For those wishing to understand the tech, here are some research notes I’ve put together:
Silicon Carbide chips. Understand that this is a specific “chip” application that is not associated with computing, but with power delivery, such as in electric vehicles. Here’s a 2-year old article on Tesla’s pioneering use of SiC chips for the inverters it uses in its cars:
Why is Tesla’s inverter innovative? Well, with the release of the Model 3 in 2018, Tesla became the first company to add SiC metal-oxide-semiconductor field-effect transistors (MOSFETs), sourced from ST Microelectronics, in an in-house inverter design. The overall design has several innovations beyond the use of SiC packages, but this is the main one. It has led to the overall weight of the inverter (4.8kg) to be under half that of the 2019 Leaf (11.15kg) and over a third less than that of the Jaguar I-PACE (8.23kg), which use Si IGBT inverters and off-the-shelf parts.
The result is Tesla’s overall inverter and permanent magnet motor combination is one of (if not the) best on the market, achieving a 97% efficiency, yielding more range without increasing expensive battery capacity. All at a similar cost to the older technologies being displaced.
Since then, other EV companies are likewise using SiC for their inverters.
Another good read is this recommendation for investing in AEHR about a year ago(!): Silicon Carbide Gold Rush - Aehr Test Systems
The 97-99% yield on manufacturing SiC at the wafer (or “substrate”) level is nowhere close to the yield of 99.99%+ for memory chips.
To the uninitiated, a failure rate of 1% on a single die may not sound bad. A standard traction inverter has 48 SiC devices on one computer board or “package.” For the traction inverter to properly function, not one of those dies can go bad. Additionally, the failure of the die can happen 6 months down the road because of the tremendous stresses of operating in high temps & voltages. This failure means the car stops working (perhaps in a dangerous situation) and expensive repairs.
So, this is why the burn-in testing of SiC is so important.
Now, I’ve seen a couple of sites claim that Tesla is using SiC chips from STMicroelectronics, and then that STMicroelectronics is a customer of AEHR. What I don’t know is whether STMicro is the exclusive supplier of SiC chips to Tesla or whether AEHR is the exclusive tester of SiC chips by STMicro.
But, it seems fair to say that with many EV OEMs and Tier1s going to SiC for their inverters, that AEHR is in a fast growing market. And if not exclusive to Tesla, just the growth in other EVs should lead to more demand for AEHR’s testing systems.
Note, there was some concern about a recent Tesla presentation that claimed their next-gen vehicle would reduce SiC use by 75%. Here’s a Twitter thread on that:
AEHR’s stock (and STMicro and other SiC vendors stock) dropped on this presentation, but recovered. It appears from that Twitter thread and this article that this is a natural evolution of chips getting better and more capable. My take is that the number of EVs with SiC inverters is growing faster than optimization is reducing SiC use per inverter, but that’s just an uneducated guess.
The Gallium Nitrite (GAN) new business is, I think, something that is mostly future oriented right now. This very technical article concludes:
Silicon carbide may be a more effective product in the short term, as it is easier to manufacture larger, more uniform wafers of SiC than gallium nitride. Over time, given its higher electron mobility, gallium nitride will find its place in small, high-frequency products. Silicon carbide will be preferable in larger power products, given its power capabilities and higher thermal conductivity than gallium nitride.
So, GAN might not be for future EVs, but for other power products, and I don’t understand the market potential for GAN at all.
Similarly, for Silicon Photonics, I don’t understand the need and ROI for AEHR testing. I do understand EV inverters, and between the relatively high failure rate of SiC chips during use (in other words, they work fine at first, but the high power and temps lead to some failures in the first 6-12 months) and the cost of replacing inverters (these aren’t chips that can be easily swapped in and out), and the issue of potentially having EVs die on the road, something like AEHR sounds very important to me. I’m sure companies using Silicon Photonics want the chips to not fail, but there has to be an ROI on the testing costs versus addressing failures in the field for it to gain traction. I just don’t understand enough of that, yet.