AXT’s miss is less the story than the export-gated backlog now defining the recovery

The Q2 report was bad on the metrics investors were underwriting into the print, but it was not bad in the way that matters for the next trade. What was priced in was a small loss-making substrate company still digesting weak demand after Q1 FY2025 revenue of $19.4 million and negative gross margin, with the Street already at a loss estimate of -$0.13. What surprised was the magnitude of the top-line miss, with revenue of $18.0 million versus $19.5 million expected, a -7.9% surprise, and EPS of -$0.15 versus -$0.13 expected, a -15.4% surprise. That would normally close the debate in a cyclical materials name. Here it should not, because the miss landed alongside two pieces of evidence that shift the debate from demand destruction to shipment timing: management guided Q3 revenue to $19.0 million to $21.0 million, and Morris Young put a hard number on the bottleneck by saying AXT is “making progress against a backlog of more than $10 million in customer orders for gallium arsenide and indium phosphide substrates.”

That backlog sentence is the print’s fulcrum because it reframes the sequential decline. Revenue fell from $19.4 million in Q1 FY2025 to $18.0 million in Q2 FY2025 on the company’s own reported basis, and the Street comparison shows that investors expected more resilience than they got. But backlog greater than $10 million is too large relative to Q2 revenue of $18.0 million to dismiss as routine order noise, especially when management connected it specifically to gallium arsenide and indium phosphide substrates rather than to a broad corporate wish list. The variant perception is that Q2’s miss is being overread as end-market weakness when the data pack points to constrained conversion: cash fell by $3.1 million to $35.1 million, inventory remained high at $80.1 million, and management highlighted an export permit process that has been slower than desired. If permits move, the model has operating leverage; if they do not, the balance sheet carries the cost of stranded inventory.

The financial trajectory supports that interpretation, but only if investors resist extrapolating the Q2 revenue line. Over the history shown, revenue has swung around a narrow small-cap base rather than compounding steadily, while gross margin has been far more sensitive to volume and mix than the top line alone would imply. Q2 FY2025 gross margin was 8.0%, which is a visible repair from negative 6.4% in Q1 FY2025 but still far below 27.4% in Q2 FY2024. That is the margin story in one sentence: the trough is probably past, but the business has not yet re-earned last year’s absorption. The setup into Q3 is therefore asymmetric. The guide of $19.0 million to $21.0 million does not require a return to the Q2 FY2024 revenue base of $27.9 million, but management still expects “gross margin improvement” and a non-GAAP net loss of $0.11 to $0.13.

The margin bridge matters because the cost base is already being pulled down, which makes shipment conversion more valuable than the headline miss suggests. Gary Fischer said non-GAAP operating expense was $7.6 million in Q2, compared with $8.5 million in Q1 and $8.9 million in Q2 FY2024, so the company is not trying to wait out a substrate cycle with a static expense load. GAAP operating expense showed the same direction at $8.2 million versus $9.0 million in Q1, while the non-GAAP operating loss narrowed to $6.1 million from $9.6 million. The objection is that Q2 FY2024 non-GAAP operating loss was only $1.2 million, which keeps this from being a clean margin recovery story. The answer is that the gap to last year is still about volume and gross margin, not expense discipline, and the next quarter’s evidence should show up first in gross margin rather than in revenue alone.

Indium phosphide is the product line that can change that mix, and management gave investors enough specificity to make the Q3 guide testable. Q2 indium phosphide revenue was $3.6 million, primarily from PON and data center applications in China, and Morris Young committed to a near-term acceleration with the wording, “Broadly speaking, we expect to grow our total indium phosphide revenue by 30% or more in Q3, as a result of growth in applications for PON, data center connectivity and various indium phosphide-based sensors.” The quote earns weight because it attaches a product category, a magnitude, and applications to the guide, rather than hiding behind generic sequential growth. It also explains why the market should separate a substrate-cycle miss from an InP mix inflection: AXT believes it is “either #1 or #2 in the world of indium phosphide substrate supply,” with an estimate of “at least 40% of the world indium phosphide supply.” If that claim is even directionally right, Q3 InP conversion is more important to forward gross margin than a one-quarter miss against $19.5 million Street revenue.

The customer and geographic mix make the recovery both more concentrated and more fragile than the headline guide implies. Asia Pacific accounted for 90% of Q2 revenue, Europe 9%, and North America 1%, so any thesis that assumes a broad Western optical recovery is not supported by this print. The top 5 customers generated approximately 30.9% of revenue, and 1 customer was over the 10% level, meaning shipment timing at a small number of accounts can move reported revenue by more than normal distribution math would suggest. For Coherent and Lumentum, the read-through is not that AXT is seeing a broad North American demand snapback, because North America was only 1% of revenue. The more precise read-through is that InP substrate availability and permit timing can affect EML laser and compound wafer supply chains even when the revenue recognition at AXT is still overwhelmingly Asia Pacific. WIN Semiconductors and IQE sit in the same InP and GaAs substrate chain, while Korea Zinc’s refined indium exposure is tied to whether AXT’s $80.1 million of inventory becomes shippable demand or remains a working-capital drag.

That concentration also explains why the balance sheet should be part of the bull case rather than an afterthought. Cash and cash equivalents and investments decreased by $3.1 million to $35.1 million, and net inventory was only down by approximately $300,000 to $80.1 million. Those two figures create the tension in the story: the company has enough liquidity to wait for permits for another quarter, but the inventory number is large enough that investors should demand proof that backlog is turning into shipments. AXT is not short of product narrative; it is short of demonstrated conversion. The print gives bulls a credible reason to underwrite that conversion because management quantified backlog and guided sequential growth, but it also gives bears a clean way to falsify the story if inventory remains elevated and cash continues to fall.

The call delivery was better than the Q2 miss, but not as bullish as management’s product commentary sounded. The tone history shows Q2 FY2025 sentiment at 0.22 versus 0.09 in Q1 FY2025, while guidance_tone was unchanged at 0.32. That combination matters: management sounded less beaten down, but the guide language did not become more expansive. The more notable change was in qa_evasiveness, which moved to -74.0 in Q2 FY2025 from 48.2 in Q1 FY2025, suggesting the Q&A was more direct even as qa_sentiment stayed flat at 0.03. This is consistent with the print’s substance. Management gave investors numbers they can underwrite, including the $19.0 million to $21.0 million Q3 range and the more than $10 million backlog, but it did not claim the export issue had disappeared.

The tone chart also warns against overpaying for confidence before the revenue shows up. Later tone history in the data set shows that guidance_tone can rise sharply, with Q1 FY2026 at 0.57, but uncertainty also sat at 86.5 in that same call. That conflict matters because a higher guidance tone alone is not the same as lower execution risk. For this Q2 event, the cleaner interpretation is that management’s delivery improved where it needed to improve, namely specificity and Q&A directness, while the uncertainty around permits remained real. The stock should not get credit for a full-cycle recovery on tone, but it should get credit for moving the discussion from vague optimism to dated, measurable Q3 targets.

Relative to substrate peers, AXT looks less like a structurally broken margin asset and more like a small, volatile supplier whose reported quarter is dominated by shipment permission and mix. The peer table’s latest AXTI line shows $26.9 million of revenue, 29.6% gross margin, and +39.1% revenue YoY, which demonstrates the model can screen competitively when volume and mix cooperate. That gross margin is below 3445.T at 32.3% and ROG at 32.2%, but above 5802.T at 22.0% and 5713.T at 20.9%. The comparison is not a valuation argument, because the peers differ in scale and currency, but it is a rebuttal to the idea that AXT is permanently trapped at Q2 FY2025’s 8.0% gross margin. The company has recently shown a margin level comparable with the better substrate group; the investment question is whether the current permit backlog is a bridge back to that profile or a recurring friction that deserves a discount.

The clean way to express the thesis after this print is to be constructive on the next-quarter setup but unforgiving on proof. The Q2 miss was real, and the Street was too high for the quarter: revenue missed by -7.9%, EPS missed by -15.4%, and Q2 FY2025 revenue was far below Q2 FY2024’s $27.9 million. But the surprise investors should care about is that management put numbers around the recovery path despite the miss, including Q3 revenue of $19.0 million to $21.0 million, indium phosphide growth of 30% or more, and backlog of more than $10 million. That is enough to make Q2 a potential trough print rather than just another disappointment. The bear case becomes dominant only if the backlog fails to convert while inventory remains around $80.1 million and cash keeps stepping down.

What to watch next is therefore concrete. On the next quarter’s report, the thesis is confirmed if Q3 revenue lands inside or above the $19.0 million to $21.0 million range, total indium phosphide revenue grows by 30% or more, and the guided non-GAAP net loss narrows into the $0.11 to $0.13 range. Gross margin must also show the promised improvement from Q2 FY2025’s 8.0%, because a revenue uptick without absorption would mean permits helped shipments but not profitability. The thesis breaks if management again cites export permits while backlog remains more than $10 million, inventory stays near $80.1 million, or cash falls materially from $35.1 million without a corresponding revenue step. For AXT, the next quarter is not about whether Q2 was weak; it was. It is about whether a quantified backlog and InP guide turn into reported revenue before the balance sheet has to fund another stalled quarter.