AOSL’s beat is a mix-shift head fake, not a cycle turn

The print says AOSL is not yet in a broad power-discrete recovery, even though the headline numbers were good enough to squeeze a short-term earnings reaction. The Street comparison basis shows EPS of $0.13 versus the $0.10 estimate, a +30.0% surprise, and revenue of $182.5 million versus the $177.7 million estimate, a +2.7% surprise. That is the part that was priced in versus the part that mattered: investors were already set up for September to benefit from seasonal Computing strength and better revenue than the prior quarter, but the market may be overpaying for a beat that came with reported revenue up only +0.3% year-over-year and a gross margin of 23.5%. The variant view is that the multiple should not be anchored on the $0.13 EPS result, because the company’s own December guide points to revenue of approximately $160 million, plus or minus $10 million, and non-GAAP gross margin of 23%, plus or minus 1%. AOSL proved it can harvest September seasonality; it did not prove the next two quarters have enough end-market breadth to lift gross margin out of the low-20s range.

That distinction matters because the September quarter’s revenue trajectory looks better in sequential terms than it does in structural terms. Revenue rose to $182.5 million from $176.5 million, a +3.4% sequential move, but the year-over-year change was only +0.3% against $181.9 million in Q1 FY2025. Gross margin moved from 23.4% to 23.5%, which is not enough evidence that utilization, pricing, or mix has inflected in a durable way, especially after the business printed 24.5% gross margin a year ago. The chart belongs here because the investment debate is not whether AOSL beat a low bar in September; it is whether the revenue and margin series can escape the pattern of short seasonal rebounds followed by renewed pressure.

The charted history should keep PMs focused on the sequence rather than the single-quarter surprise. AOSL recovered from $164.6 million in Q3 FY2025 to $176.5 million in Q4 FY2025 and then to $182.5 million in Q1 FY2026, but the next two datapoints in the table show the risk embedded in management’s seasonal language: $162.3 million in Q2 FY2026 and $163.8 million in Q3 FY2026, with gross margin at 21.5% and then 21.1%. Those later entries are not part of the Q1 beat, but they frame why a September-only earnings acceleration is not enough. The gross margin series has moved from 21.4% in Q3 FY2025 to 23.4% in Q4 FY2025 to 23.5% in Q1 FY2026, then to 21.5% and 21.1% in the subsequent periods shown. A stock that rallies on the +30.0% EPS surprise is implicitly underwriting that 23.5% is closer to a floor than a September high-water mark; the data pack does not support that conclusion.

The business mix explains both the upside and the fragility. CFO Yifan Liang put the company’s own accounts on record: “Revenue for the September quarter was $182.5 million, up 3.4% sequentially and up 0.3% year-over-year.” That quote earns its place because it forces the right base rate: the same quarter that beat consensus by +2.7% barely grew against last year. Beneath that, DMOS revenue was $108.5 million, up 1.1% sequentially and down 11.4% over last year, while Power IC revenue was $72.7 million, up 5.9% from the prior quarter and 37.3% from a year ago. The market may be missing that Power IC is carrying the growth burden while the larger DMOS line remains in year-over-year decline. If the stock is treated as a clean recovery story, the burden of proof sits with DMOS, because $108.5 million of revenue declining 11.4% over last year is a different setup from $72.7 million growing 37.3%.

That mix point gets sharper when translated into segment behavior, because Computing is the source of the September lift and the source of the December air pocket. Stephen Chang said the September quarter Computing revenue was “up 27.1% year-over-year and up 4.6% sequentially and represented the majority or 53.2% of total revenue.” The commitment that matters is not the historical percentage, but the next sentence from the same speaker: “Looking ahead to the December quarter, we expect Computing segment revenue to decline nearly 20% sequentially.” A segment that represented 53.2% of total revenue and is expected to decline nearly 20% sequentially can swamp progress elsewhere, which is exactly why the product revenue guide of around $150 million and total revenue guide of approximately $160 million, plus or minus $10 million, should carry more weight than the September beat. AOSL’s revenue base is not diversified enough in this print to let investors ignore Computing seasonality.

The non-Computing segments did not supply enough offset to change that conclusion. The call excerpts show one segment with September quarter revenue down 25.8% year-over-year and 11.6% sequentially while representing 12.9% of total revenue, and Power Supply and Industrial accounting for 15.3% of total revenue while down 12.4% year-over-year and 5.6% sequentially. Those figures are not consistent with a broad-based demand turn. They are consistent with a company whose seasonal Computing exposure and Power IC mix can produce an EPS beat in September, while other end markets remain a drag. The market’s likely mistake is to collapse those different stories into one “recovery” narrative. The numbers separate them: Power IC up 37.3% from a year ago, DMOS down 11.4% over last year, one segment down 25.8% year-over-year, and Power Supply and Industrial down 12.4% year-over-year.

Margins reinforce that the beat was not driven by a clean operating leverage inflection. On the company’s own non-GAAP basis, gross margin was 24.1% compared to 24.4% last quarter and 25.5% a year ago, while non-GAAP operating expenses were $41.4 million compared to $40.9 million for the prior quarter and $38.5 million last year. Non-GAAP quarterly EPS was $0.13 compared to $0.02 per share last quarter and $0.21 per share a year ago. That is the crux: EPS improved sharply sequentially, but it is still below last year, while gross margin is also below last year and opex is higher. EBITDA excluding equity method investment income was $19.4 million for the quarter compared to $10.5 million last quarter and $20.6 million for the same quarter a year ago, which gives bulls a real sequential improvement but not a year-over-year margin story. If investors pay for the sequential EBITDA recovery, they should demand evidence that December can hold closer to September than to the later 21.5% gross margin in the quarterly history.

Cash flow and balance sheet optics are better, but they also require cleanup before being capitalized at full value. Operating cash flow was $10.2 million, including $5 million of repayment of customer deposits, versus negative $2.8 million in the prior quarter and positive $11 million last year. The company expects to refund $8.2 million of customer deposits in the December quarter, which is a real cash headwind against the next quarter’s guide. Cash ended the September quarter at $223.5 million compared to $153.1 million at the end of last quarter, but that improvement included the JV monetization: the company divested 20.3% of its equity interest in the JV company for $150 million and received the first installment payment of $94 million. It also paid out $20.8 million for the remaining balance of its equipment loan. This is not a negative read on liquidity; it is a warning not to confuse a $94 million installment from divesting 20.3% of the JV company with operating momentum from the core power business.

Capex is another reason the December margin guide should be taken literally. CapEx was $9.8 million in the September quarter compared to $14.3 million for the prior quarter, and the December quarter expectation is $14 million to $16 million. That points to higher cash investment just as revenue steps down to approximately $160 million, plus or minus $10 million, and non-GAAP gross margin is guided to 23%, plus or minus 1%. Non-GAAP operating expenses are expected to be $40.5 million, plus or minus $1 million, which means opex is not being cut enough to fully cushion the revenue reset. The market may want to frame the quarter as EPS leverage returning; the guide says the next test is whether a lower revenue base can hold 23%, plus or minus 1%, non-GAAP gross margin without the September Computing tailwind.

The tone of the call was consistent with that more cautious read, and the tone history shows why the delivery should not be treated as a bullish signal. The Q1 FY2026 call scored sentiment 0.18, guidance_tone 0.29, tone_confidence 0.41, prepared_sentiment 0.02, qa_sentiment 0.15, ai_optimism 0.38, uncertainty 42.9, and qa_evasiveness 44.3. Relative to Q4 FY2025, sentiment stayed at 0.18, guidance_tone slipped from 0.32 to 0.29, qa_sentiment declined from 0.20 to 0.15, and ai_optimism fell from 0.45 to 0.38, while uncertainty dropped from 86.5 to 42.9 and qa_evasiveness dropped from 90.1 to 44.3. That combination says management was clearer, not more upbeat. The cleaner delivery lowers downside surprise risk around the December step-down, but it does not create upside. In other words, the tone corroborates the thesis: this was a beat paired with explicit seasonal caution, not a management team trying to pull numbers higher.

The supply-chain read-through is necessarily narrow because the data pack lists no named customers of AOSL and no suppliers to AOSL, so there is no defensible customer-specific or supplier-specific magnitude to pass through. The disciplined implication is still useful for PMs: any unnamed customer exposed to AOSL’s Computing business faces a supplier calling for a nearly 20% sequential decline in that segment after it represented 53.2% of total revenue, while any unnamed supplier tied to AOSL’s manufacturing and assembly demand sees December product revenue guided around $150 million and CapEx moving to $14 million to $16 million from $9.8 million. For named read-throughs, the absence of named counterparties is itself the constraint; inventing a customer or supplier would be less useful than recognizing that the available data supports only segment-level and capex-level implications.

The peer context also argues against paying a scarcity multiple for the quarter. In the Power_Discrete peer table, VSH reported $839.2 million of revenue, 21.0% gross margin, and +17.3% revenue YoY, while DIOD reported $405.5 million of revenue, 31.8% gross margin, and +22.1% revenue YoY. AOSL’s Q1 FY2026 revenue was $182.5 million, gross margin was 23.5%, and revenue YoY was +0.3%. That puts AOSL’s gross margin above VSH’s 21.0% but far below DIOD’s 31.8%, while its +0.3% revenue YoY trails both VSH at +17.3% and DIOD at +22.1%. The comparative point is not that AOSL is broken; it is that the September beat does not screen as an above-peer growth print. Bulls need to argue Power IC mix can pull the consolidated profile closer to peers with higher growth or higher gross margin, and the current data still shows DMOS at $108.5 million and down 11.4% over last year.

What was priced in, then, was a September quarter that could clear a low Street hurdle after Q4 FY2025 revenue of $176.5 million and prior-quarter non-GAAP EPS of $0.02. What actually surprised was that EPS reached $0.13 against $0.10, revenue reached $182.5 million against $177.7 million, Power IC rose 37.3% from a year ago, and operating cash flow flipped to $10.2 million from negative $2.8 million. What should not be extrapolated is the breadth of that surprise: reported revenue grew only +0.3% year-over-year, non-GAAP gross margin was 24.1% versus 25.5% a year ago, DMOS was down 11.4% over last year, and December Computing is expected to decline nearly 20% sequentially. The variant perception is simple enough to be investable: the beat is real, but the stock should be judged on whether the Power IC mix shift can offset a large Computing reset and a still-declining DMOS base. The data pack says that burden has not yet been met.

For the next quarter, the thesis is confirmed if December revenue lands near the high end of approximately $160 million, plus or minus $10 million, while non-GAAP gross margin holds at or above 23%, plus or minus 1%, and non-GAAP operating expenses stay within $40.5 million, plus or minus $1 million. It is strengthened if product revenue is materially better than around $150 million despite the expected nearly 20% sequential Computing decline, because that would show non-Computing offsets rather than September seasonality. It breaks if revenue trends toward the low end of the $160 million, plus or minus $10 million range, gross margin falls below the 23%, plus or minus 1% framework, or the $8.2 million customer deposit refund combines with CapEx of $14 million to $16 million to pressure cash more than the $223.5 million balance suggests. The date is the December quarter report: by then, investors should know whether Q1 FY2026 was the first quarter of a mix-led recovery or just the cleanest quarter in a still uneven power cycle.