NextFin News - Eoptolink Technology, one of the world's largest makers of optical modules for artificial-intelligence data centers, delivered a half-year profit that nearly doubled - and an inventory pile that grew 61% in six months. The Chengdu-based company reported net profit of 7.53 billion yuan ($1.05 billion) for the first half of 2026, up 90.98% from a year earlier, on revenue that more than doubled to 20.91 billion yuan. But stock on hand swelled to 11.66 billion yuan by the end of June, equal to nearly a third of everything the company owns, while money owed by customers climbed to 7.55 billion yuan. The split verdict is the story: the income statement says the AI supercycle is still roaring; the balance sheet asks how much of that roar is being paid for with working capital that has not yet come back as cash.
The report, released August 21, landed in a market that has already rewarded Eoptolink handsomely. Shares rose 6.76% that day, closing at 442 yuan and lifting market capitalization above 610 billion yuan, after a run that left the stock up nearly 80% year to date by mid-2026. Peer Zhongji Innolight, the industry's largest player, rose 4.29% on its own earnings release the same evening, its market capitalization crossing 1.1 trillion yuan. The sector's biggest names have still given back a chunk of their mid-year gains - Eoptolink pulled back as much as 17.59% from its July record of 618.87 yuan before the earnings rebound. The question the inventory build forces investors to answer is not whether AI demand is real. It is whether Eoptolink is stockpiling ahead of deliveries it has already won, or building inventory faster than the market can absorb it.
The Income Statement and the Balance Sheet Are Telling Two Different Stories
On the face of it, the half-year report is a triumph. Revenue of 20.91 billion yuan was up 100.34% year over year. Net profit attributable to shareholders of 7.53 billion yuan rose 90.98%. Earnings per share climbed to 5.41 yuan from 2.84 yuan. The core optical interconnect business, which accounts for essentially all of the company's revenue, grew 100.59% to 20.88 billion yuan and actually improved its gross margin to 48.46%, up almost a full percentage point from a year earlier. Capacity rose 86.6% to 28.36 million units; sales volume rose 61.0% to 11.19 million units.
Turn to the balance sheet and the picture gets more complicated. Inventory at the end of June stood at 11.66 billion yuan, or 31.80% of total assets - up 4.42 billion yuan from the 7.23 billion yuan carried at the start of the year. That is a 61% increase in six months. Accounts receivable reached 7.55 billion yuan. Taken together, inventory and receivables now represent roughly 19.2 billion yuan of the company's asset base, a claim on cash that grows every quarter the buildout continues.
The cash-flow statement explains the tension. Operating cash flow was 1.62 billion yuan in the first half, up 69.67% year over year - a healthy number in isolation, but only about 21% of the 7.53 billion yuan in reported profit. Investing cash flow was a net outflow of 1.29 billion yuan, driven by fixed-asset investment as the company adds capacity. In other words, for every yuan of profit Eoptolink booked, roughly 79 fen stayed on the balance sheet as inventory or receivables rather than arriving as cash. That gap is the price of feeding an AI boom where customers pay later and suppliers want to be paid sooner.
This pattern is not unique to Eoptolink. Zhongji Innolight, the industry's largest player by revenue, reported half-year revenue of 41.78 billion yuan, up 182.49%, and net profit of 13.65 billion yuan, up 241.7%. Its inventory at the end of June was 19.83 billion yuan, or 28.76% of total assets, and its accounts receivable hit 15.00 billion yuan - 21.76% of assets, up from 13.86% at the end of 2025. Zhongji's operating cash flow fell 44.08% to 1.80 billion yuan even as profit tripled. The whole optical-module complex is, in effect, financing the AI buildout with its own working capital.
Why Stockpiling Is Rational - and Why It Is Still a Bet
The immediate explanation for the inventory build is straightforward: Eoptolink is buying components and assembling modules ahead of deliveries it expects to make. In its half-year report, the company put the driver in its own words:
400G, 800G, 1.6T and higher-speed optical interconnect products have become the company's core products and performance growth engines, and are widely used in AI computing clusters and cloud data centers.
When your customers are hyperscale cloud providers building GPU clusters on fixed timelines, a missed delivery window is worth more than the carrying cost of extra inventory.
The deeper mechanism is a supply-chain arms race. The choke points sit upstream: 200G-and-above EML laser chips and high-speed digital signal processor chips remain dominated by overseas suppliers, and lead times for critical inputs have stretched to around 50 weeks. Industry estimates put 2026 shipments of 1.6T modules at roughly 15 million units against demand that implies a shortfall of about 10 million units, with tightness expected to extend into 2027. When the components you need take a year to arrive and your customers will not wait, the only rational response is to order early, pay in advance, and hold the buffer yourself.
Eoptolink's first-quarter report showed exactly this behavior: prepayments jumped 3,920% quarter over quarter to 682 million yuan, almost entirely raw-material prepayments. The June inventory figure is the physical manifestation of that same logic - cash converted into components, then into work in progress, then into finished 800G and 1.6T modules sitting in warehouses awaiting shipment. The company flagged inventory net realizable value as a key audit matter for the second consecutive year, a disclosure that signals the auditors are watching the stockpile closely even as management defends it.
The bet embedded in 11.66 billion yuan of inventory is a bet on two things: that demand holds, and that the technology mix does not shift underneath the stockpile. At a 48.46% gross margin, Eoptolink has pricing power, which means each unit that sells contributes heavily to profit. But optical modules are a fast-iterating business - the industry is already demonstrating 6.4T NPO and 12.8T XPO prototypes, and the eventual migration toward co-packaged optics could change what data centers buy. Inventory that is valuable at 800G margins can become a write-down if the customer queue moves to a generation the company has not yet shipped.
Cyclical or Structural: The Call That Determines Everything
This is the judgment that separates a trade from an investment, and getting it wrong flips the conclusion. Is the inventory build cyclical - a mean-reverting wave of front-loaded buying that will unwind once supply catches up - or structural - a permanent step-up in the working capital required to run an AI-era optics business?
The evidence points to both forces operating at once, and the distinction matters. The structural leg is real and durable. AI data-center architecture has permanently raised the optical content per rack: GPU clusters require far more optical interconnects than traditional enterprise data centers, and the migration from 800G to 1.6T and eventually 3.2T keeps the technology treadmill running. Eoptolink's overseas sales were 96.16% of revenue in 2025, and its first-half 2026 overseas optical interconnect sales of 20.46 billion yuan - 973 million units shipped abroad - show that exposure has not diminished. The company said collections on those overseas sales remain normal. This is a regime change in demand, not a one-quarter blip.
The cyclical leg is the stockpiling itself. Inventory growing 61% in six months is a front-loaded bet, not a steady-state condition. Historical cycles in the optical-module industry show the pattern clearly: companies that built inventory ahead of a genuine demand wave saw margins expand as stock turned quickly; companies that built inventory into a slowing cycle saw write-downs compress earnings for quarters afterward. The difference is never the inventory level in isolation - it is the turnover rate. If Eoptolink's inventory converts to revenue over the next two quarters at current margins, the build validates itself. If turnover slows while gross margin compresses, the stockpile becomes a liability.
There is also a currency dimension that is easy to miss. Eoptolink earns almost all of its revenue in dollars but reports in yuan. The first-half report shows financial expenses swinging from a 208 million yuan net gain a year earlier to a 644 million yuan expense, driven by exchange-rate movements. A strengthening yuan against the dollar does not change the dollar value of the inventory, but it does reduce the yuan value of future revenue - a translation squeeze that can hit margins even when unit volumes are rising. The first quarter already produced 522 million yuan of FX losses. Currency is the silent third variable in the inventory equation.
The Second-Order Question the Market Is Not Asking
The first-order read is simple and already priced in: strong demand, strong profit, buy the stock. The shares are up nearly 80% year to date, and the eight analysts covering the company rate it a buy with an average price target of 645.32 yuan, implying roughly 56% upside from recent levels. That consensus assumes the inventory converts.
The second-order question is about who is financing the AI buildout and what that does to earnings quality. Eoptolink's operating cash flow covers only about a fifth of its reported profit. The rest is working capital lent, in effect, to customers through receivables and to suppliers through prepayments. Meanwhile the company is spending on capacity - and on research and development of 4.40 billion yuan in the first half, up 31.85%, as it chases the next generation of products.
This is the context for the company's planned Hong Kong listing. In June, the board approved an issuance of H shares equal to up to 8% of post-issue capital, with proceeds earmarked for core technology R&D, manufacturing upgrades, capacity expansion, strategic acquisitions, and working capital. A Hong Kong placement executed at a strong valuation would be growth funding - a way to build more capacity without straining the balance sheet. A placement delayed, or executed at a discount after the stock has fallen, would signal that working-capital needs are pressing harder than management admits. The market will read the terms of that raise as a verdict on the inventory build itself.
The cross-asset implication runs further. If Chinese optical-module makers are collectively absorbing tens of billions of yuan of working capital to feed AI capex, then their ability to fund the next capacity wave depends on equity markets staying open and receptive. A sustained de-rating in the sector - the kind that followed the July peak, when Eoptolink pulled back nearly 18% from its record high - raises the cost of that funding and forces a choice between slower growth and more dilution.
The Counter-Thesis - and the Signal That Would Break It
The strongest argument against reading the inventory build as a warning is also the simplest: Eoptolink's customers are the highest-credit-quality counterparties in technology. The company's report describes its customer base as AI infrastructure companies, hyperscale cloud service providers, equipment makers, and distributors. If your debtors include the world's largest cloud platforms, receivables risk is low - and the company stated explicitly that overseas collections are normal. Under this view, the inventory is not a red flag but a competitive moat: only companies with the balance sheet and the supplier relationships to stockpile can capture a supercycle where demand exceeds supply through 2027. Core vendors' orders already cover all of 2026, with some customers scheduled into 2027.
This counter-thesis has institutional backing in the order visibility that management and industry sources describe, and it is reinforced by the peer data: Zhongji Innolight is carrying even more inventory in absolute terms while growing faster, yet its gross margin on optical transceiver modules held at 46.59%, up 6.63 percentage points. If the whole industry is stockpiling and the whole industry is still raising prices, the stockpile is working.
The falsifying signal is specific and observable. Watch Eoptolink's inventory turnover - cost of goods sold divided by average inventory - over the next two quarters. If turnover declines for two consecutive quarters while gross margin compresses by more than three percentage points, the stockpiling thesis breaks. That combination would mean the company is holding more stock that is turning more slowly at lower prices - the classic signature of a cycle top rather than a supply-constrained boom. A second warning sign: operating cash flow turning negative for a full half while reported profit stays positive. Either signal would indicate that the balance sheet is no longer funding growth but absorbing demand that is not there.
What Comes Next: Three Horizons, Three Scenarios
Short term - the next one to two quarters. The first real test is the third-quarter report, expected in late October. If the 11.66 billion yuan of June inventory converts to revenue without a material margin hit, the stockpiling narrative flips from risk to proof of demand, and the shares can re-rate higher. If inventory days lengthen and receivables keep growing faster than revenue, the market will start discounting the earnings quality. The specific metrics: inventory turnover, receivables days, and whether operating cash flow keeps pace with profit.
Medium term - six to twelve months. The Hong Kong listing becomes the key funding lever. A successful placement at a premium valuation funds the 1.6T capacity ramp and validates management's read of the cycle. A delayed or down-round raise would signal working-capital pressure and could force a choice between growth and dilution. Also watch the currency: a sustained yuan appreciation would keep pressuring the translated value of dollar revenues.
Long term - the structural view. The AI data-center buildout is a multi-year regime shift, and Eoptolink's 96% overseas exposure and 48%+ gross margins position it well - provided the technology transitions do not strand today's inventory. The migration from 800G to 1.6T is already underway; the next inflection, toward 3.2T, co-packaged optics, and silicon photonics, is where the current stockpile could become obsolete. The company is investing 4.40 billion yuan in R&D to stay ahead of that curve, but R&D pays off in years, while inventory turns in quarters.
Base case: AI capex holds, inventory turns in line with sales, and Eoptolink remains a high-beta instrument on the AI buildout. Upside case: the 1.6T shortage persists into 2027, pricing power holds, and the Hong Kong listing funds expansion at a premium valuation. Downside case: a hyperscaler capex pause or a faster-than-expected technology shift leaves the 11.66 billion yuan stockpile needing write-downs, and the 48.46% gross margin proves to have been the cycle peak.
Eoptolink's balance sheet is placing a leveraged bet on the AI supercycle, and for now the income statement is winning. The risk is not that demand disappears overnight. It is that in a race where every optics maker is stockpiling, the winner is decided not by who sells the most today, but by who is left holding inventory nobody wants when the music stops.
Explore more exclusive insights at nextfin.ai.
