No Gas, No Chips: The Hidden Supply Risk Beneath the Fab
In a semiconductor fabrication plant (or “fab”), production can be derailed by a shortage of a material worth US$20 as much as one worth six figures.
That was one of the uncomfortable realities mentioned during “Beneath the Fab: Securing the Semiconductor Materials Layer,” a panel discussion earlier this month hosted by the American Enterprise Institute (AEI), a Washington, D.C.-based public policy think tank.
A typical fab uses roughly 150 specialty gases, along with numerous liquid chemicals and other materials that collectively represent about $5 billion in annual economic activity. For a semiconductor industry worth about $1 trillion, that’s a rounding error.
However, an input’s value doesn’t determine its criticality, said panelist Carl Jackson, cofounder and managing director of SSoT Group, a Singapore-based builder and operator of specialty-gas production plants: “If you lose one of those 150 gases and inputs into a fab,” he said, “It’s no gas, no chips. The fab’s down.”
The other panelists were Olivier Blachier, president, materials solutions at Entegris, a Billerica, Massachusetts-based materials and chemicals supplier, and Michael Robbins, president of the Semiconductor Industry Association (SIA) in Washington. The moderator was Chris Miller, AEI nonresident senior fellow and author of the 2022 bestseller Chip War: The Fight for the World’s Most Critical Technology.
Their lessons for supply management organizations extend well beyond semiconductors: Risk cannot be measured by spend alone. The inputs that deserve the most scrutiny might be buried several tiers below a company’s largest suppliers, where visibility declines, geographic concentration increases and substitution becomes far more complicated.
“There’s a vulnerability there in terms of the complexity of the supply chain, the amount of materials that go in, where these materials are dug out of the ground, how they’re purified, and everything else,” Jackson said.
Look Beyond the Supplier’s Address
According to a recent SIA study of about 300 inputs into chip manufacturing, 95 percent of top-level semiconductor equipment is sourced from the U.S., Europe or Japan. That figure, Robbins said, “looks great on the surface.”
He continued, “But when you go down to the next level — the materials that are making up the equipment and that are going into the equipment — you start getting into a high degree of vulnerability.”
How much? Go one tier deeper, and half of the global supply chain was classified as at medium or high risk of geographic concentration, export controls or geopolitical disruption. Go deeper still, and the number of suppliers often shrinks to one or two — a “totally different picture,” Jackson said.
Consider tungsten hexafluoride (WF6), a gas used in semiconductor manufacturing, which can be purified and processed somewhere that gives a fab a seemingly diversified supply base. But tungsten and fluorine have to come from somewhere.
China accounts for about 80 percent of global capacity of tungsten and 60 percent of fluorspar (which is used to produce fluorine and its derivatives), Jackson said. And tungsten, as Miller noted, is under Chinese export control.
“That’s just one example of where the tap can be turned off,” Jackson said. “And there are many examples of (that risk) where a root mineral comes from.”
A Second Source Isn’t Always a Solution
The natural response to concentration is to find another supplier. In semiconductor materials, that can be considerably harder than it sounds.
The materials entering a leading-edge fab have to be extraordinarily pure. “Take a human hair diameter and divide it by 1 million,” Blachier said. “This is the size of the largest impurity that can get into a chip.”
Some gases are toxic, corrosive, flammable or pyrophoric, and impurities measured in parts per billion or parts per trillion can affect production. That means a fab has to be not only a highly advanced chemical plant, Blachier said, but also one of the world’s most complex contamination-control systems.
Leading-edge semiconductor processes can involve roughly 1,200 individual steps. To achieve an economically viable final yield of 80 percent to 90 percent, Blachier said, each step has to operate at approximately 99.9999 percent reliability. “You need to have perfection in every step,” he said.
As a result, supplier qualification is a critical part of semiconductor risk mitigation. A chip manufacturer cannot simply place an order with another producer and assume the material will perform identically.
Transporting highly sensitive chemicals across long distances can increase contamination risk, which is one reason the industry increasingly favors production and purification facilities near major fab clusters.
“You don’t want to ship these pyrophoric or flammable materials from one part of the world to the other part of the world,” Blachier said. “You need to package them in what we call high-purity cylinders (and) quality control them to an extreme level. Once you have these facilities built close to the clusters, (regional shipping) is pretty well-handled nowadays by the gas companies.”
Robbins said companies are increasingly requiring bills of materials that extend multiple layers into the supply chain, searching for alternate suppliers and investing in applied science and advanced chemistry to develop substitutes.
“Innovation can also come in the form … of using gases and rare earths or creating alternatives,” he said. “So, I think the industry is paying attention to this and investing in their supply chain teams and looking at this much more closely than in previous years.”
China Built the Layer Beneath the Fabs
China’s position in semiconductor materials also illustrates why building semiconductor fabs is only part of the supply chain equation.
The country has spent roughly two decades developing its semiconductor industry. In 2014, Beijing launched a government-backed program known in the West as the “Big Fund.” This initiative, Jackson said, is broader in scope than the U.S. CHIPS and Science Act of 2022 because it extends across the semiconductor supply chain.
The fund has exceeded $100 billion in total investment, compared with $38 billion allocated through the CHIPS and Science Act. China had roughly 15 percent of capacity in the specialty-gas layer in 2014, Jackson said; now it a has a majority share.
Much of that growth has come through Chinese provincial investment, with multiple provinces building capacity for the same products. The result is substantial overcapacity in some materials — and significant leverage.
“I think the U.S. has … not hollowed out, that would be unfair to say,” Jackson said. “But it’s been pretty stagnant, I think, compared with the investment in China (in) the last 10 years because of the Big Fund. So, as of today, China probably has the infrastructure to support the world for most of the high-volume gases.”
That doesn’t mean that global chip supply chains are failing, Robbins said. He pointed out that globalization has helped keep semiconductor production costs far lower than they would be in a fully indigenous supply chain in, say, the U.S. or Taiwan.
“We do need to absolutely address some of the single points of failure and some of the potential fragilities and chokepoints,” he said. “But when the global supply chain works, it works very, very well.”
Looking Beyond the Next Bottleneck
There is one more lesson for supply managers: The most important material today might not be so tomorrow.
As semiconductor manufacturers push toward smaller, more powerful chips, they are confronting another problem — heat. Advanced AI chips consume enormous amounts of power, making thermal management increasingly important.
Blachier pointed to graphite, silicon carbide and diamond as materials that could become critical to the next generation of semiconductor and AI infrastructure. China has identified silicon carbide as a future strategic battleground, he said.
“The companies that master these supply chains,” he predicted, “will be companies that are winning the AI race three to five years from now.”
That means supply chain strategy must go beyond protecting today’s materials. Procurement, engineering and R&D need to understand which materials could become constraints as products evolve.
The semiconductor industry offers a useful warning for supply managers everywhere: Knowing your supplier is not the same as knowing your supply chain.
“That’s where resilience and different supply chain concepts come into play, where you cannot nowadays rely only on one single region to export to the rest of the world,” Blachier said. “Because you would be foolish to do that in the current world we live in.”