Thrown to the Wolves: A Tale of the CHIPS Act
In January of 2020, we recommended that our Gilder Technology Report readers invest in Wolfspeed (WOLF), an American company and leader in semiconductor devices made not from silicon (Si) but silicon carbide (SiC), a so-called “wide bandgap” semiconductor, like a diamond in the rough. Think of a new diamond age, consisting of wide-bandgap diamonds made from silicon and used in wafers, microchips and electric cars rather than around the necks of wealthy women.
Wide bandgap means it can deal with power rather than merely with energy. The wide bandgap of silicon carbide, for example, enables creation of expensive convertors that can take the otherwise useless energy fits and starts from windmill totem poles and sun henge solar panels and turn them into reliable power with the help of costly batteries.
Why anyone would want to do this is beyond us. The so-called Inflation Reduction Act, however, values these eyesore energy exhibits at a putative $6 trillion in mandates and subsidies. A world delusional about energy and obsessive about climate performs many foolish, egregiously costly, and sacrificial rituals, which may well face sacrilegious setbacks in the Trump administration. But power semiconductors like those from Wolfspeed have thousands of uses beyond these religious rites—for automobiles, data centers, lasers, light-emitting diodes, and even my paradigm breakthrough of wafer scale graphene on silicon carbide. It switches a thousand times faster than silicon. Look it up in our Gilder Technology Report.
Our reasoning was sound, even precocious. We saw that more and more computers were fleeing the desktop, the laptop, the office, and other cozy controlled environments determined to make their way in a wild and dangerous world. The new power devices would conquer fields as hostile as the innards of an automobile engine, take on high voltages out on the grid, or drive “5G” wireless devices running faster and hotter than silicon chips were ever intended to endure.
In the absence of available diamonds, silicon carbide (SiC) and other “wide bandgap” materials such as gallium nitride (GaN) were far more suited than silicon to these barbaric conditions. They could operate at higher frequencies, endure much higher temperatures, and operate at much higher voltages, before experiencing “electrical breakdown” which turns semiconductors into continuous conductors, unable to switch on and off to perform operations.
We announced to our readers a new technology “paradigm,” as we call powerful technology trends dictated by the physics, forces to which resistance is futile regardless of the preferences of producers or consumers. We named this paradigm “Life After Silicon.”
Since 2020, the market for these alternatives to pure silicon has nearly doubled and will double again by the end of the decade. Electric vehicles have been a big driver of that growth but far from the only one.
Searching for companies that would lead on the wide-bandgap thruway, we found several including “II-VI” (now Coherent (COHR)), and Infineon. The standout, however, was Wolfspeed. WOLF nailed its flag to the silicon carbide mast. It sold its LED lighting business to help finance a transformation into a pure play silicon carbide enterprise. It would make the raw material, manufacture and finish the “wafers” on which microcircuits would be inscribed, and finally the chips themselves. But in the long run, most of the chips may even give way to that wafer-scale graphene on silicon carbide.
Boldly, WOLF set out to build the world’s most advanced silicon carbide chip “foundries,” processing wafers of 200 mm diameter as opposed to the more common 150 mm. That 33% increase in diameter would yield an 85% increase in the number of chips. However many silicon carbide chips the world would demand, Wolfspeed would provide.
Today, the company faces bankruptcy, or at least it should.
Being on the right side of a technology paradigm is a great advantage. It cannot overcome management incompetence, arrogance, and folly. These may be disguised for some time as routine and acceptable setbacks—a construction delay here, a bit of volatility in demand there. After all, everyone knew the transition from 150 mm diameter wafers to 200 mm would be challenging and costly. Doesn’t management deserve a little breathing room?
And breathe it did, until it had sucked all the oxygen out of the business. In five years, even as revenues rose from circa $470 million in 2020 to circa $800 million, operating losses rose from $205 million to more than $600 million. Net losses rose from $192 million to $751 million. Write-downs of obsolete or overvalued plant and equipment may reach $400 million.
In 2020, WOLF had negative net debt (the amount by which its cash and equivalents exceeded total debt) of some $440 million. Today, WOLF carries some $4.5 billion in net debt.
The obvious solution, the right solution, is bankruptcy. WOLF built too fast and paid too much, but the demand for SiC devices is real and growing. Cleared of debt, this would be an enterprise valuable to new owners and the nation.
It won’t happen. Instead, the company—and apparently the blundering management—will be saved by some $2.5 billion in “investment” and loans from Uncle Sam via the CHIPS Act. It seems that to the folks who dole out CHIPS money, grotesque overspending on factories still waiting for demand, and fraught with mismanagement, is just what America needs to remain competitive in semiconductors. Because, as everyone knows the wealth of nations comes from big, big, factories that lose big, big money, like those that made the Soviet Union—and Mao’s China—great.
Meanwhile, Nvidia is pushing past $100 billion in revenue, has been growing at more than 50% a year for the last three years, has operating margins of more than 60%, net margins (after taxes) of almost 50% and negative net debt of more than $26 billion.
NVDA has done all this without manufacturing a single chip. Nearly all its manufacturing is outsourced to Taiwan Semiconductor (TSM), the world’s leading outsourced semiconductor firm, manufacturing not only for NVDA but AMD, Apple, Broadcom, Marvell, MediaTek, Qualcomm, at times Intel (INTC), and even wafer-scale ventures with Cerebras and Tesla’s DoJo tiles for its autonomous cars and robotics.
TSM is a great company. But the U.S. government regards it as uncertain, because of its closeness to China. And we are provoking China at every chance with sanctions, moralistic posturing about Uigur terrorists, ridiculous charges of technology “theft,” and aircraft carriers in the Taiwan Straits. So, even though TSMC has earned almost 60% of the outsourced market for semiconductors, we see it as replaceable by some vastly costly copycat fabs in Arizona presumably powered by windmills and solar panels. Obsolete before they are built.
Nonetheless, TSMC’s revenues are less than a fourth of Apple’s and its operating margins are but two-thirds of Nvidia’s. TSM’s market capitalization is less than one third of either AAPL or NVDA.
By all means, let’s pour billions more into money-losing chip manufacturers to keep them from the bankruptcy that is their best hope for survival.
And waste $6 trillion on those religious rites that make our power grid more volatile and vulnerable. Just kidding, of course, but in America’s “life after capitalism,” nothing should be taken for granted. Let’s hope Trump will be allowed to fix it.