Description

AEHR Technologies is a supplier of semiconductor burn-in test capital equipment. AEHR sells tools primarily into the high growth silicon carbide (SiC) power device market. The stock has been a huge winner over the last two years, as revenue has skyrocketed four-fold, and investors have slapped a generous multiple on the company as a result. AEHR is now trading at 12x Sales and 50x non-GAAP (untaxed) EPS on FY23 (May) numbers. The problem I see for AEHR is that this strong growth has been driven by one customer, and that customer, I believe, should start declining next year. This customer, OnSemi (ON), was 80% of revenue on a TTM basis and is likely to start reducing purchases beginning in CY24, based on ON’s public disclosures regarding their SiC capacity plans. Even if we assume substantial growth from new customers, AEHR will be unable to hit the street’s aggressive FY24-FY25 revenue estimates, due to this decline from ON. The stock is up 45% YTD, ~200% in the last year, and over 1000% over the last two years on the back of their growth with ON; I expect there to be substantial downside once ON starts to roll over. I see around 50% downside to fair value around $15/share, which would represent a roughly peer average 4x Sales on my FY25 estimates, and 20x GAAP EPS. 

Background

The market for silicon carbide (SiC) power devices (chips) was estimated to be around $2B in CY22, with electric vehicles representing the bulk of that demand. TSLA has led the adoption of SiC in EVs, having first started using SiC devices from STMicro (STM) in the Model 3 in 2018. Since then, SiC has become widely adopted by multiple auto OEMs for use in their EVs. SiC chips are a good solution for EVs, as they allow EVs to be more power efficient, and thus extend battery range. Estimates have the market for SiC devices growing several-fold over the next few years, from around ~$2B in CY22 to $7B-$11B in the 2026-2027 timeframe. 

There are five major SiC device manufacturers, with STM currently being the largest. Most have put out multi-year revenue targets for $1B+ in annual SiC in revenue. Below are actual/estimated/projected SiC revenues by manufacturer pulled from company disclosures and sell-side estimates: 

While SiC chips allow for more efficient power usage and extended range in EVs, the chips are both more expensive to produce than traditional silicon chips, and also have higher material defect rates/failures, particularly when exposed to high temperatures. SiC failures follow the so-called ‘bathtub curve’ whereby the failure rate is high early on (“infant mortality”), then decreases exponentially to reach a steady-state of low failure, before eventually rising again around the end of the chip’s life. Because of the high infant mortality, and the need for chips in EVs to be able to withstand extreme temperatures without failing, SiC chips need to undergo high temperature burn-in testing to catch these early failures. EVs OEMs require that the SiC devices they purchase have already gone through high temperature burn-in so as to weed out defective chips. 

This high temperature burn-in step can either be done on the finished packaged device, or it can be done earlier, at wafer level, before the individual die have been separated and packaged. AEHR’s FOX-XP tools do this burn-in testing step at the wafer level, and allow for up to 18 wafers to be burned in at the same time. The primary benefit of doing burn-in at wafer level is catching defective die before they are packaged, and thus saving costs by not packaging bad die. SiC chip suppliers to date have been mixed on the need for adopting wafer level burn-in. ON has been to the only one that’s gone all in on wafer level test, while the largest player STM has not adopted it all. STM, and others, instead just do burn-in testing on the finished packaged devices, before those devices go to their EV customers. 

Below I’ll go into more detail on AEHR’s various SiC customers (or potential customers), but overall, of the 5 major SiC device manufactures only ON (#5 player by 2022 revenue) has widely adopted AEHR’s wafer level burn-in. STM (#1 player) has not adopted, nor do they apparently have plans to. 

AEHR has announced early engagements with two other SiC suppliers in recent months, one for initial production tool orders, and one who has ‘selected’ AEHR. 

While I believe that broader adoption of AEHR’s tools more widely beyond ON is possible, or even likely, I don’t see this adoption being anywhere near enough for AEHR to hit FY24/FY25 consensus expectations.

For additional background on AEHR’s technology, SemiAnalysis had a good write-up:

https://www.semianalysis.com/p/aehr-multi-wafer-level-burn-in-test?utm_source=%2Fsearch%2FAEHR&utm_medium=reader2

On Semiconductor (ON)

AEHR shipped its first tool to ON for SiC EV applications in CY19. Then, in CY21, orders and shipments to ON really took off as ON began to ramp its SiC capacity build in earnest. ON did $200M+ in SiC revenues in CY22 and expects to do $1B in CY23. As of last month’s earnings call, ON had $4.5B in committed Long Term Supply Agreement (LTSA) revenue over the CY23-CY25 period. Given an approximate $1B/$1.5B/$2.0B cadence for the ON LTSAs in CY23/CY24/CY25, we’d expect that ON likely had SiC revenue capacity approaching $1B entering CY23, and should be approaching $1.5B of capacity exiting CY23. ON would be targeting having something over $2B in revenue capacity for CY25 to meet their LTSAs.

To date, AEHR has shipped $75M-$80M in FOX-XP tools and related consumables (Waferpaks) to ON. On a TTM basis, ON was 80% of AEHR’s revenue. At an ASP of $2.5M for the FOX-XP tool + $1.5M in Waferpaks, that translates to about 20 FOX-XP tools to ON to date. With the most recently announced $25M FOX-XP tool order in January 2023 from ON (expected to be accompanied by $15M in Waferpaks) to be delivered over the March 1, 2023-November 30, 2023 three quarter period, AEHR will likely have shipped about 30 FOX-XP tools to ON by the end of November. ON should be approaching $1.5B of SiC revenue capacity exiting CY23 with these tools.

This should all be fine for AEHR because ON will not stop ramping their SiC revenue capacity at $1.5B, or even $2B. The problem for AEHR comes in how ON plans to further ramp their SiC revenue capacity in CY24 and beyond. Here’s what ON said on their Q322 call in November ‘22:

“To limit our long-term CapEx investments, most of the silicon carbide equipment we are installing around the world is 200-millimeter capable, and we are on track for 200-millimeter wafer qualification in 2024 and related revenue ramp in '25.”

Here is ON’s slide on their SiC capacity expansion plans, with my annotation in red:

From ON’s slide above, we can see that ON expects new wafer start additions in CY25 to be less than half of what they were in CY23/CY24. The lower number of wafer start additions in CY25 would imply meaningfully lower equipment purchases by ON beginning in CY24. As a result, I believe AEHR’s ON business should see a decline from the current ~$50M annual run-rate to perhaps half that amount, with that drop-off likely starting to occur in early to mid CY24. 

This drop will coincide with ON’s transition to 200mm. ON’s SiC production is currently being done on 150mm (6 inch) SiC wafers. But most of the equipment ON is installing, including the AEHR tools, is capable of running 200mm (8 inch) wafers also, and ON plans to transition those tools over from 150mm to 200mm. Beginning no later than CY24, this transition to 200mm should increase’s ON’s SiC revenue capacity by roughly 80%-90% on the same set of tools. 

In terms of revenue capacity, if ON were to exit CY23 with $1.5B in SiC revenue capacity when running 150mm wafers, that should translate into $2.5B+ of revenue capacity once they’re running 200mm wafers. This transition to 200mm, and the associated added revenue capacity, is the reason that ON’s purchases of additional tools from AEHR will slow beginning in CY24. I’d note that ON has said recently that they see a $6.5B SiC TAM in 2026, so it’s not clear how much further beyond $2.5B+ in revenue capacity ON would look to expand in the next couple years; ON would likely be happy if they were one of the SiC market leaders, and owned something like 1/3 of this market, in a few years.

Infineon

AEHR’s second major SiC customer, who I believe to be Infineon, has ordered 2 FOX-XP production tools so far. These orders were each announced by AEHR within the last four months, and both tools are scheduled to be delivered in AEHR’s current May ’23 (Q4) quarter. 

Infineon’s 2 ordered FOX-XP tools is a modest number relative to the 30 or so that ON will have taken by the end of this year. At this point, it’s hard to know whether this will turn into an ON-sized adoption, or whether Infineon is using AEHR for a more limited application. What I would point out here is that Infineon was already ~50% larger than ON in terms of CY22 SiC revenue, without yet using AEHR’s tools. In fact, Infineon was one of 4 SiC suppliers that did more SiC revenue than ON in CY22, each without the benefit of using AEHR tools. While it’s possible that Infineon goes all-in on AEHR like ON has, it is demonstrably true that SiC chips can be produced, tested, and sold in high volumes by multiple non-ON suppliers without using AEHR tools. 

ST Micro (STM)

STM has been the early leader in the SiC device market, beginning with the adoption of STM’s SiC MOSFETS in the Tesla Model 3 in 2018. In CY22, STM did approximately $700M in SiC revenue, making them twice as large as any other supplier. And they’ve done this without AEHR tools, which again indicates that being successful as a SiC supplier isn’t dependent on using AEHR. STM’s view on AEHR/wafer level burn-in seems to be that it’s unnecessary/redundant. Because they will always do burn-in on the final packaged device, they don’t believe that also doing burn-in at the wafer level is necessary. This view was conveyed in a Tegus call with a former technology executive at STM:

Other Customers

• AEHR announced in late January that another (3rd) SiC customer “selected” AEHR’s FOX tool, but it’s not clear that there were any associated tool orders yet.

https://www.aehr.com/2023/01/25/silicon-carbide-semiconductor-supplier-selects-aehr-fox-system-for-wafer-level-test-and-burn-in-of-silicon-carbide-devices-for-automotive-electric-vehicles/

• AEHR has had a $5M-$10M annual revenue business with INTC for silicon photonics for a number of years. This business has bounced around, but remained below $10M/year. I model some growth from INTC and other non-SiC customers, but I don’t believe INTC is particularly material to the story here.

Estimates

ON will be an estimated $40M-$50M out of AEHR’s $66M in estimated revenue in FY23 (May), with perhaps another $8M from their second major SiC customer (Infineon). Consensus estimates for $102M in FY24 and $168M in FY25 appear to embed expectations that AEHR will not only maintain/grow that level of business with ON, but also ramp multiple other SiC customer to be similarly ON-sized by FY25 (May). 

With no SiC customer besides ON being more than perhaps ~$8M in FY23, the idea of ramping multiple new customers to ON’s current $40M-$50M levels would seem to be a best-case scenario for AEHR, and one that I don’t believe will happen. Other than the 2 FOX-XP tools that will be delivered in the AEHR’s current fiscal Q4 to Infineon, we have little tangible evidence of major adoption by SiC players besides ON. 

Given that each of the other major SiC suppliers have already achieved significant production scale without using AEHR tools, I believe that they will likely use AEHR in a more limited fashion, if at all, and that these other customers likely won’t get to be the size of ON. 

I still do give AEHR significant benefit of doubt here in modeling non-ON SiC revenues going from sub-$10M in FY23 to $30M in FY24, and $50M+ in FY25. I believe this represents an optimistic scenario of multiple new SiC customer ramping for AEHR, though no single customer reaching ON’s scale. 

AEHR currently has about $22M in deferred tax assets, with a full valuation allowance against them, so aren’t recognizing income tax expense. They should start recognizing tax expense in FY25, which is how it’s modeled above.

Other points

• As the stock has run up, insiders have sold $24M in stock YTD in CY23, reducing their combined ownership stake from 8.9% to 6.5% of S/O

• Company announced they were doing a $25M ATM offering in early February, despite not needing the cash

• TSLA recently made headlines at their analyst day by claiming their will reduce SiC usage by 75% at some point in the future

  • It’s hard to know how much weight to put on this at this point, and SiC carbide bulls seem to have largely shrugged this off, but it could potentially pose a risk to those lofty projections for the SiC TAM in the out years. 

• AEHR’s TTM GAAP R&D spend was just $6M, which is incredibly low 

  • Despite its small scale, AEHR actually spent less on R&D as a percentage of TTM revenue (10%) than industry bellwether AMAT (11%) or large cap test equipment supplier TER (14%). 

  • I’m skeptical that this low level of R&D can maintain any kind of durable competitive advantage

Conclusion

AEHR is extremely expensive at roughly 12x Sales and 50X non-GAAP untaxed EPS on FY23 numbers. Investors seem to be betting that the ~60% FY24-FY25 revenue CAGR that’s in consensus estimates is achievable. I believe AEHR will fall well short of those numbers, as they are unable to overcome revenue declines from their 80% customer ON, even with a significant ramp from other SiC customers. 

I see 50% downside in the stock to $15, which would represent a roughly peer average 4x Sales multiple, and 20x taxed EPS on my FY25 estimates. 

Risks

• Could be early

  • AEHR can probably meet/beat estimates for the next few quarters, with risk to numbers more so starting in early CY24

  • I expect an imminent ON order announcement of ~$15M for Waferpaks associated with the January FOX-XP order announcement that could temporarily boost the stock once announced

    • With the stock up almost 50% YTD and trading at over 12x Sales I still think this represents a good entry point

• All major SiC players could fully adopt and AEHR achieves ~$200M of annual revenues

  • ON will have gotten to something north of $2B in (200mm) revenue capacity with $120M of spend on AEHR tools over a 2-3 year period, or $40M-$50M/year

  • So, I’d estimate that in the unlikely scenario that AEHR gets full adoption across all SiC suppliers they could theoretically get to something on the order of ~$200M in annual revenues while SiC industry capacity is getting aggressively built out. 

    • Not very realistic when the largest player (STM) doesn’t use AEHR or see much value inn the wafer level burn-in technology

Catalyst

- FY24-FY25 numbers coming down