Aehr's Record $33 Million Quarter Lands Into a 65% Drawdown
The tape voted before the earnings did
On November 2, Aehr Test Systems closed 64.7% below its August 17 peak of $145.61, a 49% four-week collapse that started mid-August and never fully reversed; the ninety-day return is -43.9%. Then, after the close, the company reported the best quarter in its history: revenue of $33 million, up 200% year over year and 75% sequentially, non-GAAP net income of $9.4 million, non-GAAP gross margin of 48.5% against 37.5% a year ago, and a raised fiscal 2027 revenue outlook of $150 million.
The divergence is not a company story so much as a sector one. The global tape's 90-day decliner list is dominated by the exact complex Aehr sells into — AI data center shows 52 negative tickers against 11 positive, high bandwidth memory is down nearly 30% across fifteen names, and advanced packaging and leading-edge logic sit in the same bucket. Aehr did not disappoint; it got de-rated with its customers. What makes this report worth reading is what management said while the multiple was being compressed.
From proving it to shipping it
The keyword fingerprint of this call has rotated decisively. Where the top of Aehr's own prior-quarter lexicon was dominated by evaluation language — benchmark customer, "second device," "potential customer," "benchmark testing" — the current quarter leads with volume production, customer program and production ramp. That is the tell. Bookings were $46.7 million for a book-to-bill of 1.4, quarter-end backlog reached $100.9 million despite shipping $33 million, and effective backlog hit $109.3 million as of November 1.
The diversification claim is new in tone, if not in fact. Gayn Erickson noted that historically Aehr "had only a handful of customers with the potential to generate significant production orders," and that “we now have a much broader range of customer and production opportunities contributing to our growth outlook” — Gayn Erickson, Chief Executive Officer · 2026-11-02. He put the count at more than a dozen across AI processors, silicon photonics, silicon carbide and gallium nitride. A year ago this was a single-customer AI story with a serial promise attached.
The accelerator markets that we have been seeing, accelerator markets in 2027 alone are somewhere around 25 million units or so total... The numbers are staggering. And sobering for us.
That arithmetic — five to ten of Aehr's nine-wafer FOX XP systems per million units shipped, roughly $35–70 million of revenue per million units including consumables — has been the bull case for three calls. In January 2026 Erickson was sketching the same math, if smaller: “a single processor for some of these big guys at wafer-level burn-in is 20, 30 systems or so. And these are $4 million, $5 million machines.” — Gayn Erickson, President and Chief Executive Officer · 2026-01-08 The difference today is that the denominator is being filled by real purchase orders rather than a paper benchmark.
The benchmark that refuses to close
To be fair to the skeptics, the marquee wafer-level opportunity is still the fourth or fifth installment of the same serial. The major AI accelerator/CPU/network processor supplier that completed a benchmark last quarter has now bought a full XP wafer-level cell for evaluation at a Taiwan OSAT, with installation planned for early calendar 2027 and a volume decision "within several months following installation and validation." In April the company was describing a progression toward wafer-level for that customer's third device; in January an analyst pressed on why the benchmark had slipped past a February target and Erickson conceded the customer had handed over package-level test vectors that didn't translate cleanly.
So the escalation is real — a paid production cell at an OSAT is a materially better signal than a lab demo — but the conversion event remains ahead of us, not behind. The genuinely new technical claim is thermal: independent per-die thermal control (DITC) inside the WaferPak, which lets Aehr hold every die at the same burn-in condition. That is the answer to thermal runaway in multi-die packages, and it connects to the infant mortality framing that has climbed to near the top of Aehr's own keyword stack. Erickson's argument is that at package level you must cool to the hottest die, leaving the coldest die effectively unscreened — a customer paying for burn-in and not receiving it on the full package.
What management stopped talking about
The more interesting silence is memory. For several quarters HBM and high-bandwidth flash consumed disproportionate airtime, generating keywords like "memory tester" and "high bandwidth flash." On this call memory is a single paragraph: “NAND flash opportunities are progressing more slowly than expected because a limited incremental manufacturing capacity” — Gayn Erickson, Chief Executive Officer · 2026-11-02, with the only live hook being the base die of next-generation HBM — a logic die that may need wafer-level screening before twelve DRAM die are stacked on top of it. Back in July the CEO had already told investors exactly how to treat this: “I currently do not anticipate any memory revenue in even the 150 number.” — Gayn Erickson · 2026-07-14 Read together, that is a theme being quietly demoted from growth driver to option value. Silicon carbide has made the same journey — from roughly 90% of revenue a couple of years ago to about 10% of a $50 million base in fiscal 2026 — though it is now re-accelerating, with the silicon carbide customer base running at full utilization and buying contactors again.
The conservative arithmetic
Strip out the narrative and the guide deserves scrutiny. The prior range was $130–150 million; the new number is $150 million. That is not a raise so much as a pointer to the top of the old band. With $109.3 million of effective backlog, reaching $150 million across the remaining three quarters requires only about $7–8 million of incremental orders — a point Lake Street's Maxwell Michaelis put directly to management. The answer was unusually candid:
We are trying to be conservative... that is currently not a supply limited, meaning we can ship more than that. That is current customers orders and when they say they would like it shipped... we just thought 150 is a pretty nice clean number for us to be conservative at this point.
The shape of the year is also back-loaded in a new way. Where fiscal 2026 was a weak first half, this year starts with a record and the CEO expects each subsequent quarter to step up, with Q4 the largest. Gross margin guidance for the full year is the ~48% range with Q2 lower on Sonoma mix.
What the numbers confirm is operating leverage that hadn't existed before. Total revenue reached $19 million in the June quarter, up 34% year over year, before the $33 million print landed. Gross margin of 42.6% in that quarter came on the back of a 15-point recovery, and R&D intensity stayed high — R&D versus revenue — as the company funded AI wafer-level validation and the memory extension it now talks about less. Effective net cash of $32 million at the last filing has since become $105.1 million of cash and equivalents with zero debt, after absorbing a $3.1 million operating cash outflow to buy long-lead components.
None of that fixes the valuation, which is the real reason the stock is down 65%. At 62x trailing revenue, the shares discount a lot of the twenty-five-million-unit arithmetic. The bull case is that consumables — 41% of Q1 revenue versus 24% a year ago, guided to ~35% for the year against a ~30% historical norm — turn the model into something recurring; Erickson even floated 50% in a strong year. The bear case is that lumpiness cuts both ways, that the Taiwan benchmark decision slides again, and that the FOX wafer-level franchise remains, for now, one and a half customers deep.
The honest summary: this was a genuinely excellent quarter delivered by a company whose language has finally caught up to its orders, dropped into a tape that has already decided AI data-center capital equipment is over-owned. That gap — record results, collapsed multiple — is the trade.