Originally published on LinkedIn, August 11, 2026.
Silicon carbide is rarely found in nature. To get much of it, somebody has to grow it.
This takes weeks.
The crystal is grown in a furnace until you have an extremely expensive cylinder about eight inches across called a “boule.” The boule is sliced into sub-millimeter-thick wafers, which are then polished to an absurd level of precision. The wafers are used in applications where losing electricity to heat is expensive, such as electric vehicles and data centers.
One of the largest and most sophisticated factories in the world that makes these wafers sits in the middle of chicken country about 60 miles west of Raleigh. Siler City has produced cotton, timber and poultry for decades.
Then somebody put a 445-acre, two-million-square-foot silicon-carbide factory in it.
It didn’t work.
Siler City has a population of 8,000 and a median household income around $46K, so you can understand why they wanted 1800 jobs at $78,000 a year.
Wolfspeed initially hired around 200 people. But then it cut 73 positions.
Then it filed for Chapter 11 bankruptcy in June 2025.
Then, the IRS sent it $700 million in cash two months later.
How does this happen?
The annoying thing about this story is that every simple explanation turns out to be wrong.
Wolfspeed started in Durham in 1987 as Cree Research. The company was initially known for LEDs. John Palmour, a founder, helped figure out how to make silicon carbide commercially. Eventually, it sold its lighting businesses and narrowed its focus to silicon carbide.
Ordinary silicon is cheap and works great for most electronics. Silicon carbide works better when things get hot or when a lot of electrical current needs to move through a small space.
An electric vehicle is one example. A battery’s energy is lost as heat as current moves. Silicon-carbide devices reduce the loss. Wolfspeed believed the world would need much more silicon carbide. It was right about that part.
The question was how much, how soon, and who would make them.
Wolfspeed’s answer was essentially: a lot, right now, and we will.
The company built the factory in Siler City to sell wafers. About 700 miles away, in upstate New York, it had built another huge factory in Mohawk Valley. Siler City would grow the crystals and make wafers. Mohawk Valley would turn some of those wafers into power devices.
There are good reasons to do this.
If you know how to grow unusually good crystals, there is no particular reason to hand all of them to somebody else and let that company make the higher-value product. Tiny defects in the crystal can also ruin the device later. Controlling both processes gives Wolfspeed more control over the result.
It was also moving from six-inch wafers to eight-inch wafers. The larger wafer has about 78 percent more surface area. If your yields hold, you get a lot more devices out of one wafer.
Everything gets better with scale. Assuming you get to scale.
By the end of 2024, Wolfspeed had spent about $1.9 billion at Siler City. It expected to reach roughly $2.3 billion through fiscal 2025. The longer-term commitment was up to $5 billion.
This is where I first thought I understood the problem. It seemed they built all of this without enough customers.
They didn’t.
Renesas signed a ten-year wafer agreement and eventually handed Wolfspeed $2.062 billion. BorgWarner put up another $500 million. Infineon reserved capacity. And there were others.
A customer giving you $2 billion makes you stop asking whether there are customers. But apparently, there are quite a few ways to say you plan to buy something without actually promising to buy much.
In Wolfspeed’s filings, they talk about purchase orders, design-ins and design-wins.
But many purchase orders carried no long-term commitment. A development agreement could back a design-in and still never produce meaningful revenue. Even a design win was not a guarantee that all the expected sales would materialize.
Perhaps lawyers protecting themselves from imaginary future problems isn't that bad when the imaginary problems happen.
Renesas abandoned its own silicon-carbide program. Turns out, the $2 billion deposit was refundable and unsecured. What had looked like customer demand became a loan. Renesas eventually took a large loss and emerged from Wolfspeed’s bankruptcy owning debt and equity in the company.
In other words: a customer gives its supplier $2 billion to build capacity for something the customer expects to buy. The customer changes its mind. The factory is already built. The supplier goes bankrupt. The customer ends up owning part of the supplier.
The silicon-carbide wafer story gets stranger the further you follow it.
A semiconductor company builds devices on the wafer. Those devices may end up in an inverter for a vehicle that won’t be produced for another three years. So a factory in North Carolina can depend on an engineering decision somebody somewhere else hasn’t made yet.
Meanwhile, it takes years to build the factory.
Nobody really knows what the market will look like by the time it opens. But somebody still has to pour the concrete.
For a while, the timing looked great. Electric vehicles were growing quickly, and silicon carbide was scarce.
Then everybody noticed silicon carbide was scarce.
Chinese manufacturers built more capacity. Prices fell. Much of the fastest EV growth happened in China, where automakers increasingly bought from Chinese suppliers.
And customers changed the product. In 2023, Tesla said its next-generation drive unit would use about 75 percent less silicon carbide. They had figured out how to build one using less of the material Wolfspeed was spending billions to produce.
By 2024, Wolfspeed was still the largest silicon-carbide substrate producer in the world. But industry revenue for N-type substrates fell 9 percent.
The technology was good, but Wolfspeed was in trouble.
This is where the two-factory strategy unraveled. Siler City needed people to buy wafers. Mohawk Valley needed people to buy Wolfspeed’s devices. Some of the material from Siler City could go to outside customers. Some would feed Wolfspeed’s own plant.
When demand is running ahead of supply, this looks brilliant. When demand slows, you have two very expensive factories waiting on each other.
If outside customers buy fewer wafers, Siler City loses an outlet. If Wolfspeed sells fewer devices, Mohawk Valley needs fewer wafers. If there's too much capacity, wafer prices fall.
Meanwhile, the plants remain indifferent to this.
There is no button on a two-million-square-foot factory marked Tesla changed its drivetrain.
I don’t know that vertical integration was a mistake. It may have been the right way to get eight-inch silicon carbide to scale.
But the forecast was incorrect. And perhaps the financial entities were not properly decoupled.
Wolfspeed needed factories to ramp, yields to improve, and customers to commit. The market needed to use roughly the amount of silicon carbide everybody expected. And prices had to behave.
This is a lot to ask of the future.
By 2025, Wolfspeed had more than $6 billion of long-term debt. The new factories were underused, and prices had fallen. In one quarter, Wolfspeed reported $47 million in underutilization costs across Siler City and Mohawk Valley.
Underutilization is an interesting and somewhat benign word.
You built a massive factory. There isn’t enough work for it. But you still have the factory.
Wolfspeed filed Chapter 11 and eventually wiped out about $4.6 billion of debt. Creditors became the principal owners. Existing shareholders were left with a small piece.
The financial structure disappeared. But the factory is still there.
Then the IRS sent $698.6 million in cash.
This is the thing I originally thought was ridiculous.
Section 48D gives companies a tax credit for qualifying investment in domestic semiconductor manufacturing. It doesn’t matter if the company is bankrupt.
But, in a sense, America got what it wanted. The government is incentivizing creative destruction. It happened, but it destroyed Wolfspeed.
So what exactly failed?
The shareholders certainly lost. The original capital structure failed. The jobs have not come close to what was promised.
But there is now a massive silicon-carbide factory in North Carolina that is producing crystals.
America says it wants more of that kind of capacity here.
China has made the problem easier on itself in one respect. The state can decide that an industry matters and build far ahead of ordinary economics.
The American version is different.
We want the factory, but we also want the market to prove the factory deserves to exist now. The problem is that by the time the market proves it, you’re years late.
So we try to solve the gap with customer deposits, tax credits, state incentives, debt, clawbacks, and very long contracts containing very careful definitions of what everybody did and did not actually promise.
Maybe that is the right system. It’s good for lawyers at least. It certainly produces a lot of paperwork.
And somehow, after everything, the factory is still there growing crystals.
Also, there is something remarkable about putting it in Siler City. It feels like progress. A different kind of bet.
America says it wants to build more factories like this.
I think it probably should.
But before we put another 500-acre bet on top of somebody’s chicken farm, it would be useful to know who is actually betting on what.
Daniel Sexton is a Partner at Vanguard Industrial Partners and founder of Arkvera, a deal advisory firm.