AAOI’s miss was CATV timing, but the investable argument is whether datacenter can absorb a $15 million cable air pocket

The Q3 print should be read as a mix-shift test, not a simple demand miss. What was priced in was $131.6 million of revenue and a loss of -$0.10, which framed the quarter around whether Applied Optoelectronics could extend the Q2 FY2025 revenue base of $103.0 million into a cleaner optical growth quarter. What actually surprised was different: revenue of $118.6 million missed by -9.8%, while EPS of -$0.09 beat by +10.0%. That split matters because the income statement did not behave like a company losing control of costs, but the revenue mix did not deliver the datacenter acceleration investors were positioned for. The thesis from this print is therefore uncomfortable but actionable: AAOI is not broken on demand, as Q3 revenue still rose +15.2% sequentially and +82.1% year-over-year, but the equity should not be paid simply for aggregate growth until the company proves the datacenter ramp can replace CATV volatility at a higher margin structure.

That distinction is visible in the historical arc. Revenue has moved from $65.2 million in Q3 FY2024 to $118.6 million in Q3 FY2025, but gross margin only moved from 24.4% to 28.0% on the reported quarterly history, and diluted EPS worsened from -$0.42 in Q3 FY2024 to -$0.28 in Q3 FY2025 rather than crossing into profitability. The quarter was also below the next-step cadence implied by the Street’s $131.6 million estimate, even though management characterized its own result as within the company’s prior range. Stefan Murry put the company-accounting frame plainly: “In Q3, we delivered revenue of $118.6 million, which was in line with our guidance range of $115 million to $127 million.” The wording matters because it explains the debate after the print: management delivered against its own range, but investors were underwriting something closer to the high end plus incremental momentum, and the -9.8% revenue surprise is the part the stock has to digest.

The financial trajectory is still improving enough to keep the debate alive. Quarterly revenue has climbed from $99.9 million in Q1 FY2025 to $103.0 million in Q2 FY2025 and $118.6 million in Q3 FY2025, with the data pack showing a further $134.3 million in Q4 FY2025 and $151.1 million in Q1 FY2026. Gross margin, however, has not followed in a straight line: 30.6% in Q1 FY2025, 30.3% in Q2 FY2025, 28.0% in Q3 FY2025, 31.2% in Q4 FY2025 and 29.1% in Q1 FY2026. The market may be missing that AAOI’s revenue path can be directionally correct while the margin path remains uneven, because the company is simultaneously scaling transceiver capacity, absorbing tariffs, and leaning on a CATV product cycle that management says will moderate in Q4. A revenue-only bull case is not enough here; the investable case requires Q4 revenue of $125 million to $140 million to arrive with non-GAAP gross margin discipline and without opex expanding faster than the incremental datacenter dollars.

The reason the Q3 revenue miss did not translate into a larger EPS miss is that the company protected non-GAAP loss despite the top-line shortfall. On the company’s own basis, non-GAAP gross margin was 31%, in line with the guidance range of 29.5% to 31%, and above 25% in Q3 2024 and 30.4% in Q2 2025. Non-GAAP operating expenses were $47.1 million, or 40% of revenue, versus $27.9 million, or 43% of revenue, in Q3 of the prior year. That is the constructive part of the print: opex as a percentage of revenue compressed even as absolute spending rose by $19.2 million, using the two company-disclosed opex figures. But the next quarter will not be forgiving, because management guided non-GAAP operating expenses to $48 million to $50 million per quarter. If revenue lands near the low end of $125 million while non-GAAP operating expenses land near $50 million, the EPS bridge has much less room for mix disappointment than Q3 had.

The mix problem is that Q3’s headline growth was carried by cable rather than the optical datacenter products the market wants to capitalize. CATV segment revenue was a record $70.6 million, more than triple year-over-year and up 26% sequentially, while datacenter revenue was $43.9 million, up 7% year-over-year but down 2% sequentially. Telecom added only $3.7 million, though it was up 34% year-over-year and 93% sequentially. This is the core variant perception: the revenue miss was not a signal that AAOI lacks demand in total, but it was a warning that the wrong revenue stream did too much of the work in Q3. In datacenter, 83% of revenue came from 100G products, 9% came from 200G and 400G transceiver products, and 7% came from 10G and 40G transceiver products. If investors are paying for a transition into 800G or 1.6 terabit, a Q3 datacenter base still dominated by 100G at 83% is not yet the same as proof of that transition.

The Q4 guide is where management attempted to answer that concern, and it is the most important part of the call. Murry said, “We expect Q4 revenue to be between $125 million and $140 million, accounting for a sequential decrease in CATV revenue as well as a more substantial sequential increase in our datacenter revenue.” The phrase is valuable because it commits the company to the specific mix inversion investors need: CATV down, datacenter up more. Management separately said CATV should moderate to $50 million to $55 million next quarter after Q3’s $70.6 million, and later framed cable TV as an implied decline of roughly $15 million while overall revenue would be up roughly $15 million within the specified ranges. That means the Q4 setup is not merely “growth continues”; it is a test of whether datacenter can swing enough sequentially to offset a cable decline and still lift total revenue from $118.6 million to a range whose midpoint the data pack does not require us to calculate. Chih-Hsiang Lin’s summary was even more direct: “So revenue will increase by $25 million to $40 million in Q4.” That statement is aggressive relative to the Q3 revenue miss and is the clearest line investors can hold management to.

The capacity discussion explains why management is willing to put that mix shift in guidance, but it also explains why the market may discount the promise until shipments show up. Murry said the expansion should culminate later this year with “the largest domestic production capacity for 800G or 1.6 terabit transceivers, approximately 35,000 transceivers per month or roughly 35% of our overall capacity for these advanced optical transceivers.” The commitment is significant because it ties the datacenter recovery to a concrete monthly capacity number, not a loose product roadmap. But it also increases execution risk: the company has already made $124.9 million in capital investments this year and is tracking at or above its $120 million to $150 million total CapEx projection, while the completed capital raise brought in $147 million net of commissions and fees and the revolving loan facility with BOK Financial is $35 million. The balance sheet resources exist, but the earnings power has to catch up. Q3 GAAP net loss was $17.9 million, or -$0.28 per basic share, and non-GAAP net loss was $5.4 million, or -$0.09 per share, so the capacity build is still being funded ahead of visible profitability.

That funding bridge has a cost and a macro sensitivity that should not be ignored. Tariffs already created a direct impact on capital equipment of $1.9 million in Q3, or roughly 4%, and management warned that tariff rates and equipment import mix may cause future results to vary materially. This is not a reason to reject the thesis, but it sets the burden of proof: the 800G or 1.6 terabit ramp needs to generate enough datacenter revenue to absorb both the $48 million to $50 million quarterly non-GAAP operating expense run-rate and tariff variability on a capex program that is already at $124.9 million year-to-date. The EPS guide shows that management is not promising a clean inflection yet, with Q4 non-GAAP net income expected between a loss of $9 million and a loss of $2.8 million and non-GAAP EPS between a loss of $0.13 per share and a loss of $0.04 per share using a weighted average basic share count of approximately 70.3 million shares. In other words, Q4 can be a revenue validation quarter without being an earnings validation quarter.

The read-through to the supply chain is unusually limited because the data pack identifies no named customers and no named suppliers for AAOI, which itself matters for how PMs should size the signal. Without customer names, the $43.9 million datacenter revenue figure cannot be allocated to any hyperscale buyer, and without supplier names, the $124.9 million of year-to-date capital investments and $1.9 million tariff impact cannot be mapped to a named equipment vendor. The only customer-side implication that can be defended from the pack is categorical: buyers of AAOI datacenter optics did not pull enough product in Q3 to prevent datacenter revenue from being down 2% sequentially, while CATV customers pulled enough to drive a record $70.6 million segment. The supplier-side implication is also categorical rather than issuer-specific: production equipment suppliers tied to the Texas expansion have an AAOI-funded demand signal through the $147 million net capital raise and $120 million to $150 million total CapEx plan, but the pack provides no named counterparty to monetize that signal.

The peer comparison reinforces why the market is unlikely to give AAOI full credit until datacenter mix and margin catch up. In the latest reported quarter table, AAOI shows $151.1 million of revenue, 29.1% gross margin and +51.4% revenue YoY. That growth is far above COHR at +20.5% revenue YoY and below LITE at +90.1% revenue YoY, but AAOI’s 29.1% gross margin is lower than COHR’s 37.7% and LITE’s 42.4%. LASR is closer in scale at $80.2 million of revenue and shows +55.2% revenue YoY with 33.1% gross margin. The comparative point is not that AAOI lacks growth; it is that AAOI’s growth is being valued against peers that either have much larger revenue bases, higher gross margins, or both. The market’s skepticism after a -9.8% revenue surprise is therefore rational, but it may be misdirected if investors treat the Q3 miss as an end-demand failure rather than a timing and mix handoff into Q4.

The call tone fits that interpretation: management sounded less upbeat in Q3 even while the numbers embedded a more ambitious Q4 mix shift. In the tone history, Q3 FY2025 sentiment fell to 0.25 from 0.40 in Q2 FY2025, guidance_tone fell to 0.26 from 0.33, and ai_optimism collapsed to 0.03 from 0.65, while uncertainty rose to 81.2 from 76.3 and qa_evasiveness rose to 83.5 from 78.6. That is not the transcript profile of a management team simply celebrating an +82.1% year-over-year revenue quarter. It is the profile of a call where delivery was more guarded because the forward story required explaining why datacenter declined 2% sequentially just before management guided to a more substantial sequential increase in datacenter revenue. The later tone entries show the burden of proof easing only when delivery catches up: Q4 FY2025 guidance_tone rose to 0.42 and Q1 FY2026 guidance_tone rose to 0.45, while uncertainty moved to 76.5 and then 76.3.

That tone sequence matters because it gives investors a behavioral tell, not just a financial one. In Q3 FY2025, the management message had to bridge a Street revenue miss, a CATV-heavy mix, and a datacenter product base still 83% 100G. By Q1 FY2026 in the tone table, sentiment was 0.34, qa_sentiment was 0.36, ai_optimism was 0.68, uncertainty was 76.3, and qa_evasiveness was 76.8, with call-over-call deltas versus Q4 FY2025 of sentiment +0.00, guidance_tone +0.03, tone_confidence -0.05, prepared_sentiment -0.01, qa_sentiment +0.01, ai_optimism +0.07, uncertainty -0.1 and qa_evasiveness -0.1. The contradiction is useful: management’s Q3 words were cautious, but the forward tone history becomes less uncertain only after the company has the opportunity to demonstrate that the Q4 and Q1 revenue trajectory is real. PMs should treat that as a confirmation framework rather than a sentiment overlay.

The conclusion is that AAOI’s Q3 is a buyable controversy only for investors willing to underwrite a very specific Q4 mix thesis. The market expected $131.6 million of revenue and got $118.6 million, so the -9.8% miss deserves a discount. But the EPS result of -$0.09 versus -$0.10, the non-GAAP gross margin of 31%, and the Q4 revenue guide of $125 million to $140 million argue against the bear case that demand broadly rolled over. What the market may be mispricing is the shape of the recovery: Q3’s CATV record of $70.6 million is supposed to fall to $50 million to $55 million, while datacenter revenue of $43.9 million is supposed to rise enough to push total company revenue higher. If that happens, the stock’s debate shifts from cable cyclicality to the value of 35,000 transceivers per month of domestic 800G or 1.6 terabit capacity. If it does not, the Q3 miss was the first sign that capex is running ahead of sell-through.

What to watch next is therefore narrow and numerical. For Q4, revenue must land inside or above the $125 million to $140 million guide, and a result near the low end is not enough unless CATV is within the $50 million to $55 million moderation band and datacenter shows the “more substantial sequential increase” management promised from the Q3 base of $43.9 million. Non-GAAP operating expenses need to stay within the $48 million to $50 million quarterly range, and non-GAAP EPS needs to remain within the guided loss of $0.13 per share to $0.04 per share on approximately 70.3 million shares. Gross margin confirmation is the harder tell: reported history shows Q4 FY2025 gross margin at 31.2% and Q1 FY2026 at 29.1%, so any datacenter acceleration that fails to hold margin around those levels would weaken the thesis that advanced transceivers are improving the mix. By the next call after the Q3 FY2025 event dated 2025-11-07, the confirmation is simple: datacenter must replace the roughly $15 million CATV decline, 800G or 1.6 terabit capacity must begin showing up in revenue rather than only CapEx, and tariff impact must not expand materially from the Q3 capital-equipment hit of $1.9 million or roughly 4%.