There's a small company in Fremont, California, that almost no one outside the semiconductor industry has heard of.
It has fewer than 200 employees.
Its annual revenue is smaller than what some hedge fund managers make in a year. Its CEO is a soft-spoken engineer who's been there for over a decade.
And yet, three of the most consequential technology trends of the next ten years — Tesla's electric vehicles, the AI data center buildout, and the rise of silicon photonics — all run through machines this company builds.
The company is called Aehr Test Systems.
The ticker is AEHR.
And if you understand what they do, you understand something important about how modern technology actually gets manufactured.
What They Actually Do
Aehr makes burn-in equipment. That's it. That's the whole business.
Burn-in is a process most people have never heard of, but it's the reason your phone doesn't randomly die in the first month, the reason a Tesla's power inverter doesn't fail on the highway, and the reason an AI training chip costing tens of thousands of dollars doesn't crash a data center the day it's installed.
Before any high-stakes semiconductor goes into a product, it gets stress-tested under elevated voltage, elevated temperature, and elevated current — for hours, sometimes days — to weed out the small percentage of chips that have hidden defects and would have failed in the field.
The defective chips fail during burn-in, in a controlled environment, instead of in your car. The good chips get shipped. The whole industry depends on this happening reliably, at scale, for every chip that matters.
Aehr's specialty is doing this at the wafer level — meaning before the individual chips are even cut apart from the silicon disc they're manufactured on. Their FOX-XP system can test 18 wafers in parallel, with each wafer holding hundreds of chips.
The economic advantage is enormous.
Wafer-level burn-in costs roughly $250,000 per wafer of capacity using Aehr's equipment, versus $700,000 to $1 million per wafer using competing approaches.
That's a 60%+ discount on a piece of equipment that semiconductor manufacturers must use, not optionally buy.
For years, this was a small, unglamorous business with declining revenue. Aehr had spent decades building expertise in a niche corner of semiconductor capital equipment, and the niche stayed niche. Then, almost simultaneously, three things happened.
The Three Trends That Changed Everything
The first trend was silicon carbide.
Traditional semiconductor chips are made from silicon.
Silicon carbide is a different, harder material that handles much higher voltages and temperatures, which makes it the ideal material for the power inverters that convert battery DC into the AC current that spins an electric vehicle's motors.
Tesla pioneered this in the Model 3, and the rest of the EV industry has been racing to catch up. Silicon carbide chips are critical for EV range, charging speed, and efficiency. They are also notoriously difficult to manufacture reliably, which means they require extensive burn-in testing — at the wafer level, exactly what Aehr specializes in.
By 2022, Aehr had landed onsemi (formerly ON Semiconductor) as a major customer, then added STMicroelectronics, then Wolfspeed. The legacy oilfield-services-like decline turned into a growth story almost overnight.
The second trend was AI processors.
The chips that train large language models — the H100s, the GB200s, the custom ASICs that companies like Google and Amazon are designing in-house — are some of the most expensive and complex semiconductors ever made.
A single Nvidia GB200 NVL72 rack costs roughly $3 million.
When you're paying that kind of money for a chip, you absolutely cannot afford to discover a manufacturing defect after it's been installed in a data center running a critical AI workload. So hyperscalers and AI chip designers have begun demanding wafer-level burn-in for their highest-end processors, where historically only packaged-part testing was done.
In late 2025 and into 2026, Aehr started landing AI processor orders.
First, a $14 million order from a lead AI processor customer in February.
Then, a record $41 million follow-on order from a hyperscale AI customer for their Sonoma package-level burn-in systems, used to test custom AI ASICs for training and inference workloads.
By the third fiscal quarter of 2026, Aehr was reporting record bookings of $37 million and a book-to-bill ratio over 3.5x — meaning new orders were coming in three and a half times faster than they could ship product.
The third trend was silicon photonics.
As AI clusters get bigger, the bottleneck stops being the chips themselves and becomes the data movement between them. A modern AI training run involves hundreds of thousands of GPUs talking to each other constantly, and the copper wires that traditionally carried that data are running into thermal and bandwidth limits.
The replacement is silicon photonics — chips that send data using light instead of electricity. Companies like Coherent and Lumentum make these, and the market is exploding.
Silicon photonics chips are tested using burn-in too.
And in March 2026, Aehr announced its first major silicon photonics customer win — a global networking leader and major supplier of optical transceivers for AI data centers.
The stock jumped 23% in a single day. By April, follow-on orders had pushed second-half bookings above $92 million.
The Company’s Numbers Here's where the company stands, as of the most recent reporting.
Quarterly bookings of $37.2 million. Effective backlog of $50.9 million. Cash position of $37.1 million. The book-to-bill ratio is the most telling number — at 3.5x, it means orders are accumulating faster than the company can deliver, which historically is a leading indicator of explosive revenue growth one to two quarters out.
The stock reflects this.
AEHR started 2026 at around $20.
By late March, it had reached $30.
By early April, it had eclipsed $65 after the silicon photonics announcement and the record AI processor bookings. The trailing twelve months have been one of the strongest performances in the small-cap semiconductor equipment space.
What you're really buying with AEHR is a small, focused company that found itself sitting at the intersection of three of the most important secular trends in semiconductors, with a structural cost advantage that competitors haven't been able to match.
The business model has operating leverage built in — most of the cost is in designing and building the systems, and once they're shipped, the consumables (the wafer contactors, the test cards) generate recurring revenue at high margins.
So, if you're going to invest in the AI buildout, the EV transition, or the silicon photonics shift — and most thoughtful investors are doing some version of all three — it's worth knowing about the small company in Fremont that quietly tests the chips that make all of it work.