In coming days, the House of Representatives will consider expansive legislation in part designed to prevent future microchip shortages like the one that has reportedly drawn down U.S. stocks to a five-day supply. The bill, now called the America Competes Act, would appropriate around $52 billion to restore U.S. dominance in semiconductor design and manufacturing. But to succeed, congressional leaders must develop a clear operational plan for spending the funds, make sure that the resources are not spread too thinly across competing organizations, and focus on supporting leading-edge manufacturing capabilities.
While semiconductors drive nearly everything in our networked world, the United States produces only 12 percent of the world’s chips, down from about 40 percent two decades ago. Supply-chain breakdowns and tensions with China have rightly led officials to seek to shore up the U.S. position. However, any foray into industrial policy is fraught with risks. The stakes are high, and Congress will have only one shot at deploying resources at this scale. Legislators should ask several questions.
First, does the legislation have a clear operational concept to deploy these resources effectively? In the military, an operational concept refers to how resources — such as soldiers, units and equipment — are brought together to achieve a desired strategic outcome. A good starting point would be to focus on two recommendations made by the bipartisan National Security Commission on Artificial Intelligence : ensure that the United States stays at least two generations ahead of China in semiconductor design, and maintain multiple sources of cutting-edge microelectronics fabrication at home.
Second, does the bill run the risk of allocating the billions of dollars for research and development too thinly, across duplicative organizations? The bill would appropriate about $13 billion over five years, to create four new organizations. The first, the National Science and Technology Center, makes the most sense. It would be a consortium of government scientists and private sector experts that is loosely modeled on the late-1980s Sematech consortium. Supported by the Defense Department, that organization helped to establish U.S. dominance in semiconductors. Prior to Sematech, it took some "30 percent more research and development dollars to bring about each new generation of chips. Thus, that collaboration created important advances in the sector.
Less promising are the other initiatives: the National Advanced Packaging Manufacturing Program, a microelectronics research program, and the U.S. Manufacturing Institute. The legislation suggests that all would tackle chip design, assembly and packaging. Given the complexity of these processes, does it make sense to spread resources across multiple centers? Particularly since TSMC and Samsung, the world’s only producers of advanced chips, spent some $9.3 billion on research and development in 2020 alone. And will there be enough talent to populate the potentially competing centers? Not only is there a shortage of around 70,000 skilled workers across the industry, but the most advanced researchers are also in short supply.
Building fabrication facilities, or “fabs,” to manufacture the chips should be the second part of the operating concept. The U.S. Innovation and Competition Act (USICA) authorizes $39 billion for that purpose, and the question is whether that will achieve the desired result of establishing leading-edge manufacturing capabilities.
While this sum might seem generous, it’s a modest amount in the capital-intensive semiconductor industry. Leading or near-leading edge fabs can cost close to $20 billion, depending on the chip size. Unless complex national security waivers are granted, USICA will offer a subsidy of up to $3 billion per fab, which would be 15-25 percent of the cost. That’s not nearly enough to bridge the cost disadvantage facing U.S. firms. Depreciation plus operating cost differences between Taiwan- and U.S.-based fabs range from 30 to 50 percent.
Top policymakers, including Sen. Chuck Schumer (D-N.Y.) and Secretary of Commerce Gina Raimondo, have said funding would help build seven to 10 new fabs. However, the legislation does not specify what kind of fabs should be built. Many chips in devices from cars to mobile phones are not leading-edge.
Increasing domestic production of these chips could alleviate some supply chain problems. But to stay two generations ahead of China, and to meet national security demands, the United States needs leading-edge chip capacity at scale and inside the United States.
Today, Intel, the only near-leading edge U.S.-domiciled company, is still about two generations behind the Taiwan-based Taiwan Semiconductor Manufacturing Company (TSMC). Support for leading edge fabs means that the $39 billion would need to focus on funding Intel, Samsung and TSMC so that they build facilities here. All have recently announced plans for large U.S. manufacturing fabs, which is positive. However, we should only provide grants to foreign companies if their facilities are located in the United States and are run and managed by Americans.
Finally, how long is the life cycle for the fabs? Speed is not a strong suit of our government. By the time the Commerce Department establishes a process to vet businesses that have a “documented interest” in constructing, expanding or modernizing a fab, how much time will have passed? The plan should abide by metrics to begin funding within three months of passage and to last for the amount of time necessary, which might be longer than five years.
The sad reality is that, while the Manhattan Project took four years to develop a nuclear bomb, it’s been over seven years since the Congressional Semiconductor Caucus began to draft legislation related to this critical sector. Every other country in the world subsidizes its semiconductor sector. America needs $52 billion to keep its competitive edge. But before we congratulate ourselves too much, Congress must make sure the money is well spent, or the billions of dollars under consideration will amount to much less than the sum of its parts.
Read in The Hill