Aehr Test Systems (AEHR) | The Buildout — AI Infrastructure
The Verdict
Aehr Test Systems makes semiconductor test, burn-in, and stabilization systems and consumables used on wafers, singulated die, and packaged parts. Its wafer-level burn-in tools screen AI processors before they enter advanced packaging, while its package-level burn-in platforms handle high-power devices after assembly. The AI infrastructure buildout depends on reliability screening to catch early-life failures before scarce packaging capacity and expensive modules are committed, and Aehr has become embedded in production screening flows at key AI customers.
| Market Cap | — |
| Revenue (TTM) | $51M |
| Revenue Growth | −13.6% |
| EBITDA Margin (TTM) | -22.2% |
| Net Cash | $106M |
| Earnings Beats | 5 of 7 |
| P/E (TTM) | — |
| EV/EBITDA (TTM) | — |
What We Like
- Lead AI processor customer moved all production burn-in screening to Aehr wafer-level burn-in and no longer performs system-level screening.
- AI accelerators, CPUs, and network processors were ~71% of FY26 revenue; optical device test/burn-in was ~20%.
- FY27 guidance initiated at $130M–$150M revenue with non-GAAP pretax income of 18%–22%.
- Year-end backlog was $80.6M, up from $15.2M; effective backlog after year-end was ~$100.6M.
- Balance sheet ended FY26 with $116.5M cash and restricted cash, total debt of $9.9M, and only $2.1M capex.
What We’re Watching
- Customer concentration: Q4 FY26 had three customers >10% of revenue; FY27 guidance is mostly built on current customers.
- Fiscal Q2 FY27 is heavily loaded with Sonoma deliveries; a slip could compress the full-year timeline.
- Power-supply vendors used in Sonoma raised prices 40%; the 10-K lists sole/limited-source components.
- Labor is the soft constraint: management says it does not have the people to ship 10-fold; Fremont was still on one shift.
The demand thesis is strengthening: record bookings, a completed top-tier AI benchmark, and a step-change FY27 guide all point the same direction. The operational thesis remains the open part—execution, hiring, and supply-chain inflation determine whether the revenue ramp lands on time. The key open question is whether Aehr can deliver a 2.6–3x revenue year from a concentrated order book without a delivery or labor stumble.
Earnings Beat
Aehr Test Systems reported fiscal Q4 FY2026 (ended May 29, 2026) revenue of $19.8 million, reported gross margin of 40.6%, and reported net income of $1.4 million; non-GAAP gross margin was 45% and non-GAAP net income was $3.6 million. Bookings were $60.7 million, up more than 500% from $11.1 million in the prior-year quarter.
| Metric | Q4 FY2026 | Q3 FY2026 | Q4 FY2025 | YoY |
|---|---|---|---|---|
| Revenue | $20M | $10M | $14M | +40.4% |
| Gross margin | 40.6% | 32.7% | 30.3% | +1030bps |
| EBITDA | −$1M | −$3M | −$1M | −38.5% |
| EPS | $0.05 | $-0.10 | $-0.10 | −146.9% |
| Bookings | $60.7M | $37.2M | $11.1M | >500% YoY |
| AI + silicon photonics share of revenue | >80% | n/a | 56% | vs 56% in Q4 FY25 |
We are expecting revenue of between $130 million and $150 million representing 2.6x to 3x the just-completed fiscal 26 revenue.— Gayn Erickson, CEO, July 14, 2026
Management tone: Management moved from caution to confidence. In earlier quarters it was cautious on silicon carbide and the top-tier AI benchmark; on the Q4 FY26 call, management said demand from AI-related applications continues to accelerate and described silicon carbide as showing encouraging signs of recovery.
Management Guidance
On the July 14, 2026 call, management initiated FY27 guidance of $130 million to $150 million in revenue, representing 160% to 200% growth from FY26, and non-GAAP pretax net income of 18% to 22% of revenue. The base case assumes no memory revenue, and management said the bulk or majority of the range comes from current customers; the newly completed top-tier AI benchmark customer is excluded.
Trajectory
Revenue has turned sharply: the latest quarter's $19.8 million is nearly double Q3 FY26's $10.3 million, after a low of $9.9 million in Q2 FY26. Gross margin expanded to 40.6% from 32.7% in Q3, and EBITDA improved to -$0.8 million from -$3.4 million. The company attributes the improvement to higher revenue, better capacity utilization, and favorable mix; earlier pressure came from lower-margin package-level products, assembly and warranty costs, freight, and tariffs.
The Model
The model projects FY+1 revenue of $160.0 million with EBITDA of $30 million (18.8% margin), and FY+2 revenue of $265 million with EBITDA of $67 million (25.4% margin). The FY+1 projection sits above management's $130M–$150M guide and is anchored by the year-end backlog and AI processor/hyperscaler orders; the FY+2 step assumes continued wafer-level burn-in adoption plus contribution from outside-the-guide options such as the top-tier benchmark customer or memory.
| Metric | FY2026 | Next FY (E) | Following FY (E) |
|---|---|---|---|
| Revenue | $51M | $155M | $240M |
| YoY Growth | — | +203.9% | +54.8% |
| EBITDA | −$11M | $23M | $46M |
| EBITDA Margin | -22.2% | 14.8% | 19.1% |
Projections are the median of 5 independent model runs. The model’s revenue sits 94.3% above analyst consensus.
On the July 14, 2026 call, management initiated FY27 guidance of $130 million to $150 million in revenue, representing 160% to 200% growth from FY26, and non-GAAP pretax net income of 18% to 22% of revenue. The base case assumes no memory revenue, and management said the bulk or majority of the range comes from current customers; the newly completed top-tier AI benchmark customer is excluded.
What Could Go Right — and Wrong
- Top-tier AI benchmark customer converts from pilot at the Taiwan contract manufacturer into system and WaferPak orders.
- Memory development agreement closes, adding memory-optimized FOX blade revenue beyond FY27.
- Hyperscaler's third device moves to wafer-level burn-in, expanding wallet share beyond package-level Sonoma.
- Silicon photonics customer calendar-2026 forecasts turn into additional system orders.
- Advanced-packaging bottlenecks intensify, accelerating known-good-die screening adoption.
- A major AI customer delays or re-plans its burn-in flow; the lead AI customer has moved all production burn-in to Aehr.
- Power-supply or other sole-source component shortages prevent on-time Sonoma deliveries.
- Hiring fails to keep pace with a 160–200% revenue ramp; Fremont was still on one shift.
- Fiscal Q2 FY27 Sonoma delivery concentration slips, compressing full-year revenue timing.
- Competitors developing full-wafer or bare-die burn-in products gain traction at a major customer.
Looking Ahead
The next 12 months center on converting a record backlog into shipped systems. Management has concentrated Sonoma deliveries in fiscal Q2 FY27, expects silicon photonics orders during calendar 2026, and says memory orders could arrive in FY27 with ramps in FY28. The top-tier AI benchmark customer's pilot at the Taiwan contract manufacturer is the clearest outside-the-guide catalyst.
- Fiscal Q2 FY27Sonoma delivery concentration — Bulk of hyperscaler Sonoma backlog ships; management calls Q2 a big quarter.
- Calendar 2026Silicon photonics system orders — New networking customer forecast additional systems during CY2026.
- FY27Memory development agreement — Orders possible FY27; ramps FY28; 12-18 months after agreement.
- FY27Top-tier AI pilot conversion — Pilot at Taiwan CM for current high-volume device; not in FY27 base guide.
- FY27Hyperscaler second-device ramp — Device has twice power per package; Sonoma expansion forecast.
Financials
Annual Summary
| Metric | FY2025 | FY2026 | TTM | YoY |
|---|---|---|---|---|
| Revenue | $59M | $51M | $51M | -13.6% |
| Gross Margin | 40.9% | 33.2% | 34.7% | 767bps |
| EBITDA | −$1M | −$11M | $7M | -707.1% |
| EBITDA Margin | -2.4% | -22.2% | -22.2% | 1,978bps |
| Net Income | −$4M | −$7M | −$7M | -86.8% |
| Free Cash Flow | −$12M | −$5M | −$25M | — |
| Net Cash | — | — | — | — |
Key Ratios (Trailing)
- P/E TTM—
- EV/EBITDA TTM—
- EV/Revenue TTM—
- Price/FCF TTM—
- Gross Margin (TTM)34.7%
- EBITDA Margin (TTM)-22.2%
- Net Margin (TTM)-13.9%
- ROIC-9.9%
- SBC / Revenue13.3%
The Company
Aehr Test Systems provides semiconductor test, burn-in, and stabilization systems and consumables used on wafer-level, singulated-die, and package-part devices. Its FOX-XP full-wafer systems can test up to nine 300mm wafers in parallel, and the WaferPak contactor is a proprietary consumable; Sonoma systems handle ultra-high-powered packaged-part burn-in at up to 2,000W or more per device. The products screen for early-life failures in AI processors, silicon photonics, memory, power semiconductors, and related devices.
AEHR operates one segment, with a principal Fremont, California facility of 51,289 square feet, a Philippines subsidiary facility of 6,458 square feet, and a newly signed Hsinchu, Taiwan office. The company has roughly 150 employees and about 250 people in direct contract manufacturers in lower-cost regions. An existing Southeast Asia contract manufacturer is now shipping Sonoma systems, adding more than 20 systems per month; FY26 capex was $2.1 million.
Business Segments
Competitive Landscape
The 10-K discloses that several companies have developed or are developing full-wafer and single-touchdown probe cards, and several are developing products for test and burn-in of multiple bare die and small modules. A Taiwan silicon carbide customer chose AEHR over Semi/Nexus Test. Management also says AEHR was first to demonstrate and ship wafer-level burn-in for AI processors.
- Semi/Nexus TestDisclosed as the competitor AEHR won against for a Taiwan silicon carbide customer; no further details in the supplied sources.
Supply Chain
AEHR sits between component suppliers and semiconductor fabs/OSATs, selling burn-in systems and consumables into AI processor, silicon photonics, and power-semiconductor production lines. The sourced disclosures list sole/limited-source component categories but do not identify individual suppliers by name.
More on AEHR: Earnings recap