AEHR
Aehr Test Systems
Summary
What they do:
Design and manufacture the only commercially scaled wafer-level and package-level burn-in test systems for semiconductors — the machines that stress-test AI processors, silicon photonics devices, and power semiconductors before they ship, catching defects that would otherwise cause failures in production data centers.
Why they matter:
Burn-in is a hard gate. No AI chip ships without passing burn-in testing. Aehr is the only pure-play provider of this capability, and their FOX-P system is now being adopted by hyperscalers for custom AI ASIC production testing. A record $41M order from their lead hyperscale customer and a 3.5x book-to-bill ratio signal that the market has recognized burn-in as a structural bottleneck.
Recent performance:
Fiscal Q3 2026 (ending Feb 2026) revenue $10.3M, down 44% YoY due to shipment timing, but bookings hit $37.2M (3.5x book-to-bill) and record effective backlog of $50.9M. Stock ~$83, market cap ~$2.6B. Received record $41M follow-on order from lead hyperscale AI customer in April 2026.
Our Verdict
The only pure-play burn-in test company riding a 3.5x book-to-bill inflection from AI chip and silicon photonics demand — bookings are exploding but revenue has not caught up, creating a backlog-to-revenue conversion setup where FY2027 shows the real earnings power.
Structural trends
Structural
80
/ 100
Moat
7/10
Only commercial-scale burn-in provider, recipe lock-in, installed base consumables stream
AI Exp.AI Exposure
High~45% AI
Play Type
EmergingAI Growth
~50%
Rel. Value
17
PREMIUMBefore an AI processor ships from a packaging house to a hyperscaler's data center, it must be stress-tested. The chip is heated to operating temperature (often 85–125°C), driven at full electrical load, and run through millions of signal cycles. Chips that would fail in the field — infant mortality defects — are identified and rejected. This process is called burn-in, and for AI chips costing thousands of dollars each, it's non-negotiable.
Aehr Test Systems builds the machines that perform this burn-in testing. The product line splits into two families. The FOX-XP is the flagship wafer-level system — each unit tests nine 300mm wafers in parallel, delivers thousands of amperes per wafer (the highest power per wafer in the market), and can be paired with a FOX WaferPak auto-aligner for fully automated, hands-free production. The FOX-NP is a smaller-footprint wafer-level system used for engineering qualification, new product introduction, and lower-volume production. On the package-level side, the Sonoma platform handles burn-in of packaged parts and advanced multi-die modules (e.g., AI ASICs with HBM stacks in CoWoS-type packages). Consumables — WaferPak full-wafer contactors for FOX systems, and burn-in boards/modules (BIMs/BIPs) for Sonoma — generate recurring revenue from the installed base. Management expects consumables to consistently reach 30%+ of total revenue over time.
The company is tiny by semiconductor equipment standards — fiscal 2026 revenue expected around $45–50M — but sits at a critical chokepoint. Their lead hyperscale customer (widely believed to be Google or a Google partner for TPU-class devices) placed a record $41M follow-on production order in April 2026 for package-level burn-in of custom AI ASICs. On the call, CEO Gayn Erickson disclosed that this customer is already ramping Sonoma on a first-generation AI ASIC, has awarded Aehr production for a second-generation higher-power ASIC (production later in CY2026), and is in road-map discussions for a third-generation device — which the customer has asked about doing at wafer level on FOX systems. A major silicon photonics win with a new customer (advanced transceivers for hyperscale data center optical interconnects) extends the platform into 800G/1.6T optical networking — this customer ordered multiple FOX-XP and FOX-NP systems for both engineering qualification and high-volume production, all scheduled to ship in fiscal Q4 (by May 2026), with a forecast for additional XP production systems over the next year. Meanwhile, the lead silicon photonics customer placed a follow-on order and has integrated Aehr systems with autonomous-guided robots for fully lights-out operation. Power semiconductors provide a fourth vector: a new SiC customer in Taiwan (targeting the greater China EV market) ordered a FOX-XP this quarter, and Aehr reports an uptick in SiC activity as Japanese and German OEMs prepare new EV launches. GaN wafer-level burn-in challenges on silicon substrates have been solved, per management.
Aehr's fiscal year ends in late May (changing to late June starting FY2027). For FY2026, management targets the high end of $45–50M in revenue and expects bookings near the high end of $60–80M for H2. Cash stood at $37.1M at quarter-end after fully utilizing a $40M ATM equity program (1.13M shares at avg. $35.38). To support the ramp, a contract manufacturer is being brought online with capacity for 20+ Sonoma systems per month, with first customer shipments expected by end of fiscal Q4 — additive to Fremont's own ~20/month Sonoma capacity plus all FOX product builds. The real inflection is fiscal 2027, where overlapping production ramps from current and next-generation AI processors at the lead customer should drive a significant revenue step-up.
Supply Chain Dependencies
The Catch
AEHR is a $50M revenue company with a $2.6B market cap — 55x trailing revenue. The entire thesis depends on a single lead hyperscale customer that represents the vast majority of bookings. If that customer delays a chip program, brings burn-in in-house, or shifts to a different test methodology, AEHR's order book collapses. Revenue is extremely lumpy — Q3 showed $10M revenue on $37M bookings — and the gap between bookings and recognized revenue creates quarters that look terrible on paper. The stock has priced in perfection; any execution stumble will be punished severely.
If They Win
If the AI chip reliability thesis plays out — that every hyperscaler custom ASIC, every NVIDIA GPU, and every silicon photonics transceiver requires burn-in testing before deployment — Aehr becomes the de facto reliability infrastructure company for the AI era. FY2027 revenue reaches $150M+, FY2028 pushes toward $250M as multiple hyperscalers adopt the FOX platform. The installed base creates a recurring consumables stream (WaferPaks, ContactPaks) worth 20–30% of system revenue annually. At that scale, AEHR is a high-margin, high-growth monopoly in a structural market — the ASML of semiconductor testing, owning a step that cannot be skipped.
Others in Package the Chip
Not financial advice. All scores generated via AI algorithms using public data.