in the 800G phase, and 1.6T pluggable modules will remain the market's main line. What this kind of company is best at is "fast cadence," rather than "complete product coverage." Compared with Eoptolink, Accelink has similar platform depth, but on frontier cadence and overseas major-customer penetration, the market is clearly more willing to place chips on Eoptolink.

Changxinbochuang plays a different role. It is more like a segmented leverage name inside AI optical interconnect, rather than a mirror-image competitor to Accelink. In 2025, the company recorded revenue of CNY 2.533 billion and net profit attributable to shareholders of CNY 335 million. Among this, data communications, consumer electronics, and industrial interconnect business revenue was CNY 2.039 billion, up 89.86% year on year, already accounting for 80.5% of total revenue, while the telecom market business declined. Its decisive levers are AEC, AWG, connectivity, and local amplification after the consolidation of Changxinsheng, not "comprehensiveness." Compared with Accelink, it is narrower, but also more likely to produce high elasticity in a specific sub-track.

Coherent and Lumentum provide another global reference. Coherent's fiscal 2025 networking business revenue was USD 3.421 billion, up 49% year on year, with networking business profit of USD 644 million. Its annual report clearly states that the company makes not only transceivers, but also VCSELs, DMLs, EMLs, CW lasers, ICs, and passive optical components, and internally has GaAs and InP wafer fabs in multiple locations. In fiscal Q3 2026, the company recorded revenue of USD 1.81 billion, up 21% year on year, with management directly attributing growth to exceptionally strong demand in datacenter and communications. Lumentum emphasized in its 2025 annual report that it makes not only high-speed optical transceivers, but also key optical module components, narrow-linewidth lasers, VCSEL arrays, and OCS. By fiscal Q3 2026, Lumentum revenue reached USD 808.4 million, up 90% year on year, and management specifically said subsequent profit elasticity would also come from OCS and CPO. Why do customers choose these companies? Because they have not only products, but also key materials, chips, components, and global factory networks. Compared with them, Accelink has strong domestic system capability, but still has distance in globalized capacity, overseas customer mindshare, and voice in upstream key links.

Accelink's Ecological Position

Putting these companies together, Accelink's ecological position is clear. In the China market, it is a platform-type leader. In the most core global AI datacom profit pool, it is more a strong challenger than the dominant player with the greatest pricing power. It fills the gap of "understands carriers and access while also being able to make high-speed data-center modules." The profit pool it most directly competes for is the domestic integrated optical communications platform pool, not the highest-share global hyperscaler pool already highly bound to Zhongji Innolight, Eoptolink, Coherent, Lumentum, and others.

This also determines a very important judgment. If the industry continues to grow at high speed, Accelink will benefit, but the way it benefits is more like "growing with the broader optical communications market and domestic computing-power buildout." If the industry enters a price war or inventory adjustment, its platform revenue can be more resilient than that of a pure AI leader, but share-price elasticity will also be compressed earlier. If the market treats it as a steady platform, the valuation center can move up. If the market treats it as the second or third pure AI optical module leader, the mismatch risk will be very large.

Current Fundamentals, Valuation, and Risks

What Has Actually Happened in Recent Quarters

This cadence shows two things. First, the company's growth did not immediately peak after the 2025 annual report; 2026Q1 still maintained year-on-year revenue and profit growth. Second, although cash flow has already repaired, quarterly volatility remains large, showing that delivery, collections, and inventory preparation still follow a manufacturing cadence rather than an asset-light model of linear growth. The market is therefore betting that "Q2-Q4 will continue to accelerate," not that "Q1 has already completed the delivery."

The theme being traded by the market now can almost be compressed into one sentence: can Accelink become the stock that "finally closes the valuation gap" in this round of A-share AI optical module trading? In its abnormal-volatility announcement, the company disclosed that the share price had cumulatively risen 77.92% from the end of April to May 20. At the same time, it emphasized in investor communications that growth drivers over the next 2-3 years would still come from high-speed optical modules and high-end optical components. This combination easily leads capital to view it as a target where "profit improvement has been confirmed, while share catch-up has not yet been fully priced in."

But there is still a gap between real fundamentals and market narrative. The real fundamentals are: data and access grew fast in 2025, growth continued in 2026Q1, and inventory and prepayments also rose at the same time. The market narrative often directly extrapolates these facts into "overseas cloud-vendor share will rise materially, margins will quickly approach those of the leaders, and valuation should be close to pure AI leaders." The former has happened; the latter still needs to be proven.

Where the Bull-Bear Split Is Most Concentrated

The strongest evidence for bulls is that the industry and the company are resonating in the same direction. On the industry side, under Omdia's measure, the data communications optical components market reached USD 12.9 billion in the first three quarters of 2025, accounting for 68% of the global total. 800G has been fully commercialized, and 1.6T is being introduced. On the company side, 2025 data and access revenue grew 65.89%, 2026Q1 net profit grew 59.76% year on year, and management continued to list high-speed optical modules as the core driver for the next two to three years. As long as this line continues, Accelink is gradually producing the financial trajectory a growth stock should have, rather than merely "telling a story."

The core bear evidence lies in "price and position may not match," not in "the company is bad." As of 2026-06-12, Accelink's PE (TTM) had reached 159.59x and static PE 174.72x. On the same day, Zhongji Innolight's TTM PE was about 85.72x, while Changxinbochuang's forward PE was about 127.43x. For Eoptolink, because different data providers used different total-share-capital bases after the 2025 bonus share issue, public market-cap and PE pages are not fully consistent. But regardless of which basis is used, its valuation level is not as high as Accelink's range. This shows capital is probably pricing Accelink on the basis of "may close the gap with other leaders in the future," rather than "current earnings power." As long as the pace of gap-closing falls short of expectations, valuation will be more fragile than earnings.

Valuation Analysis

There is no need to circle around historical valuation. Based on 2025 net profit of CNY 946 million and the closing market capitalization of CNY 165.344 billion on 2026-06-12, the company's current static valuation is already typical high-expectation pricing. Looking through to cash flow: 2025 operating cash flow was CNY 1.628 billion, capital expenditure was CNY 825 million, and owner earnings on the most conservative measure were about CNY 802 million. Converted by current total share capital, owner earnings per share were about CNY 0.99, which means the share price is about 206x owner earnings, and the owner-earnings yield is only about 0.5%. The company has not disclosed a precise split between maintenance capex and expansion capex. I can only judge from industrial-park investment, equipment prepayments, and inventory expansion that the expansion portion of capex is clearly higher than maintenance at this stage. Therefore, valuing the company on net profit will be more optimistic than the real shareholder cash return.

Peer valuation comparison also gives an uncomfortable conclusion. Zhongji Innolight and Eoptolink already delivered 2025 profits of CNY 10.797 billion and CNY 9.532 billion respectively, with fuller evidence in 1.6T, LPO, OCS, and overseas major-customer penetration. Accelink's 2025 profit was only CNY 946 million, yet the market has attached a very high-multiple "AI datacom challenger" premium to it in advance. In other words, what Accelink has now is "high valuation waiting for homework to be made up," not "low valuation waiting for delivery." The risk-reward profiles of these two states are completely different.

The table below is scenario valuation under the research framework, not investment advice. To avoid being misled by accounting profit, I use owner earnings as the main valuation basis and then constrain it with high multiples acceptable for growth stocks.

The direction of these scenarios is clear: even after giving quite generous growth assumptions and relatively high growth-stock multiples, the current price still sits above the bull scenario. The truly fragile assumption is that "the company will quickly turn platform breadth into share depth and profit depth in pure AI datacom." If that assumption is marked down by 30%, the base-case value would move back from around CNY 94 to the CNY 74-78 range.

The margin-of-safety review is even simpler. The current price is at a significant premium to my bear-case intrinsic value, not a discount, so the margin of safety is zero. If profit is completely flat over the next three years, the current 174.72x static PE implies a yield of only about 0.57%, clearly below the 1.74% yield on China's 10-year government bond on 2026-06-12. This is the classic case of a "good company, but bad price." Therefore, the disciplined conclusion for this part can only be: there is no margin of safety.

Risks, Catalysts, and Tracking Dashboard

The first risk that could cause permanent capital loss is that the price and share trajectory for AI high-speed optical modules is not as smooth as the market imagines. I assign the probability as "medium" and the impact as "high." The indicators are straightforward: data and access revenue growth, consolidated gross margin, 1.6T introduction cadence, and the expansion and product cadence of Zhongji Innolight and Eoptolink. The transmission path is also direct. If 800G/1.6T unit prices fall faster than volume rises, or if overseas major-customer certification advances slower than expected, Accelink's revenue can still grow, but margins may not rise. At a TTM PE above 160x, the market has very low tolerance for "growth, but not fast enough."

The second risk is working capital swallowing cash. The probability is "medium" and the impact is "high." In 2026Q1, inventory rose materially both year on year and quarter on quarter, while prepayments, accounts payable, and contract liabilities all rose together. This shows the company is in a heavy-investment phase of stocking, expanding capacity, and rushing delivery. If the cycle continues, these are the prelude to growth. If demand turns, they become inventory and cash-flow problems. What needs to be watched is OCF/net profit, inventory/revenue, contract liabilities, and prepaid equipment payments, not the income statement alone.

The third risk is valuation compression. The probability is "high," and the impact is also "high." The valuation the market gives Accelink now already contains a high-intensity theme premium, not a normal valuation for a traditional manufacturing growth stock. Once sector preference rotates back to low-price, high-delivery names, or industry sentiment moves from "watch the catch-up" back to "watch already-delivered leaders," the company may suffer valuation mean reversion even if it continues to grow. This risk cannot be hedged by fundamentals alone.

The fourth risk is the speed of proof for overseas customers and global supply capability. The probability is "medium" and the impact is "medium-high." From the 2025 structure, the company's overseas revenue share was still below 30%. What truly widened the valuation of Zhongji Innolight and Eoptolink is scale shipment already formed in overseas hyperscalers and high-value AI switching networks, not merely having products. If Accelink's overseas breakthrough is slower than expected over the next two years, the market will eventually reprice it under the framework of an "integrated platform" rather than a "core supplier to AI cloud vendors."

Positive catalysts are not lacking: continued quarterly earnings beats, rapid uplift in 1.6T revenue share, consolidated gross margin stabilizing above 25%, and a sustained increase in overseas revenue share would all make the market willing to keep raising Accelink's role in AI datacom. Negative catalysts are also clear: more conservative guidance, inventory continuing to rise, cash flow deteriorating again, or peers obtaining clearer overseas major-customer incremental orders in 1.6T/OCS.

The point of tracking these indicators is to break the abstract phrase "AI narrative overheating" into financial and operating signals that can be checked. The most important question is whether high-end product ramp-up simultaneously brings gross-margin improvement, cash-flow improvement, and inventory-turnover improvement, rather than whether profit is a little higher or lower in one quarter. If there is only revenue without these three things, the valuation story is not solid.

Cross-Sectional and Longitudinal Synthesis

What Accelink has truly proved over its journey is the ability to "avoid falling behind in every cycle," rather than the ability to be the most aggressive in every cycle. From a domestic components platform under the Posts and Telecommunications Research Institute system, to a large product matrix covering transmission, access, and datacom, and then to today's ability to keep up with the 800G and 1.6T product generations, the company has proved a capability that capital markets rarely price patiently: platform-type survivability. It can cross the shift in industry focus from carrier network expansion to cloud data centers, maintain a meaningful position in the domestic optical communications supply chain, and quickly lift revenue when industry momentum returns. The 2025 annual report and 2026Q1 have already shown that this company has not been left behind by the AI optical communications wave.

But platform-type survivability is not the same as pricing power in pure AI datacom. Its past success came more from product breadth, customer stickiness, state-owned system resources, and long-cycle R&D accumulation than from rapidly taking the fattest profit pool of the entire industry through the most advanced product generation like a primary supplier to pure cloud vendors. These success factors still exist today, but the capital market has already taken one step ahead and priced it as a company whose profit pool will also migrate in sync. This creates the most important crack in this round of research: the company's real commercial position is a domestic platform leader; the stock's traded position is increasingly close to a high-elasticity AI datacom challenger. The two can coexist for a while, but it is hard to explain them forever with the same valuation framework.

The market's most likely misjudgment now is the difference between catching up and dominating. Being able to make 800G and 1.6T does not mean being able to obtain the same share as the leaders in the most core global AI cloud-vendor profit pool. Having complete platform capability does not mean high-end segment margins will automatically converge toward the best peers. The scarcity of a domestic asset does not mean it is worth owning at any price. The most important variable over the next 1 year is whether later quarters can continue to deliver the dual evidence of "high growth + high-quality cash flow." The most important variable over the next 3 years is whether overseas customers and high-end datacom product share can rise materially faster than access and telecom businesses. The most important variable over the next 5 years is whether the company can truly translate platform advantages into mid-to-high-end margins in global AI optical interconnect, rather than staying at the level of "growing with the industry."

If you ask under what conditions Accelink would become a better investment target, my answer is very simple: let the price return to a range that can absorb uncertainty, and see two hard pieces of evidence land, instead of telling a bigger AI story again: the high-end datacom revenue share continues to rise, and cash-flow and inventory indicators do not deteriorate because of capacity expansion. Conversely, if over the next two quarters, data and access revenue growth slows, inventory stays elevated, and gross margin does not rise, investors should reexamine the current high-valuation framework and even overturn it. At that point, what you own is no longer a "platform stock in the process of delivering growth," but a "manufacturing stock whose high-level narrative failed to deliver."

Bull and Bear Cases

Bull case:

• Accelink has already shifted from "transmission-led" to "data and access-led." In 2025, the latter accounted for 70.94% of revenue and grew 65.89% year on year, so the growth direction is not empty.

• The company's high-speed data-center products already cover 100G to 1.6T and have entered higher-value forms such as LRO and LPO, showing that it is at least on the track of this generation upgrade.

• Operating cash flow repaired from CNY -641 million in 2024 to CNY 1.628 billion in 2025, indicating improving profit quality.

• Under Omdia's measure, the company had a rolling four-quarter global optical components share of 5.9% and ranked fourth globally in datacom optical components, showing it is not a marginal participant.

Bear case:

• The current closing price corresponds to PE (TTM) of 159.59x and static PE of 174.72x, so valuation has already far exceeded delivered profit capability.

• 2025 net profit was only CNY 946 million, while Zhongji Innolight and Eoptolink reached CNY 10.797 billion and CNY 9.532 billion respectively. The earnings gap remains an order of magnitude.

• In 2026Q1, inventory rose to CNY 7.478 billion, while prepayments and equipment payments rose at the same time, showing that this growth round has high working-capital requirements.

• The company's overseas revenue share in 2025 was only 26.69%, while the logic of being a primary supplier to pure AI cloud vendors needs stronger global customer depth as support.

Pre-mortem

The first scenario that could make me lose 50% three years from now is that the 1.6T transition in 2027 does not send Accelink into the overseas high-value profit pool. Zhongji Innolight and Eoptolink continue to win deeper certification and larger shipment volumes at overseas cloud vendors. To defend share, Accelink follows price cuts of 25%-30% in 800G/1.6T. Data and access revenue can still grow, but consolidated gross margin falls from around 23% back to 20%-21%, and net margin retreats to 5%-6%. The market then rerates it from an "AI high-growth challenger" back to a "platform manufacturer," compressing valuation from 170x static PE to 45-60x and sending the share price back to the CNY 70-100 range. This scenario does not require operating failure. It only requires growth quality to fall short of the current imagination.

The second scenario is that in 2026-2027, neither domestic computing-power investment nor overseas order progress has a problem, but the company itself moves faster than demand in capacity expansion and inventory preparation. Inventory continues to pile up, operating cash flow weakens again, and the market starts to realize that "income statement up" and "free cash flow up" are not the same thing. Because the current owner-earnings yield is only about 0.5%, once sector sentiment ebbs, capital will first sell the names with the least margin of safety. Even if the company remains a good company at that time, the investment return may still be cut in half.

Final Research Conclusion

Accelink Technologies is a company worthy of respect. It is one of the few true platform companies in China's optical communications industry that spans transmission, access, and datacom, rather than an AI concept stock assembled around a theme. The 2025 and 2026Q1 statements have already proved that it does stand in a beneficiary position in this industry upswing, and datacom upgrades are indeed driving revenue and profit improvement. The problem is that the share price has already priced "will participate in the AI optical module cycle" as "will significantly narrow the gap with leading players in the high-profit AI optical module pool" ahead of profit. This leap has not been fully proven commercially, yet it has been largely realized in valuation.

What I worry about most is the more common mismatch, not a sudden disappearance of industry momentum: the company continues to grow, but growth is not fast enough to support a TTM PE above 160x; the company continues to expand production, but cash flow and inventory do not improve in sync; the company continues to tell the story of 1.6T, LPO, and high-end optical components, but the market discovers that the real profit elasticity still sits with peers that are purer and deeper. For a company like this, the most important research discipline is to "recognize that a good company also needs a good price," not merely to "recognize it is a good company." If the price later returns to a range that can cover execution error, I would be very willing to revisit it. At the current price, I believe waiting is more important than owning.

【Company Profile Score】

• Fundamental quality: Medium

• Growth: High

• Moat: Medium

• Financial resilience: Medium

• Management credibility: Medium

• Valuation attractiveness: Low

• Risk level: High

• Suitable investor type: Not suitable for ordinary investors

【Investment Rating】

• Rating: Avoid

• One-sentence investment thesis: AI datacom growth is real, but the current price has pulled forward too much of the catch-up story.

• 【Ideal/Fair Buy Price】55-70 CNY Rationale: This corresponds to the bear-case intrinsic value with an additional margin of safety of about 20%.

• Holdable price: 80-108 CNY

• Clearly overvalued price: 135-155 CNY

• Current price classification: Clearly overvalued

• Worth waiting for a better price: Yes. A more reasonable trigger would be the share price entering the area around CNY 70, while later quarters can still maintain high growth in data and access and cash flow does not deteriorate.

• Target holding period: 3-5 years

• Expected annualized return: buying at the current price implies about -67% / -54% / -33% under the bear / base / bull scenarios

• Maximum loss risk: if high-end module share improvement falls short of expectations and valuation returns to a reasonable 45-60x range, the share price could fall 50%-65%

• Signals that trigger reassessment: Data and access revenue YoY below 25% for two consecutive quarters

• Consolidated gross margin below 21% for two consecutive quarters

• Inventory/annualized revenue staying above 60% while operating cash flow/net profit is below 0.8x

• Overseas revenue share still materially below 30% over the next two years

• Management no longer listing high-speed optical modules and high-end optical components as future core growth drivers

【Valuation Range】

• current: 204.97 (as of the 2026-06-12 close)

• bear (conservative · ideal buy zone): [55, 70]

• base (reasonable · acceptable holding zone): [80, 108]

• bull (optimistic · above the clearly overvalued line): [135, 155]

Key Data Table

This table puts the two most important facts in the entire report together. On one side, revenue and profit are indeed stepping up. On the other side, cash flow, inventory, and capital expenditure are also expanding volatility. It explains why Accelink is worth researching, and it also explains why, at the current price, risk control matters more than growth imagination.

Research Uncertainties

• The company has not broken out revenue and gross margin for products such as 800G, 1.6T, LPO, and LRO in current public disclosures. I can only infer from the product catalog, management statements, and total growth rate, and cannot precisely quantify each generation's contribution.

• The company has not disclosed a precise split between maintenance capex and expansion capex, so owner-earnings estimates can only use a conservative method.

• For the market narrative around the "self-sufficiency rate of domestic optical chips," current public materials do not provide a verifiable ratio. Therefore, this report discusses it only directionally and does not provide a precise number.

• After Eoptolink's 2025 profit distribution and bonus share issue, different public data vendors use different bases for market capitalization and TTM PE. Horizontal valuation comparison is therefore more suitable for directional judgment than precise subtraction.

Reference Sources

• Accelink Technologies' 2025 annual report, 2026 first-quarter report, May 2026 investor relations activity record, and abnormal-volatility announcement.

• Accelink Technologies' 2009 IPO prospectus, used to trace the listing path, issuance parameters, and early competitive structure.

• Zhongji Innolight, Eoptolink, and Changxinbochuang's 2025 annual reports or annual-report summaries, as well as Eoptolink's 2026 investor relations materials.

• Coherent and Lumentum's fiscal 2025 annual reports and latest fiscal 2026 quarterly results materials.

• Eastmoney Choice, ChinaBond, and other public market data pages, used for current share price, market capitalization, valuation multiples, and China's 10-year government bond yield.

Other Companies Mentioned in the Report

• 300308.SHE: Zhongji Innolight, directly compared on AI datacom optical module scale, profit elasticity, and high-end customer depth.

• 300502.SHE: Eoptolink, directly compared on 1.6T, LPO, OCS, and the pace of overseas cloud-customer breakthroughs.

• 300548.SHE: Changxinbochuang, compared on segmented components, AEC, and a higher-elasticity small-to-mid-cap optical interconnect path.

• COHR.US: Coherent, as a reference for a global integrated photonics platform and vertical integration capability.

• LITE.US: Lumentum, as a global optical components reference in high-end components, OCS, and CPO.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.