The market interpreted the collective decline in the optical module chain on October 29 as a weakening of North American AI orders. Zhongji Innolight's and Eoptolink's Q3 reports will be disclosed before October 31; of contract liabilities quarter-on-quarter and inventory turnover, which one can prove that orders have not weakened?

The market interpreted the collective decline in the optical module chain on October 29 as a weakening of North American AI orders. Zhongji Innolight's and Eoptolink's Q3 reports will be disclosed before October 31; of contract liabilities quarter-on-quarter and inventory turnover, which one can prove that orders have not weakened?

The third-quarter report was already posted after market close on October 29 (the announcement is dated October 30), so there is no need to wait until October 31, and no need to speculate further. Putting the two companies under the same ruler, the conclusions are clearly divergent: the contract liabilities line item proves nothing, inventory turnover can prove something, but only for Zhongji Innolight. The curve for Eoptolink that jumped from 144 days to 225 days records the account of the Thailand floods, not the account of North American orders.

Contract liabilities carry no information for either company

At the end of the third quarter, Zhongji Innolight's contract liabilities were 1128.8 wan yuan, versus 2070 wan yuan at the end of the second quarter, down 45% quarter-on-quarter, with 941 wan in cash removed. It sounds like "advance-order halving," but this 1128 wan in total is not even enough for one hour of its revenue—its single-quarter revenue was 277.69 yi.

An item that accounts for only four ten-thousandths of quarterly revenue, with inflows and outflows of a few million yuan, is not in the same world as the 277 yi scale. A more direct comparison: the company's revenue for the first three quarters rose 178% year-on-year, while contract liabilities instead fell 57% from 2632.7 wan at the beginning of the year. Revenue and this line item are not moving in the same direction, which only shows that it has nothing to do with orders.

For the same line item, Accelink accounts for a quarter of single-quarter revenue, TFC for two tenths, Eoptolink for less than 3%, and Innolight for four ten-thousandths—a 600-fold difference across the four. It measures whether customers are willing to pay first, and is determined by customer structure—telecom customers and component customers are accustomed to prepaying, while North American cloud vendors operate on payment terms and do not prepay in the first place. Using it to compare order strength is equivalent to using payment method to compare demand.

Inventory turnover can speak, provided it is compared with operating cost

The transmission order of this chain is fixed: the capex of North American cloud vendors and Nvidia turns into purchase orders for 800G and 1.6T, first landing in optical module makers' prepayments and inventory (asset side, leading by one to two quarters), then landing in single-quarter operating revenue and operating cost (income statement, current quarter), and finally landing in gross margin. To judge whether orders have weakened, one must look at the most upstream segment.

Inventory itself is an ambiguous indicator; an increase may be stockpiling or may be accumulation. The way to distinguish is not to look at absolute days, but to see whether inventory rises faster or operating cost rises faster—cost is the part already sold, inventory is prepared for the part not yet sold; comparing the two growth rates brings out the direction.

Zhongji Innolight's inventory rose from 198.26 yi to 238.78 yi, +20.4% quarter-on-quarter; in the same period, single-quarter operating cost rose from 119.38 yi to 148.26 yi, +24.2% quarter-on-quarter. Cost rose faster than inventory, and turnover days edged down from 135 days to 134 days. Stockpiling is increasing, and shipments are increasing faster; this combination does not look like accumulation.

Eoptolink is the reverse: inventory rose from 116.55 yi to 174.13 yi, +49.4% quarter-on-quarter, while single-quarter operating cost instead fell from 65.42 yi to 58.78 yi, -10.2% quarter-on-quarter, and turnover days surged from 144 days to 225 days. It is the only one among the four where three things happened at the same time: revenue turned negative quarter-on-quarter, inventory had the sharpest quarter-on-quarter increase, and days jumped.

Therefore, the statement "orders in the optical module chain are weakening" can, when落到 the statements, at most fall on one company. TFC's single-quarter revenue was +28.4% quarter-on-quarter, Accelink +32.2%, Innolight +24.6%. If North American capex were weakening, it would not single out just one company.

Eoptolink's 225 days contain the same flood

Both companies have factories in Thailand and hit the same flood disaster in late September.

At the end of September, some shipments caused revenue recognition to cross periods due to external objective circumstances... As of now, the affected shipments have been recognized as revenue in October.

Late September was a concentrated shipment period, highly overlapping with the flood, causing some products to fail to be shipped or exported through customs.

These are Eoptolink's exact words at its October 29 conference call. Zhongji Innolight's call that same evening said the same thing:

Rain in Thailand affected logistics, and some finished product shipments were delayed... After entering the fourth quarter, the relevant impact has been eliminated, and shipments in early October have returned to normal.

Once shipments are delayed, the same thing squeezes this indicator from both ends: goods completed remain in the warehouse, so the numerator goes up; revenue is not recognized and cost is not carried forward, so the denominator goes down. How much of those 225 days is flood, the company itself says it is "difficult to quantify," but the direction is clear. Using it to prove demand weakening is equivalent to counting one thing twice.

There is also corroborating evidence. Eoptolink's asset impairment loss for the first three quarters was 9936 wan yuan, versus 2.03 yi in the same period last year; the financial report gives the reason for the change as "inventory impairment decreased," while inventory in the same period rose 140.7%. A company truly piling up goods would not put it this way.

Harder than these two line items is prepayments

Zhongji Innolight's prepayments rose from 1.34 yi at the beginning of the year to 15.67 yi at the end of the third quarter, +64% quarter-on-quarter versus the second quarter; Eoptolink's rose from 0.17 yi to 13.22 yi, +52% quarter-on-quarter. This is money paid upstream in advance to lock DSPs and lock lasers; once paid, it cannot be recovered, and it is far more credible than a contract liability balance in the tens of millions.

Major customer orders have basically covered the full year of 2027, and no customer order cuts or downward revisions to 2027 demand guidance have been seen.

At the same meeting, the company was also pressed on whether that "covering the full year" for 2027 was an order or guidance, and the answer was "a certainty PO given by the customer, strictly different from demand guidance." Even the company itself has to split PO and guidance apart and explain them separately; using the contract liability balance to replace PO erases that distinction once again.

This was not said only after the stock price had already fallen. On October 23, the very day of the previous round of decline, both companies issued voluntary announcements: Zhongji Innolight wrote "orders on hand are full" and "major customer orders have been placed through 2027," while Eoptolink wrote "orders on hand are sufficient" and "1.6T optical modules have entered a stage of sustained volume ramp."

As for those four bearish candles on October 29—that day they fell 8.5%, 11.1%, 19.8%, and 9.1% respectively. The day before, these same four rose 10.3%, 8.1%, 6.8%, and 3.0% respectively; and before that, on October 22, the four fell 8.8%, 12.3%, 8.4%, and 6.0%. Within five trading days, two rounds of 8% to 12% selloffs sandwiched a round of 8% to 10% rebounds. If orders really were weakening, what would show up is a downward staircase, not this kind of back-and-forth chopping.

Under what circumstances would I be wrong

If Eoptolink's fourth-quarter single-quarter revenue does not return above 125 yi (the second quarter was 125.72 yi), it means that batch of goods at the end of September was not a cross-period shipment but simply never shipped out; the flood explanation is void, and the 225 days must be recalculated as backlog. Conversely, if Zhongji Innolight's fourth-quarter inventory is still up more than 20% quarter-on-quarter while single-quarter operating cost growth drops below 10%, the 134 days will turn upward, and "stockpiling materials" must also be changed to "stuffing the channel."

Source: Zhongji Innolight Co., Ltd. "2026 Third Quarter Report," October 30, 2026, announcement no. 2026-101; "Investor Relations Activity Record Form," October 29, 2026, no. 2026-009; "Announcement on the Company's Recent Production and Operating Conditions," October 23, 2026, announcement no. 2026-100; Chengdu Eoptolink Technology Inc., Ltd. "2026 Third Quarter Report," October 30, 2026, announcement no. 2026-065; "Investor Relations Activity Record Form," October 29, 2026, no. 2026-006; "Announcement on the Company's Recent Production and Operating Conditions," October 23, 2026, announcement no. 2026-061; Suzhou TFC Optical Communication Co., Ltd. "2026 Third Quarter Report," October 29, 2026, announcement no. 2026-064; Wuhan Accelink Technologies Co., Ltd. "2026 Third Quarter Report," October 26, 2026, announcement no. (2026) 061.