AI burn-in is now the whole story
- Aehr has shifted from EV silicon carbide tools toward AI processor and data center test systems.
- Management says a top-tier AI processor supplier passed its benchmark and is moving to pilot production.
- FY27 revenue guidance is $130M-$150M, about 2.6x-3x the prior level management is using as the base.
- Memory is not helping yet, with management expecting zero memory revenue in FY27.
- The risk is no longer only whether the AI test works. It is whether pilots become large orders at good margins.
AI passed the big test
Aehr makes machines that burn in chips. Burn-in means running chips under stress so weak parts fail before they reach customers. That job matters more when chips are costly, hot, and hard to replace, which is why Aehr is chasing AI processors, high-performance computing, silicon photonics, and memory.
The bull case improved after Q4. Management said a top-tier AI processor supplier completed a wafer-level burn-in benchmark, beat expectations, and got results better than package-level testing. That customer now wants pilot production validation on a current high-volume device. If that pilot turns into volume orders, Aehr could grow much faster than a normal equipment company.
The SiC story also stopped getting worse. SiC, or silicon carbide, is used in electric vehicles and power electronics. That market had been soft, but Aehr said it received about $8 million of new SiC orders in the last month as global EV programs accelerate.
The bear case is still real. Memory has moved out in time, and management said it does not expect any memory revenue in the FY27 revenue guide. Customer concentration is also high. The stock needs the AI ramp to happen, because the valuation leaves little room for missed orders or weaker margins.
Systems first, consumables after
Aehr sells test systems and the parts that touch or hold chips during tests. The main wafer-level systems are FOX-XP and FOX-CP. The key consumables are WaferPaks, which are contactors that connect to wafers, plus BIMs and BIBs used in package-level burn-in.
The attractive part is the follow-on spend. When a customer adds a new chip design, it may need new WaferPaks or boards. In Q2 FY26, management said contactor revenue, including WaferPaks, BIMs, and BIBs, was $3.4 million, or 35% of total revenue for that quarter.
The weaker part is mix. Recent gross margin fell because production volume was lower and sales shifted toward package-level burn-in products. The company said those products carry lower gross margins than wafer-level products. Tariffs and higher assembly and warranty costs have also hurt margins.
Aehr is adding capacity for Sonoma package-level systems through a contract manufacturer. Management says this can support more than 20 additional Sonoma systems per month. That matters if AI orders arrive, but it also raises the need for clean execution.
Tools for hotter chips
FOX-XP wafer-level systems
FOX-XP tests and burns in many wafers before chips are packaged. It is central to Aehr's AI processor, silicon carbide, and silicon photonics push.
FOX-CP single-wafer systems
FOX-CP handles single-wafer production test. Aehr also sells it into the hard disk drive market.
WaferPak contactors
WaferPaks are custom contactors used with wafer-level systems. They can create repeat sales when customers launch new chip designs, though recent consumable demand has been pressured.
Sonoma package-level burn-in
Sonoma tests packaged chips, not full wafers. Aehr has built a higher-power version that supports up to 2,000 watts per device for AI processors.
HBM and HBF roadmap
Aehr is working with memory suppliers on high-bandwidth memory and high-bandwidth flash. This is a future option, not a near-term revenue driver, because management expects no memory revenue in FY27.
Silicon photonics test
Silicon photonics uses light to move data in data centers. Aehr reported a major new customer win with an initial order for multiple high-power FOX-XP systems.
AI now dominates the mix
The mix uses the company view cited in the latest internal thesis for annual revenue: AI accelerators, CPUs, and network processors were about 71%, optical device test and burn-in was about 20%, and the rest included SiC and other markets. Customer concentration remains a major caveat.
What can still break
Pilot does not become volume
High impact · Medium oddsThe biggest benchmark risk has improved, because the top-tier AI processor supplier passed testing. The next risk is commercial. Aehr still needs the pilot production validation to turn into high-volume orders.
Too much revenue from too few customers
High impact · High oddsAehr's order book depends on a small number of large AI and hyperscale customers. In fiscal 2025, two customers were about 39% and 15% of net sales. A pause by one buyer can move revenue and margins quickly.
Lower-margin mix takes over
Medium impact · Medium oddsAehr is selling more package-level burn-in products after expanding the Sonoma platform. The company has said these products have lower gross margins than wafer-level products. If the AI ramp is mostly lower-margin systems, revenue can rise while profit disappoints.
Memory stays a future story
Medium impact · High oddsHBM and HBF could be large markets for Aehr, but they are not contributing yet. Management said it expects zero memory revenue in FY27, even in the $150M guide. That means the near-term thesis rests on AI processors, silicon photonics, and SiC.
China IP and trade friction
Medium impact · Medium oddsAehr has filed a lawsuit against a local Chinese supplier. It also faces barriers in China from geopolitics and U.S. tariffs. These issues can limit sales, raise costs, or distract management.