By Jeff Ferry, CPA Chief Economist
Joe Biden is getting lots of advice from many quarters for his new administration. But last week he got some from an unexpected quarter: Bob Swan, CEO of Intel, published an open letter urging Biden to pursue a national manufacturing strategy for the US semiconductor industry. “Rising costs and foreign government subsidies to national champions are a significant disadvantage for US semiconductor companies,” Swan wrote.
Swan is right. The US badly needs a national manufacturing strategy, and a national R&D strategy, for the US semiconductor industry. The figure mentioned by the Boston Consulting Group (BCG) in a recent report, $50 billion, is a good estimate of the order of magnitude required. As Swan points out, US companies account for about half of the world’s microchips in terms of design but only 12 percent of the world’s chips in terms of manufacturing. In other words, half (and in my view, much more than half when it comes to cutting-edge applications like smartphones or 5G wireless systems) of the chips sold each year were designed by a US company, mostly by what are called fabless semiconductor companies, i.e. companies that design chips but don’t manufacture them. But only 12 percent of the world’s chips are manufactured each year in the US. And that 12 percent is sure to shrink as China, Taiwan, and other countries invest more in their chip manufacturing industries. China is investing the eye-watering sum of $100 billion in building a world-class chip manufacturing industry. And Nikkei recently reported that three new Chinese startups have gone on a hiring spree, wooing engineers out of the top US makers of chip design software.
What would a US manufacturing strategy entail?
- First, we need to invest billions of dollars in supporting our existing chip “fabs,” as semiconductor fabrication plants are known, and building new ones. We need a strategy that includes more than one US fab in each important area of chipmaking: processors, artificial intelligence (AI) processors, memory, analog chips, and so on.
- Next, we need government policies that incentivize the purchase of US-manufactured chips. A bipartisan bill from Senators Cornyn and Warner takes some good steps in this direction. But it relies too much on the US tax system. The “CHIPS” bill incentivizes investing in expensive chipmaking equipment, but it does not do enough to guarantee there will be customers for US fabs. Tax credits are not enough. Most US chip companies pay very little tax, and what tax they do pay is often a voluntary decision. For example, Nvidia, the hottest chip company in the world right now, paid just 2.6 percent tax on its pretax profit of $2.94 billion in the first nine months of this year. And every single startup chip design company is required by its venture investors to lose money and therefore pay no tax. Making a profit as a startup would probably get a CEO fired immediately. The government must develop a financial mechanism to directly compensate chipmakers for using US fabs, while penalizing them in ways that will cost them real money for using foreign fabs. This is crucial for national security and economic reasons.
- Third, the government needs to focus on incentivizing a number of universities to build up great technical departments focused on chips and supporting technologies, especially material science. We had these resources in the early days of chip development and we have lost them. The lack of skilled professionals is a serious handicap for our ability to build and staff new fabs and generate cutting-edge products from them. I’m not talking about the obvious glamorous places like Stanford and MIT. At those places you can practically feel the money oozing out of the walls—and see the self-important young students rushing around, powered by their dreams of becoming the next Mark Zuckerberg. I’m talking about state universities, designed to attract and teach kids who understand engineering is not easy but that the starting salary for a 22 year old engineer in a Silicon Valley company is in six figures. And if you have those skills, you will never want for work. At the chipmaking startup where I worked we made optical chips. The engineering department at UC Santa Barbara played a key role in supplying us with the engineers that designed and manufactured our breakthrough cutting-edge products, and made our IPO possible. After the 2007 IPO was complete, all our senior engineers were millionaires.
- Fourth, we need to seek competition in chip fabs. It’s great that TSMC (world’s leading chip fab company, based in Taiwan) is building a fab in Arizona. But we need more choices. We need multiple fabs in each segment of the industry. Venture capitalists know you need to make multiple bets because any single bet can fail. We have some great lesser-known US fabs, like GlobalFoundries, that can be supported to develop into mainstays of the entire US chip industry, and we need some greenfield startups alongside the existing players. The Boston Consulting Group is in the right ballpark in suggesting we should target some 19 new fabs. But we need to back this up with the policies that can ensure those fabs will have customers, revenue, and profit.
The situation in the US is actually much worse than Bob Swan let on in his open letter. Intel is in big trouble. A couple of weeks ago, Apple announced that all of its new laptops will from now on run on Apple-designed processors fabbed in Taiwan, instead of the Intel Core chips they have been using for years. And Microsoft is reported to be looking at replacing the Intel processors in its Surface line of tablets with new chips that would be fabbed in Taiwan.
On top of this, Intel’s dominant position in the crucial data center market is under siege from Nvidia, and its efforts to move to smaller, denser processors (10 nanometer and 7 nanometer) are behind schedule. Intel is the only US-based manufacturer of processors. If it decides to outsource building those smaller line-width chips to another manufacturer, that manufacturer is almost certain to be in Asia. Swan knows the firestorm this would create in Washington, and that is part of the motivation for his open letter.
We need Intel to succeed. But its current problems suggest that the fully vertically integrated model (i.e. chip design and chip manufacture in one company) is not working. We need to build a disaggregated industry in this country.