Title

Presentation

Aehr Test Systems beats earnings expectations. Reported EPS is $-0.01, expectations were $-0.015.

Operator: Greetings. Welcome to the Aehr Test Systems Fiscal 2025 Fourth Quarter and Full Year Conference Call. At this time, all participants are in a listen-only mode. A question and answer session will follow the formal presentation. Please note this conference is being recorded. I will now turn the conference over to your host, Jim Byers, of Pondell Wilkinson Investor Relations. You may begin.

Jim Byers: Welcome to Aehr Test Systems Fiscal 2025 fourth quarter and full year financial results conference call. On today’s call are Aehr Test Systems’ President and CEO, Gayn Erickson, and CFO, Chris Siu. Before I turn the call over to Gayn and Chris, I’d like to cover a few quick items. This afternoon, right after market close, Aehr Test Systems issued a press release announcing its fiscal 2025 fourth quarter and full year results. That release is available on the company’s website at aehr.com. This call is being broadcast live over the Internet for all parties, and the webcast will be archived on the Investor Relations page of the company’s website. I’d like to remind everyone that on today’s call, management will be making forward-looking statements that are based on current information and estimates and are subject to a number of risks and uncertainties that could cause actual results to differ materially from those in the forward-looking statements.

These factors are discussed in the company’s most recent periodic and current reports filed with the SEC. These forward-looking statements are only valid as of this date, and Aehr Test Systems undertakes no obligation to update the forward-looking statements. With that, I’d like to turn the conference call over to Gayn Erickson, President and CEO.

Gayn Erickson: Thanks, Jim. Good afternoon, everyone. Thank you for joining us on Aehr Test Systems’ fiscal 2025 fourth quarter full year earnings conference call. Before we begin, I’d like to thank our customers, employees, and partners for their dedication throughout this transformative year of execution, expansion, and strategic diversification for Aehr Test Systems in fiscal 2025. I’ll start with an update on the primary markets Aehr Test Systems is targeting for semiconductor testing and burn-in as well as the significant progress we’ve made this year in new markets. After that, Chris will give a detailed review of our financial performance, and finally, we’ll open up the floor to your questions. I just want to also point out that we’ve had a lot of inbound questions related to the AI market and what that means to Aehr Test Systems.

So we’ll be doing a deep dive as we have often done in other markets today to hopefully let people really understand how Aehr Test Systems is playing in that. This past year, we made significant progress expanding into additional key markets for semiconductor burn-in solutions, including artificial intelligence processors, gallium nitride power semiconductors, data storage devices, silicon photonics integrated circuits, and flash memory. This diversification of our markets and customers is significant, given our revenue concentration in silicon carbide devices used in electric vehicles during our previous fiscal year. Silicon carbide wafer level burn-in accounted for over 90% of our revenue in fiscal 2024, whereas it made up less than 40% of our revenue this fiscal 2025.

In contrast, the burn-in of artificial intelligence processors represented 0% of our revenue last year but this year accounted for over 35% of our revenue. And we had three companies representing over 10% of Aehr Test Systems’ revenue this year with two of these representing new markets and customers. As we grow, we expect that expanding into new markets and customers will not only allow us to grow faster but also do so sustainably. The main growth areas for us in markets beyond silicon carbide included production wafer level burn-in of AI processors, package part burn-in for qualification and ongoing process monitoring of AI processors, and also production package part burn-in and screening of those AI processors. We also had wafer level burn-in of gallium nitride semiconductors and silicon photonics integrated circuits wafer level burn-in.

And while there was only a small amount of revenue in the fiscal year from wafer level burn-in hard drive components, about 10% of our order bookings for the fiscal year came from this new market, all of which we expect to ship and generate revenue from during this fiscal year now 2026. Looking back on the year, we’re excited about the significant progress we’ve made with the key initiatives to expand these total addressable markets, diversify our customer base, and develop new products capabilities, and capacity all aimed at driving revenue growth and increased profitability. One of our biggest milestones this past year and what we believe is currently the most important for future revenue growth was the completion of development validation shipment, and customer acceptance of the first-ever wafer level burn-in system for AI processors.

Delivering the industry’s first wafer level burn-in solution for the AI processor market, the only one of its kind in the world, marks a major technological and commercial milestone and significantly expands the market potential for our FOX wafer level test and burn-in systems. Our new high power FOX XP wafer level burn-in system can test up to nine 300-millimeter AI processor wafers at the same time. This achievement is a result of extensive development efforts over the last decade in test technology, particularly in delivering massive amounts of power and current to a wafer during test, wafer contacting technology, and thermal control and heat removal. As well as wafer handling and automation. When our lead AI customer first approached us about testing and burning their AI processors at the wafer level, it wasn’t clear if this was even technically feasible.

Even with our proprietary and unique technology. We leveraged the technical capabilities we had developed over the years along with the technology and design for test methods used in state-of-the-art wafer foundries to meet these unique test and burn-in stress requirements. This included applying thousands of amps of current to a single wafer to test devices capable of withstanding thousands of watts and then not only doing it with one wafer, but nine wafers simultaneously. We also expanded our proprietary WaferPak contactor to support very high current testing capabilities including the ability to adapt the thermals to create uniformity across a wafer that, by definition, is not otherwise uniform during burn-in conditions. With Aehr Test Systems demonstrating and now shaping the first-ever solution for wafer level burn-in AI processors, in partnership with this customer’s outsourced assembly and test partner, one of the largest OSATs worldwide, we have proven that our high power FOX XP Multi wafer systems improve proprietary WaferPak contactors are a viable solution for high volume testing and burn-in of AI processors and wafer form.

This approach eliminates the need to burn-in these devices in package or system form where test costs and yield losses due to failing devices during burn-in are significantly higher and have a much more significant impact on overall manufacturing yield. Many AI processor companies are talking about billions of dollars of device a year, with the largest AI processor company in the world shipping over $100 billion worth of processors in the data center applications this year alone. Even a 0.1% increase in yield by shifting the burn-in of devices from the system or heterogeneous package level to wafer level is very significant. Today, burn-in related screening on early life failures at the system or package level causes the entire package or system to be discarded, due to the inability to repair these devices at this stage.

Moving this screen to wafer level allows devices that would otherwise fail during screening at the or system level to be removed before they’re packaged or worse put into the system level. We believe using wafer level burn-in will result in savings in manufacturing costs, increases in revenues associated with the limited supply of these devices, and a reduction in field-related failures and warranty costs. As you can afford to put more screens in place or burn-in longer to ensure the highest quality and reliability of devices. We’re also receiving feedback from potential customers that doing this screening at wafer level is not only cheaper overall but requires less electrical power from the grid which has significant benefits. An important part of our story we have discussed for years is the evolution of semiconductors that has driven the increased need for wafer level burn-in.

This includes the fact that semiconductors are becoming less reliable with the transition to smaller geometries, devices are physically larger, and more devices are being developed on compound semiconductors like silicon carbide and gallium nitride, which require additional burn-in and stress testing such as Aehr Test Systems provides to meet the stringent quality and reliability needs of their customers in the end markets. AI processors and other high-performance central processing units or CPUs and network processors are also facing limitations related to their physical size due to the reticle limit in semiconductor manufacturing. The reticle limit is the maximum area that can be exposed in a single pass of the lithography equipment. Due to the reticle limit, a single chip or die cannot exceed this maximum area.

This is a physical constraint imposed by the manufacturing process. AI models, especially large language models, require massive amounts of computation and memory. This translates to the need for increasingly large and complex chips. The GPUs used for AI training have already reached the reticle field limit. In order to overcome the reticle limit that prevents them from building a single massive chip, manufacturers are using chiplets. Which are smaller die that can be interconnected to form a larger system. This approach effectively circumvents the reticle limit and allows for much larger total transistor counts. Advanced packaging technologies such as COOS or chip on wafer on substrate and SOIC or system unintegrated circuits enable the integration of multiple chiplets onto an interposer allowing for complex high-performance systems that exceed the size of a single reticle.

As AI processors require increasingly large and complex designs, chip-like architectures and advanced packaging technologies are being used to overcome this limitation, and enable the continued scaling of AI compute power. The reason these matters to Aehr Test Systems is that these devices all need production burn-in screening to remove early failures that would otherwise occur during the lifespan of the AI processor. These failure rates are unacceptable and costly, impacting the end customer and increasing warranty costs for the supplier. In many cases, they can also pose safety issues, especially for processors used in autonomous or driver-assisted vehicle technologies. The screening and burn-in durations vary by process and device, but generally range from one to several hours or even twenty-four hours or more depending on the desired quality and reliability level for the end application or customer.

These GPUs, particularly those used in data centers and in the creation and use of large language AI models, are not the commodity consumer semiconductors of the nineties and two-thousands. These are not chips used in $200 graphics cards for gaming. Instead, these are nodes worth tens of thousands of dollars operated in parallel thousands at a time. If one node fails, you can completely disrupt the development or construction of the entire language model. As we’ve discussed with automotive and other applications, reliability is critically important for these customers. Aehr Test Systems now offers a high volume production solution for package level burn-in with our new Sonoma product line following the acquisition of InCal Technology last year.

For customers seeking to perform production screening of these devices in package form, we now provide a highly cost-effective solution with an upcoming fully automated Jetix trade to trade device handling and testing. Before our FOX XP wafer level test and burn-in system, the only solution for doing the screening was products like our Sonoma system. And while this screening step is cost-effective, and we believe our Sonoma systems offer the lowest cost solutions on the market for packaged part burn-in of AI processors, it not only weeds out defective processors, or memory, but also results in discarding highly expensive advanced packages. Such as the COAS package and substrate along with all of the other devices packaged in the failed device.

For example, multiple new AI processors feature two or more AI, a six, or GPO processors in the advanced package. With each AI processor die containing up to four high bandwidth memory or each HBM stacks. Totaling up to eight or more HBM stacks per package. Keep following here. Each of these HBM stacks can beat eight die or more in the future, meaning it has eight stack memory dies per HBM stack. You go through all the math, and there’s a total of up to 64 or more HBM dies and two or more AI processors plus a very expensive co-op package substrate per package. If one of the AI processors or one of the HBM die fails during the production burn-in after packaging, all of the other die plus the coas substrate are discarded. You can see the cost impact of performing burn-in at the package part level.

If you take this step if you take this a step further, companies also perform a burn-in at the system level when the GPU or AI processor multi-die package is installed on a computer system, printed circuit board. Along with the device power supplies, heat sinks, and supporting infrastructure, such as all the high-speed interconnect technology for AI processor to AI processor communication. Performing burn-in at this stage impacts cost and yield even more significantly. You can see why the industry is showing such interest in our ability to test these devices at the wafer form. In addition to this lead customer for AI wafer level test, we have now received multiple inbound requests from several high-profile processor companies that are very serious about wafer level testing.

This ecosystem is very small, and having demonstrated that wafer level burn-in of high power AI processors is feasible, we’re gaining visibility beyond our own sales and marketing efforts with the growing recognition that moving the AI processors and CPUs to wafer level is overwhelmingly advantageous for both a cost perspective and for yield. We’re extremely busy right now engaging with multiple companies who are asking can you test my parts? The conversation is can Aehr Test Systems do it? Not do we want it. I’m very excited to report today that one of these companies has now asked us to move forward with an evaluation for wafer level testing of their devices with one of their current high volume processors. This feasibility study will allow them to see the real advantages in performance of doing a production burn-in at wafer level instead of in package or system level form as they do today.

Based on what they have shared with us, we believe that if evaluation succeeds, they plan to transition to high volume production wafer level testing which would be a significant opportunity for Aehr Test Systems. While this evaluation involves new wafer packs for their specific wafers, we believe that Aehr Test Systems can address the needs in the near term with our proprietary AI processor optimized wafer packs designed specifically for these devices. We also have systems needed to run their wafers on our floor today. Demonstrating this capability as well as fully automated aligners to showcase the automation the 300-millimeter wafer handling. We expect this evaluation to take one or two quarters to complete fully. At the same time, we’re working with this customer to determine their production capacity needs and discussing lead times to meet their requirements.

Their capacity requirements are significant. And we feel we have the manufacturing capacity for systems aligners, and wafer backs to meet the potential demand if we’re successful with this evaluation and they decide to move to wafer level burn-in using our solution. We also expect to move to evaluation phases with other AI companies during this fiscal year and believe we can capture a meaningful share of the total production burn-in for AI processors with our FOX wafer level test and burn-in systems proprietary WaferPak contactors. So let me spend a few minutes on some of the other markets quickly. And I’ll start with packaged part burn-in. As I mentioned earlier, we also offer customers the option to perform their package part burn-in screening as well as their qualification reliability characterization for their GPUs, AI processors, CPUs, and network processors.

We completed our acquisition of Intel last July 31, expanding our product portfolio to include their highly regarded packaged part reliability burn-in and test solutions. Especially their ultra-high power capabilities for AI processors, GPUs, and computing processors. Since that date, Aehr Test Systems has shipped more packaged part burn-in systems than Intel did in the last three years. It is a record-breaking sales achievement for the qualification and production burn-in of AI processors. We’re very excited that with the added capabilities and resources from Aehr Test Systems, we have been able to ramp up production to levels that Intel had never achieved, meeting the demand from AI processor companies for the qualification and production of their devices.

With the addition of a significant number of people, processes, and scale, we’ve been able to shorten lead times, maintain low cost, address quality, and do this in a high reliability platform. Which has been overwhelmingly positive for customers. As a result, we won our first production AI processor customer for packaged part burn-in during the fiscal year, receiving initial volume production orders for the multiple Sonoma Ultibi cars systems. This customer is one of the premier large-scale data center hyperscalers. That is making their own AI processors and is growing this capacity significantly. These are their first devices that use a production burn-in system at the package part level instead of at the system level. They plan to ramp this device over the next year and are already discussing their next generation process as well as the one after that with Aehr Test Systems, to ensure we can meet their production capacity needs.

We said before that one of the best things about this acquisition is that it gives us a front-row seat to the future requirements of a large number of these AI processor customers providing us with visibility into the production burn-in needs. As a result, some of these customers from the packaged parts side are coming to us asking about our wafer level burning capabilities. Aehr Test Systems is the only company in the world that offers both a wafer level and packaged part burn-in system for both qualification and production burn-in of AI processors. We can provide them with options and show direct side-by-side comparisons of cost of test capacity output, footprint, operational cost, and impact on yield based on how they decide to do their burn-in.

We’re in the perfect position to help them while also remaining balanced so we can tell them yes regardless of how they want to do their burn-in. We’re very excited about all our new AI product offerings and the expanded total addressable market they bring to Aehr Test Systems and we look forward to discussing our progress to further capitalize on this new market as we move through our new fiscal year. Another key milestone this past year was expanding the production level burn-in for gallium nitride power semiconductors. We secured an additional order for our FOX XB high power wafer production system with high volume, for volume production of GaN devices from a leading automotive semiconductor supplier and a key player in the GaN nitride power semiconductor market marking their commitment to advancing volume production wafer load burn-in of their GaN devices using our XP platform.

This achievement expands our production wafer level burn-in market for power semiconductors beyond silicon carbide applications used in electric vehicles, data center power conversion, and solar to now include GaN, a high-performance compound semiconductor optimized for mid-power applications such as data centers, solar energy, automotive systems, and considerable electronics and PCs. Additionally, we’re in discussions and engagements with multiple other potential new GaN customers about their needs. GaN’s a new and exciting semiconductor technology with high-value applications including automotive power conversion, solar inverters, and solid-state transformers and breakers. GaN offers a much broader application range than silicon carbide. And is poised for significant growth in the coming decade.

We’ve also made significant progress in the hard disk drive market. This past year, a lead customer began ordering multiple FOX CP single wafer production test and burn-in systems, reaching an integrated high power wafer prober for their new high volume parts. In a new application for burn-in and stabilization of new devices in hard disk drives. These are follow-on system orders to the first order received all the way back in 2019. As we stated in previous calls, their plans for this new product were delayed during the pandemic, but they continue to work on this new device continuously over the five years to ensure the performance and reliability of their devices. We understand from several analysts and shareholders that this customer has publicly called out our Fox Systems as a key contributing factor in helping them achieve a long-term reliability need to this market.

This customer is one of the top suppliers of data storage devices. We’re very excited to start this production ramp after all these years of working with them on qualification and process development. During our last earnings call, I noted that the high power probers for our FOX CP for this HD customer are sourced from Japan, and it was unclear how the tariff uncertainty might affect the timing of receiving these probers. At that time, we were hoping to receive the shipment by May, but because of tariff uncertainties, these probers did in fact get significantly delayed. We just received the first one last week. We’re working quickly to do the integration and engineering steps needed to finalize the test cell to be able to make the first shipments this quarter.

In addition to multiple systems and backlog, the customers told us that they will be purchasing additional systems both in the short term and over time. I know it must be a broken record to hear terms like uncertainty around tariffs, on many company earnings calls, but this is still the case. Despite this, we’re extremely excited about our growth opportunity for our wafer level solution for the HDD market and look forward to updating you further on progress next quarter. Now turning to silicon photonics ICs. This market continues to demonstrate market adoption for optical chip-to-chip communication and optical network switching. Several companies, including AMD, NVIDIA, Intel, TSH, and GlobalFoundries, have announced product roadmaps for devices that utilize optical chip-to-chip communication.

We have several customers in this space. At last count, it was five to six customers. With one of the customers being an OSAT that purchases our tools for one customer but sparking it to others. We’ve seen a significant number of new WaferPak designs from our installed base of systems for new designs that they use for qualification development work on their FOX wafer level test and burn-in systems. We also now offer a new system with higher power 3,500 watt per wafer configuration to meet the needs of new higher power wafers for optical IO and chip-to-chip communication devices. This is also available as an upgrade to our FOX NP systems, for low volume production as well as for our FOX XP nine wafer production systems. Recently received another order for an upgrade to one of the Fox XPs we shipped a few years ago, that includes upgrading to include our new integrated WaferPak auto liner.

Which provides fully hands-free factory automation of silicon photonics integrated circuit wafers. We also have forecast for new systems for incremental capacity this fiscal year for both systems and WaferPaks. We’re well prepared with expanded manufacturing capacity for Fox high power systems remain enthusiastic about the silicon photonics market. Especially for the new application of silicon photonics integrated circuits and optical chip-to-chip communication which we see as a significant market opportunity for our products. Seems odd to wait this long to talk about silicon carbide, but let me talk a little bit about that market as well. The silicon carbide power semiconductor market remains a significant opportunity for Aehr Test Systems, and we believe we’re well positioned to continue to grow with our current customers in this sector as well as add some additional customers in this space over time.

Despite a slowdown in the growth of electric vehicle shipments, electric vehicles are still growing significantly worldwide, and we believe the silicon carbide market continues on a robust long-term growth trajectory. Demand for silicon carbide remains significantly driven by battery electric vehicles, but silicon carbide devices are also gaining traction in other markets, including power infrastructure, solar, and various other industrial applications. This quarter, we shipped our first configuration of the FOX XP, which can test 18 wafers at a time in a single system with support for our high voltage test resources that can test devices up to 2,000 volts in wafer form. The system also includes a proprietary arc suppression technology that prevents the devices from electrically arcing at these high voltage while testing all devices at a time in a single insertion.

On each of 18 wafers. This capability has already proven in our nine wafer configuration, but is now extended to the 18 wafer system. It’s also capable of being directly docked to our fully automated WaferPak aligner that takes industry standard wafer cassettes and FUPs to allow full factory integration. We believe we’re well positioned in the silicon carbide market as we have a large customer base and the industry-leading solution for wafer level burn-in. So lastly, a little bit on our flash memory proof of concept project that we’ve been working on this year. As noted in earlier calls, we’re collaborating with one of the world’s leaders in flash memory to demonstrate the capability and cost-effectiveness of our FOX XP platform for high volume production wafer level testing and burn-in and flash memory wafers.

This is very exciting because we believe Aehr Test Systems can successfully demonstrate how to create a high density, high power, fully automated test cell which will help us move to the next development phase. That next step involves working together to develop a next-generation test system specifically designed to meet this customer’s needs. Although this memory validation benchmark has taken us a bit longer than expected, due to shipment delays in some of the components of this integrated system, the new MEMS-based fine pitch wafer packed full wafer contactor is in-house and ready to complete the benchmark. We’re very encouraged, and we hope to generate data and results this quarter with the aim of completing the benchmark by next quarter.

New technologies in NAND are driving new requirements for wafer level burn-in to address the manufacturing and negative yield implication of testing these devices at package or system level. We believe that Aehr Test Systems’ Fox Waveflow Test and Burn-in platform combined with our proprietary WaferPak full wafer contactors is well positioned to offer competitive and cost advantage solution in this market. Looking ahead, and concluding, Aehr Test Systems is well positioned to capitalize on growth in the overall semiconductor market. We remain focused on addressing the critical reliability requirements of next-generation applications and leveraging key megatrends shaping our industry. Today, reliability is a vital priority across diverse sectors including combustion, electric vehicles, data centers, infrastructure electrification, and an expanding range of AI applications.

As we enter fiscal 2026, we’ve established the and capacity to support significant growth. This was the purpose of the investments we made this past fiscal year. Including the upgrade of our manufacturing facility, including upgrades to power and infrastructure, consolidating package and wafer level burning under one roof, and implementing the necessary processes to support a very high volume of both wafer level and package level test and burn-in systems, our proprietary WaferPak contactors. These foundational efforts are now complete. In the year ahead, we plan to increase our research and development investment support further product enhancements, expand our R&D resources, hire additional talent to serve our growing AI customer base, and enhance automation to improve scalability.

Beyond these initiatives, the new fiscal year will focus on securing and executing orders. We believe that nearly all the opportunities and market verticals we discuss today will experience order growth in fiscal 2026. The one exception may be silicon carbide, as customer forecasts for this market are back half loaded with stronger growth expected in our fiscal 2027. Still, there are many variables, and silicon carbide may end up growing faster than expected. Given the market share shifts currently underway and our lead customers increasing market share in the industry. During our previous earnings call, we announced the temporary withdrawal of our financial guidance following the US administration’s tariff announcements just a few days earlier.

At that time, we were concerned about the potential secondary impacts on our current and prospective customers as well as the possibility of pauses or delays in customer orders shipments, or supply chain deliveries. We remain very confident in Aehr Test Systems’ long-term outlook but we are still seeing the impact of tariff-related uncertainty on the timing of specific orders. Particularly in our first quarter. As a result, we’ve chosen to maintain a cautious approach and are not reinstating specific guidance at this time beyond what we’ve already stated, which is that we anticipate order growth across all segments in this fiscal year with this possible exception of silicon carbide. We’re very optimistic about our growth opportunities in all the segments we’ve discussed our ability to meet the potential demand of these markets.

And with that, I’ll turn it over to Chris.

Chris Siu: Thank you, Gayn. Before I review our financial results, I’d like to provide an update on the integration of our InCal acquisition. Which we completed on July 31st of last year. Since the acquisition, Aehr Test Systems has dedicated significant financial and human resources to ensure a successful integration of InCal into our operations. We have fully migrated InCal’s financial records into our Oracle NetSuite cloud ERP system. And integrated their HR and manufacturing functions into Aehr Test Systems’ product information system. Additionally, we have completed the transfer of all inventory as well as comprehensive documentation of product designs source code, and assembly and test instructions into Aehr Test Systems’ processes.

I’m pleased to report that our plan to consolidate personnel and manufacturing into Aehr Test Systems’ Fremont facility was complete by the fourth quarter of fiscal 2025. And we successfully closed the InCal facility on May 30, 2025, ahead of our original schedule. I want to express my sincere thanks to our teams for their commitment, dedication, and outstanding execution throughout this integration. As a result of the consolidation, we incurred one-time restructuring charges of $864,000 in our fiscal fourth quarter related to the closure of the InCal facility. With the integration of the two companies, we’ll be able to create synergy and reduce our facility cost by over $800,000 per year going forward. Turning to the full year results. We reported revenue of $59 million, down 11% year over year.

Our full year non-GAAP gross margin was 44% compared to 49.6% in the prior year. Our full year non-GAAP net income was $4.6 million or $0.15 per diluted share in fiscal 2025. Compared to non-GAAP net income of $35.8 million or $1.21 per diluted share in fiscal 2024. Which included the impact of a one-time tax benefit of approximately $20.7 million resulting from the release of the company’s full income tax valuation allowance. Our annual bookings in fiscal 2025 were $61.1 million, up over 24% compared to $49 million in the prior fiscal year. The increase in bookings was primarily related to the sales of AI processor burning systems. Wafer packs and burning module boards, partially offset by lower customer orders related to silicon carbide wafer packs.

Our backlog as of year-end was $15.2 million with $1.1 million bookings received in the first five weeks of the first quarter of fiscal 2026. We now have an effective backlog of $16.3 million. Turning to our Q4 performance. We’re excited about our continued momentum in penetrating the artificial intelligence market. With AI processors burning now accounting for over 35% of our business this year. Compared to zero last year. For the fourth quarter, we had three customers representing over 10% of total revenue. And two of these customers target the AI market. Revenue for the fourth quarter totaled $14.1 million, a 15% decrease compared to $16.6 million in Q4 last year. The year-over-year decrease was primarily due to a delayed shipment of a FOX CP system that was forecasted to be shipped to our hard disk drive customer.

Because of tariff-related uncertainties, COVID source from Asia to support the Fox CP system were delayed. We now expect to complete this shipment in our current quarter Q1 of fiscal 2026. WaferPak revenues were $4.2 million and accounted for 30% of our total revenue in the fourth quarter. WaferPak revenues continue to represent a sizable revenue stream for our business. Driven by the ongoing demand for new WaferPak designs from both existing and new customers, as they secure new end customer designs and strive to meet their market requirements. We are pleased with the significant progress we made integrating products from our InCal acquisition into our product portfolio to on emerging opportunities in the AI market. Sales of our Sonoma, Tahoe, and Echo package part burning systems continue to contribute strongly accounting for 44% of our fourth quarter revenue.

We believe our strategy to expand Aehr Test Systems’ product offerings and diversify beyond silicon carbide applications is gaining meaningful traction in the marketplace. Non-GAAP gross margin for the fourth quarter was 34.7% compared to 51.5% in the same period last year. Non-GAAP gross margin decreased primarily due to lower overall revenue level. Compared to Q4 last year. And less favorable product mix. Additionally, we incurred high manufacturing overhead due to under absorption as our manufacturing capacity utilization was lower during the renovation of our Fremont site. And the consolidation of inventory from the InCal facility. Non-GAAP operating expenses in the fourth quarter were $5.4 million reflecting a 6% increase from $5.1 million Q4 last year.

This year-over-year rise is primarily attributed to the inclusion of InCal’s operating expenses in our financial results. As well as higher legal and professional service fees. We anticipate incurring additional legal expenses in the upcoming quarters as we continue to protect our intellectual property rights in China. Non-GAAP net loss for the fourth quarter excluding the impact of stock-based compensation amortization of intangible assets and restructuring charges, $248,000 or negative 1p per diluted share. In line with the Street consensus. This compares to a non-GAAP net income of $24.7 million or $0.84 per diluted share in the fourth quarter of fiscal 2024. Which as a reminder included a one-time tax benefit of approximately $20.7 million resulting from the release of the company’s full income tax valuation allowance.

Moving to the balance sheet. At the end of Q4, our cash cash equivalents and restricted cash totaled $26.5 million down from $49.3 million at the end of Q4 fiscal 2024. During fiscal 2025, we used $11 million to acquire InCal, which enabled us to enter the AI market and secure key customer relationships with its superior packaged part burn-in product, Sonoma. Additionally, we spent $5 million on CapEx to support the consolidation and upgrade of our Fremont manufacturing facility and headquarters. And used $3.6 million to procure inventory and $3.8 million for the remaining working capital. We have no debt and continue to invest excess cash in money market to generate interest income. Let me provide an update on the action claims and derivative lawsuit filed against Aehr Test Systems on December 3, 2024.

We are pleased to report that on May 16, 2025, the court-appointed lead plaintiff for the class action lawsuit elected to voluntarily dismiss the case. With all parties bearing their own fees and costs. On June 9, 2025, the court also dismissed the derivative lawsuit without prejudice. The company believes the claims in all these cases were without merit. Following the voluntary dismissal of the shareholder class action, and the court’s dismissal of the consolidated derivative action, no related proceedings are currently pending. As Gayn mentioned, considering the secondary effect of the tariff announcements on our current and potential new customers, along with the uncertainty this quarter regarding possible pauses or delays in customer orders shipments or supply chain delivery delays.

We’re temporarily withholding our guidance for our fiscal 2026 year. And we’ll reassess our guidance policy as clarity develops. Despite the uncertainties around tariffs, we’re excited about the long-term growth opportunities across our more diverse target markets. We’re especially encouraged by the progress we have made in addressing the artificial intelligence processes market and other potential high-growth sectors, including gallium nitride power semiconductors, data storage devices, silicon photonics integrated circuits, and flash memory. To succeed in these new markets, we plan to increase our research and development investments by expanding our R&D resources and hiring additional talent in the US and The Philippines to support our growing AI customer base and increase automation for scalability.

With more R&D talent, a strong infrastructure, and enhanced manufacturing capabilities in place, we’re more prepared than ever to achieve our growth objectives. Lastly, you may have noticed that we are holding this earnings call one week earlier than usual. Because of the CEO Summit has been pushed back from July to October. And will be held in Phoenix instead of San Francisco. Looking at the investor relations calendar, Aehr Test Systems will be participating in two upcoming conferences over the next month. We’ll be meeting with investors virtually at Needham’s 6th Annual Semiconductor and Semicap one-on-one conference on Wednesday, August 20. The following week, we’ll meet with investors in person on Tuesday, August 26. At the Jefferies Technology Summit Conference in Chicago.

We hope to see some of you at these conferences. This concludes our prepared remarks. We’re now ready to take your questions. Operator, please go ahead.

Operator: Thank you. At this time, we will be conducting a question and answer session. If you would like to ask a question, please press 1 on your telephone keypad. A confirmation tone will indicate your line is in the question queue. You may press 2 if you would like to remove your question from the queue. For participants using speaker equipment, it may be necessary to pick up your handset before pressing the star keys. One moment please while we poll for questions. Once again, please press 1 if you have a question or a comment.

Operator: Our first question comes from Christian Schwab with Craig Hallum. Please proceed.

Christian Schwab: Question Gayn, I have received a lot of regarding your most recent slide in your investor deck with a lot of well-known marquee names. And I’m just is that should investors think of that list as a list of current and previous customers? Or does it include, you know, maybe names of prospective customers you know, such as new AI customers that you’re working with or new silicon photonics customers, etcetera. How should we be thinking about that slide?

Gayn Erickson: Okay. So when we introduced that slide, we updated the customer slide a few weeks ago. Compared to the deck that we’ve had on our website prior to that. And we were actually at a conference that had a public speaking. It is recorded and available to all people through our website. In that specific conference, I pointed out that we have updated the customer list to reflect the current customers from the InCal package part burn-in side of our business. Those are customers, not prospects. And, yes, there’s some notable names on there. Not every single customer is still on there. Actually, there’s been some historical 10% customers that we had we had specific agreements with or limitations of ever saying their names publicly other than reference them when they were doing 10% of our revenue. But, yeah, there’s some good names on there.

Christian Schwab: Great. And that is it’s released to 10% customers. Are you, since you didn’t mention a few times without naming the names, is that going to be in your K when you file it, or are you not gonna name the 10% customer?

Gayn Erickson: So we’re, yeah, we’re the new SEC rules do not require you to name it. So unless we already have prior ranged agreements with the customers to name them, we’re no longer doing that. Prior to that SEC rule, we could name them even if the customers objected, if you will, but we’re not doing that now.

Christian Schwab: Great. And then as the you know, the AI opportunity seems quite significant. I know you in previous conference calls said that that market could be materially bigger than what you had previously said that the silicon carbide opportunity could be over time if wafer level burden was used by, you know, many people. Do you have can you give us any idea of not this year or next year, but, you know, over time, you know, have you walked through the math? To what you Yeah. We as you imagine, we have. We have. So similar to how we built up the original silicon carbide models, that took a look at, say, the target applications silicon carbide, which were primarily the electric vehicles, how many EVs, how many components would be in it, etcetera, etcetera.

You could come up with how many wafer starts that would require in, say, 2030. And I know that you had put some models together at that time. There were about 4 million wafer starts. We looked at twelve-hour burn-in times. Single insertion with our systems. Long story short, we saw that the total market was somewhere three, 350 of our systems with ASPs, about $4 million apiece or something like that. If you look at the AI market, the AI market, interestingly, in that same timeline, may actually be half as many wafers, which seems a little odd. But the 300-millimeter wafers and actually don’t test them in one single touchdown. It’ll take multiple touchdowns to test these wafers because they might have 20,000 watts of power on them, and we’re testing three, 4,000 watts at a time.

You can go through the math. At even significantly lower average burn-in times than, say, the silicon photonics I’m sorry, the silicon carbide is. They you go through the math and the market is you know, three to five times larger. Than the silicon carbide was. One of the things that we’ll be looking at is what are the burn-in times You know, we have we’re doing burn-in of customers around the world, and we have customers that are doing one-hour stress time you know, some four-hour burn-in times, and twenty-four-hour burn-in times, for example. All the way up. So it’ll be interesting to see, and we think that’s related to you know, how much the longer the burn-in time, the higher the quality ends up being to the end customer. And so it kinda depends on both availability, and what the target customers required and how critically important it is to get to the quality levels.

So there’s a little bit of dynamics in here, but there’s no way to do the math and not come out with it being significantly larger.

Christian Schwab: Fantastic. No other questions. Thanks, Gayn.

Operator: Thank you.

Operator: The next question comes from Jed Dorsheimer with William Blair. Please proceed.

Jed Dorsheimer: Hi. Yes. Thanks for taking my question. I guess first one, either Gayn or Chris, is just the step down in gross margin. Quarter to quarter or year over year too. I’m assuming that it’s a mix issue in terms of drop off of, wafer pack consumables. And a mix towards InCal. Is there anything else that should be called out with respect to shift in margins, of course, lower revenues and to spread the fixed cost. But is it primarily mix in revenue levels?

Gayn Erickson: There’s two types of mix in this thing too. There definitely was a mix in that we actually with some of the wafer level burden orders, didn’t materialize as we were kind of thinking at the time. We actually pulled in some package part systems and the packaged part systems have a little lower margin. The consumables have a lot less margin than our wafer packs, for example. But the other piece is during that quarter, we had the full burden of the InCal facility as well, which we talked about now not having kind of carrying forward. So we had the double whack of both facilities going on with, you know, primarily just packaged part burn-in revenues. And some wafer packs. So I think as we go forward, you know, the same quarter would be materially better. And, of course, we have the potential of much higher volumes as we go forward.

Chris Siu: Yeah. Just to add on that, also the utilization was not as high as you before because we were moving. Some of the folks actually were helping with the moves in the operations. Not building in Not building product instead of so we got to expense those labor costs.

Jed Dorsheimer: Got it. That’s helpful. Thanks. Then, Gayn, just maybe if you could you know, I know your tenor around AI processors is certainly skewed much more positively over the past year. And you know, I know that silicon photonics, you took time to call that out. Copackaged optics are becoming a more meaningful part of the design. I’m just curious what was you know, is it just the size of the TAM There seems like there was a specific shift in terms of your excitement around this market, and I’m is it a function of your technology and the mode that you have If you could wouldn’t mind just explain and maybe I have that wrong. So but Oh, no. I yeah. I think you’re I believe I think it’s people off that’s you know, what what they makes you lose sleep at night and what makes you excited.

There’s a material difference in our story related to AI from you know, nine months ago, maybe six months ago. Nine months ago, we were serious when we said we’re doing this evaluation with the first AI customer And while we believe it will work, we still have to work through it. We also didn’t finalize the burn-in times, etcetera. Now that we’ve proven that it works, understand the size of the market through the burn-in times, and candidly, delighted the customer, if you will, We’re now also recognizing that the features that we implemented in that machine are applicable to other customers. So at the time, if you said, okay. Nine months ago, it’s like, I think this is gonna work, and the customer was cheering us on. Then it worked, and then the customer, you know, two quarters ago gave us an order, and we shipped it.

You know? Weeks later, felt like. Okay? To now, we have inbounds from customers to we’re now evaluating on paper the qualification of those devices and recognizing we can test your parts and this is why. Right? So you’re starting to see in the eyes of the looking back at you the value that you have in doing this, and the sincere interest in trying to make this happen. They’re like, how can you make this happen? If it’s not coming across clear enough, and I know as a CEO, you have to be careful of you know, I’m always a cheerleader, but this is very real in their eyes. And we used to talk about, you know, stacked memories and how important it would be to burn them in before you stack them. Okay? You’re taking devices that are maybe $2 and you’re putting them eight high to make a 16 or $20 part.

Okay? Then we talk about silicon carbide devices that were going into these modules for EVs. We were taking $10 parts putting 10 of them into a package, and selling them for $200. And look at the value added by doing that. Now we’re talking about HBM stacks where the stack of memory is hundreds of dollars. Times eight stacks, processors that are a thousand dollars of copy out of a TSMC fab They’re all stacked together, and they all need production burn-in. The math is so obvious about the value of being able to move this to wafer level that it’s just more about what can you do to prove it. And we have the ability now to not only physically show it, we can bring customers in We have tools in house that are configured for AI, high power, 300 millimeter wafers with our aligners They can see wafer packs.

We’ve upgraded the facility enough power to be building you know, 10 to 20 systems at a time if needed That’s not to imply that that’s how big our forecast is. But we have that capacity they come in and look and go, okay. I get this. So the other thing and I made some comment, kind of a snide comment, you know, that the customer enthusiasm exceeds our ability to sell and market, and that’s not a dig on my sales and marketing team. What we believe has happened is that these customers are being walked by the tool. And tools, I should say, because it’s tools actually, and seeing the feasibility of it. And so the biggest OSAT is marketing this capability and we’re getting these people calling in and say, woah. Woah. Hold on a second. I mean, I guess I got the press release.

But I just saw it or understand it. What do I need to do to do that? So you can imagine you’d be pretty excited about that.

Jed Dorsheimer: Yeah. No. Certainly. That’s helpful. Last question for you. Along those same lines, you mentioned the OSAT that’s starting to market. So you’ve got an OSAT in your silicon photonics. An OSAT or two that you’re working with in the AI processor And mostly customers that own the design for the AI processors What you haven’t mentioned you did mention TSMC but I’m just curious from a foundry perspective, what’s the real what where are the discussions with foundries in the process? Are you seeing some cross-pollinization like you are with OSATs?

Gayn Erickson: For sure. Yeah. So I’m gonna use the I’m gonna use a little different vocabulary. In this world, there’s only really two foundries. Well, to be respectful for Intel, maybe three that can build them. Okay? So GlobalFoundries isn’t really building AI processors. Because of the nodes they’re on. So you’re really talking about TSMC and potentially Intel? Okay? Would TSMC be having right now, I think everybody but Intel? Okay? And then you then you talk about design houses. Design houses are the companies that are actually working with the hyperscalers in designing their new ASICs or AI process So you’re talking about you know, Marvell, Broadcom, all chip, as examples. Okay? Those are the big guys that are actually doing the new designs.

Okay? Then, of course, some of the big hyperscalers have in-house design. They then work with the models, say, for example, with TSMC that will provide them with the models based upon what particular node they’re at. That will have DFT designed for testability features and functionality inside of it that interestingly make them common, And then they can outsource that to really two or three of the big OSATs, the biggest one being ASC who also bought Spiel, but people think those are two different companies, but candidly, they’re really the same. And then you have Amcor, You have Japan that is trying to get some of their business JCCT, and then they’re not an OSAT. They’re just test, and that’s KYEC. So it’s a pretty small community. And a lot of cross-pollinization of ideas, I would say, and awareness.

And so once the system got pulled put into production, the lights started to go on and the phone call the phone started to ring with people saying, hey. Wait a minute. What do you I thought that was just a marketing thing. This is real. Can you do my part too?

Jed Dorsheimer: Thank you.

Gayn Erickson: I hope that helps.

Operator: The next question comes from Igor Domachev with Freedom Broker. Please proceed.

Igor Domachev: Hello, gents. Thank you for Exhaustive Market comments. I wanted to get a glimpse into your outlook for 2027. Especially for TSMC. Matter. Recently, TSMC reported that there were reports that TSMC plans to wind down gallium nitride foundry services by 2027. Could you share your thoughts on how this shift in market can impact your business and addressable markets?

Gayn Erickson: Okay. So, I mean, we chose not to give guidance on ’26. I’ll be struggling with ’27. But, you know, we do believe we can give back to the track of growing significantly over this period of time. The size of the market, for example, between AI, silicon photonics, for optical chip-to-chip communication, and you add in silicon you know, even silicon carbide are all kicking in pretty hard in ’27. So, you know, we’re planning for you know, pretty significant growth. I’ll leave it at that. Okay? Now you specifically talked about GaN and TSMC. TSMC, mostly I wake up every day thinking about all AI wafers that TSMC is doing. GaN, you know, gallium nitride can be put onto multiple different process substrates. The most, kind of interesting one would be on silicon, large silicon substrates.

But a lot of the they’re both IBM folks around the world, like Infineon, for example, or people that have dedicated foundries And then TSMC is also doing a foundry. I don’t know what TSMC’s market share is, but I candidly, when I talk to customers, I hear a lot more about the other foundries or their own IDM sources than TSMC. So I don’t believe that it has a negative impact on us if they start to wind down that business. I think it that just shows up somewhere else. Either at one of the IDMs or one of the other foundries.

Igor Domachev: Okay. Great. Thanks. And in terms of your new AI clients, how long will it take to pass all qualifications to reach your final decisions? For your client after you get the best client delivered.

Gayn Erickson: So we already have the first client. We announced that we were I think this year exactly, we announced that we had a commitment from a first customer to evaluate our solution for AI. We said we’re very excited about it, but there were still a lot of uncertainties. The customer seemed extremely willing and pulling us in. I think I used the term they’re more excited than we are and more hopeful about getting the business which, of course, was good. Then within a quarter, we said it was progressing pretty well. The following quarter, we said we are now testing wafers and the data looked good. And within that quarter, they placed an order, and we shifted it same time. That was the first customer which there’s some argument is takes longer than the second one.

I will tell you our enthusiasm and confidence is a lot more higher because of actual evidence. But there’s still some variables and things that we still need to worry. My attorneys will always tell you careful. There’s always risk, and you know, you never know. Till the order’s in hand. But, you know, we think that a decision could be made you know, within, say, six months or so. And, you know, potentially orders placed somewhere thereafter.

Igor Domachev: Yeah. I got it. Thank you. It was useful. Next question. You. Thank you.

Operator: Next question comes from Larry Chlebina with Chlebina Capital. Please proceed.

Larry Chlebina: Hi, Gayn. That first AI customer at the OSAT, so there are are you under the belief that they’re really pleased with it? And do you expect more orders from them in the near future?

Gayn Erickson: Yes. To both of those. Wait a minute. You said near future. I wanna be careful of setting any timelines, but I’ll go out and say we expect more just more this year, though.

Larry Chlebina: And then now you have another AI customer in evaluation. So that’s the second one for wafer level burn-in. And then you have a third one that’s going after the production in the package part burn-in. Are they three distinct AI customers? Or are they Yes.

Gayn Erickson: Yes. Are they so different. Totally different.

Larry Chlebina: And in the write-up, you said that, you shipped an 18 wafer high power silicon carbide system in the was that did that ship and book in the make order?

Gayn Erickson: Yes. Yes. And it was and it was an upgrade to one of the systems that they had purchased earlier. Part of our strategy and one of our commitment with customers is this commitment to a platform that allows people to maintain forward backwards compatibility. It was a really big deal. They shipped the system here. We reworked the system and then shipped it back to them. Capable of testing 18 wafers at a time at high voltage, which is amazing.

Larry Chlebina: And that also included the automated aligner?

Gayn Erickson: Actually, in this case, it didn’t it did not. It shipped back without the automated aligner.

Larry Chlebina: Do you anticipate upgrade on the today. They by the way, they have automated they have automated aligners, but we only upgraded the system basically.

Larry Chlebina: I see. Do you anticipate, going forward at all the, silicon carbide systems will be high power?

Gayn Erickson: Actually, we refer to them as high voltage but I do not I know what you mean. Mostly. Yes. We’ve been working really closely with the OEMs. And making them aware. The like, the car suppliers, the EV guys. To making them aware of the capability to be able to continue to provide lower and lower cost solutions that include the high voltage insertion as well if they wanna do it. I would say that it’s mostly likely that people will buy it with that. And it allows you to do it with, you know, two multiple different stress conditions. But this allows you to also do the stress condition with high voltage, which is very valuable.

Larry Chlebina: Does that include a premium? Does it do you get a premium price on that? Going forward? Or not? Yeah. The option does cost additional dollars. That’s correct. Okay. I gotta get my memory question. And you talked up how it makes so much sense for HBM to go to wafer level burn-in. When are you gonna get a valuation going for an HBM application?

Gayn Erickson: That’s a good question. We’ve talked about walking before we run, and the walking to begin with was the wafer level burn-in application for the NAND flash. But one of the critical things that we mentioned last time, but I’ll mention again today, One of the most specific things we were implementing was a new fine pitch MEMS WaferPak that allow us to get to very attractive price points at extremely high pin counts. Okay? But it also allows us to go to, you know, four times or one fourth the pitch. These are distances, if you will, than what we have today. That fine pitch is technically fully capable of doing DRAM as well. We knew that on purpose. So one of the critical things is there was no way to contact a Deepgram HPM device. With our previous wafer packs, but we now enable it through this wafer pack. Right. So now There’s also things on the AI road map that are reusing the technology developed under this as well.

Larry Chlebina: So now that you have the MEMS fine pitch wafer pack, and it’s proven out on flash. Is that correct? You it seems like Speed work through this. This is we’re trying to get it finalized throughout this quarter here.

Larry Chlebina: Okay. So it seems like that would entice one of the HBM guys, the say, boy, this could solve a big problem that we have as we go from eight eight stacks to 12 possibly eventually up to 24.

Gayn Erickson: Yeah. That’s I think I think you’re right on that.

Larry Chlebina: Alright. Well, we’ll be waiting to hear when you do that. That’s all the questions I have.

Gayn Erickson: Thanks, Larry. Thank you.

Operator: Take care.

Operator: Okay. We have no further questions in the queue. I’d like to turn the floor back to management for any closing remarks.

Gayn Erickson: Alright. Well, I thank everybody for their patience, and I fully recognize that our prepared comments were at least ten minutes longer. We added about a fifteen-minute chunk in the middle to do detail and get on record. The information around the AI that we won’t need to do next time as well. And, hopefully, we can just give you an update on the successes that we’re having there. So I appreciate everyone’s patience and listening through this. If you have any follow-on questions, you can reach out to us. We’ll be happy to take the call. Thank you very much.

Operator: Thank you.

Operator: This concludes today’s conference, and you may disconnect your lines at this time. Thank you for your participation.