Intel’s plans to invest up to $100 billion in a new group of chip factories outside Columbus, Ohio, will have a much greater impact on the future of its manufacturing division compared to any short-term political or supply-chain concerns it might solve.
To hear President Joe Biden, U.S. Commerce Secretary Gina Raimondo and Ohio Governor Mike DeWine tell it, the new factories — known as fabs in this world — are going to help fix inflation, make the U.S. more competitive, drive down the soaring cost of cars, ease the chip supply-chain shocks and improve U.S. national security. That’s a lot, even for one of the biggest projects in Intel’s storied history. It will be years before that capacity comes online, and whether a new chip factory in Ohio could actually solve any or all of those issues is debatable.
Put aside the politics; the stakes for Intel’s efforts in the Buckeye state couldn’t be more clear.
The company missed out on manufacturing billions of smartphone chips over the last decade, failing to make its own processors or win manufacturing contracts from other mobile giants. That cost the company enormously — Intel didn’t benefit from the technical manufacturing know-how that comes along with making chips at even higher volumes than its PC or server businesses produce. After a prolonged period of disarray, chip manufacturers such as TSMC and Samsung seized upon Intel’s weakness and leapfrogged ahead of the Silicon Valley giant’s one-time manufacturing leadership position.
Under Intel CEO Pat Gelsinger, the company has vowed to regain that former status, turning around its own products’ performance in the short run and regaining manufacturing supremacy down the road. “Intel is committed to restoring end-to-end leadership, innovation and manufacturing here in the U.S.,” Gelsinger said Friday. “We are doing our part, but we can't do it alone.”
To get there, Intel is doubling down on its own contract manufacturing business for fabless chipmakers. And it has embraced the extreme ultraviolet lithography technology others such as TSMC have used to successfully make advanced chips.
Ohio is a significant component of that comeback plan. As part of the company’s renewed push in manufacturing, Intel has already announced a $20 billion expansion of its sprawling factory site in Arizona as well as a $3.5 billion expansion of its packaging site in New Mexico. It has also said recently that it plans to announce two new sites in Europe in the coming months, and may invest as much as $95 billion there over the next decade.
By themselves, Intel’s Ohio plans aren’t a sure bet that it can regain its position of manufacturing leadership, but they are a start.
The first factory isn’t set to come online until 2025 and it will probably take $20 billion in investment to buy Intel enough capacity to start producing 40,000 wafers a month of advanced manufacturing capacity, according to estimates by Harvard Business School professor Willy Shih. Right now, Shih estimates Intel’s highest-volume factories are cranking out around 100,000 wafers a month.
But getting to high volume production in Ohio will be quite expensive as Intel invests in even-more-advanced manufacturing technology. And it’s not the only chipmaker striving to lead this industry.
Moore’s second law
Intel’s bets on Ohio, Arizona and elsewhere are a huge risk for the company. But today’s massive price tag will almost certainly be dwarfed by what it will have to spend in the future to stay competitive, according to a lesser-known law articulated in 1995 by Intel co-founder Gordon Moore.
His first “law” is widely known and cited: Moore observed about 60 years ago that the number of transistors printed onto an integrated circuit will roughly double every two years, and the machines they power will get cheaper. That observation held true for a very long time but today, its veracity is a matter of debate among the chip giants.
Moore published another important paper in 1995 with a different and potent observation about semiconductor manufacturing. Moore had begun to worry about the costs of the tools needed to make chips, which had begun to increase at an exponential rate. He was concerned the trend would continue to the point where it would be economically impossible to build new factories.
Intel's proposed chip-making campus in Ohio is a key part of its future plans.
Image: Intel Corporation
“The rising cost of the newer technologies is of great concern,” Moore wrote. “Capital costs are rising far faster than revenue in the industry. We can no longer make up for the increasing cost by improving yields and equipment utilization. Like the ‘cleverness’ term in device complexity disappeared when there was no more room to be clever, there is little room left in manufacturing efficiency.”
Moore’s second-most-famous observation hasn’t quite come true for Intel and others, yet. But the latest generation of tools has an eye-popping sticker price, and are only getting more costly.
‘A high stakes game’
Intel has been secretive about the mix of chips it plans to fabricate at its forthcoming Ohio sites, but has said that it plans to deploy next generation EUV tech there. Existing EUV machines are expensive, costing roughly $180 million a piece. The next generation, called high-numerical-aperture EUV, will cost roughly twice that, and Intel has already committed to purchase the first tool exclusively made by the Dutch ASML.
Lithography tools such high-NA EUV machines have long been the most expensive bottleneck for manufacturing output. But fabs need hundreds of other tools, too, which can cost tens of millions each.
“You thought $150 million for a tool takes your breath away, but then you see how much Intel is paying for the high-NA [equipment] too, and it’s crazy,” Shih said. “At the [size of Intel’s Ohio fabs] they’ll have like 1,000 different tools, and they cost a couple million to tens of millions each.”
But while those prices seem exorbitant, the real risk is that Intel might not be spending cash on manufacturing at the same rate as some of its rivals. Late last year, Intel said the company’s capital spending plans for 2022 would be as much as $28 billion, likely including some of the costs of beginning construction in Ohio. The other six Ohio fabs, which represent the remainder of the $100 billion overall cost, hinge on Congress funding a $52 billion subsidy program that will help Intel defray the price.
“We think $20 billion is a lot of money,” Shih said. “The easiest thing for people to think is that Intel is going to put up all this money and catch up. But, is it going to catch up, or just keep up?”
TSMC already can produce many more wafers than Intel — roughly a million a month — and has promised to spend as much as $44 billion in 2022. Those tens of billions will expand its fabs in Taiwan and China and also includes part of a $12 billion factory plant outside of Phoenix.
Samsung, one of the other companies that uses EUV technology to make advanced chips, will spend $42 billion on capital equipment this year. Samsung also has plans for U.S. manufacturing, and has committed $17 billion to a new facility in Taylor Texas that is expected to begin operations in 2024.
“This is a high-stakes game, and it takes a lot of money, just because the equipment is so expensive and the technology is so hard,” Shih said.
