Robert Michael/picture-alliance/dpa/AP Images
Employees of the chip company Infineon walk along the clean room of its factory in Dresden, Germany, April 26, 2023.
There’s a latent tension at the heart of the Bidenonomics agenda between industrial policy and competition policy. Efforts to reshore manufacturing can lead to government-protected national champions. Some policy architects favor this for the purpose of efficiency, but concentrated industrial producers in the U.S. can then wield market power without substantially improving supply chain “resilience,” the main principle cited by Biden administration officials for their reshoring initiatives. And for other potential downsides, just think about Boeing’s current engineering problems.
This threat looms over the implementation of industrial-policy programs, especially the $53 billion in the CHIPS and Science Act earmarked for grants, loans, and other incentives to bolster U.S.-based semiconductor production. The firms that the Commerce Department selects for those programs, and the conditions attached, will structure the domestic market, to favor either bigness or resilience.
So far though, the CHIPS Program Office (CPO) has only awarded a small number of grants and issued few directives on what will determine the department’s review process, other than a very general paper put out last year called “Visions for Success.” One concrete proposal is to create geographic clusters of chipmakers.
With the Biden administration eager to deliver victories before the presidential election, a major round of funding is expected within the next several weeks. Recent reports indicate this round will entail several multibillion-dollar grants that will likely go to two of the largest semiconductor players: U.S.-based Intel and the Taiwanese semiconductor giant TSMC, the most dominant player for logic chips. TSMC holds a 57 percent market share for global foundry revenues, almost five times that of the nearest competitor Samsung, and even higher profit pools at around 80 percent.
Intel, one of the few remaining U.S.-based chipmakers, is both in the top ten globally for foundries and also holds an effective duopoly with AMD over CPUs, a specific niche of logic chips.
These big block grants were in some ways inevitable. The industry is already highly concentrated across most logic chip market niches with high barriers to entry. The funds can’t all go to incubate startups if the goal is to fundamentally reshape the market in the near term. But the new funding still raises the issue of whether the CHIPS Act can successfully revamp the American semiconductor industry, or if it will just amount to corporate welfare that builds concentrated power.
If companies had been smarter about putting earnings into capital investments over the past decade, then the government wouldn’t have to step in.
A timely new paper from the American Economic Liberties Project addresses this challenge head-on and provides a road map for how lawmakers can fix chip markets moving forward. The focus is not only on reshoring as a proxy for addressing the supply chain shortages that the U.S. experienced during the pandemic, but actually restructuring the markets so those shortages can be wrung from the system.
The paper argues that the goals of U.S. semiconductor policy should be to undo the “fabless” model of outsourcing overseas, which is favored by Wall Street, and to break the chokehold of TSMC on foundries. It will also require a “thickening” of the market of suppliers with footprints in the U.S.
FIRST, THE AUTHORS MAKE A DISTINCTION between the economic structure of the leading-edge processing chip market and the mature-node market. Leading-edge logic chips are the most advanced chips used for high technology and operate in a highly concentrated market. The market for the mature-node segment, which are still the most common ones used in consumer electronic devices, is more diffuse, and if anything suffers from gluts in overproduction because of foreign dumping practices and commodification.
Similar historical forces drove the development of both markets. The U.S. pioneered semiconductor technology in the middle of the 20th century, largely on the back of government-funded research and development projects at centers like Bell Labs. As a result, American chipmakers both powered the tech industry and were held to competition regulations such as open licensing requirements to allow for new entrants.
At that time, chipmakers were almost all vertically integrated and did everything from the design of the chips to the production in-house.
But trade agreements opened up the global economy to foreign competition. Under pressure, chipmakers both offshored production and scaled up by acquiring competitors, which were allowed by a lack of antitrust enforcement and changes to IP protections.
By 2020, when the pandemic hit, 60 percent of global semiconductor production and 90 percent of the most advanced chips were all centered in Taiwan, which put enormous strains on supply chains.
The main legacy of this structural shift in semiconductor production is the fabless model. It explains why companies like Nvidia, Qualcomm, Broadcom, and even Apple are the highest-valued chip players despite not actually being in the foundry business. This “capital light” model means that they only design the chips, own the intellectual-property rights, and then outsource the actual production to foundries overseas, where it’s cheaper to manufacture.
Wall Street prefers this method of companies earning high profit margins without needing to make real capital investments, all of which leads to exceptionally high payouts to shareholders, through dividends and stock buybacks.
According to the paper’s calculations, over the past five years, nine of the top ten largest U.S. semiconductor firms plus Apple have spent $698 billion in earnings on investor payouts, which amounts to 14 times the total spending in CHIPS. If companies had been smarter about putting earnings into capital investments over the past decade, then the government wouldn’t have to step in.
Both ends of this fabless model have been monopolized. The fabless firms are concentrated, and most of the production is carried out by TSMC overseas along with just a handful of other firms.
The rollup has happened rapidly. The paper finds that the number of top U.S. chipmakers declined by 44 percent just since 2010, in part due to a period of intra-sector mergers and acquisitions, which for the most part were cleared by federal regulators.
“Both foundries and fabless firms have substantial market power so there is a real cost concern even beyond the threat to resilience from outsourcing fabrication,” the authors write.
Fabless production mostly afflicts the leading-edge logic markets. Mature-node chips, by contrast, do not have the same monopoly problem, and instead face a race to the bottom from excess foreign competition.
While logic chips earn high profits, the margins for mature-node manufacturers in the U.S. are so thin that in many cases investment costs can outstrip returns. That’s in part because of foreign dumping practices. Most production for this segment of the market, which includes memory and analog chips, is concentrated in China, where firms work closely with the government and receive subsidies so they can dump products into global markets, which plummets prices.
Because there’s highly competitive pricing, most mature-node producers, such as Micron or Intel, opt for more vertical integration so they can capture as much profit as possible. The competition in this segment of the market is instructive, according to the paper’s authors, despite the many problems that policymakers need to address.
One particular focus of the paper is the immense power that Apple has over the entire semiconductor market, as the largest single buyer for all its iPhones and Mac computers. The tech giant’s exclusive arrangement with TSMC to manufacture all of its most advanced logic chips is the paradigmatic example of why the fabless model has been so destructive.
That deal takes Apple out of the market as a buyer for competitors and gives it first-in-line status with special discount pricing from TSMC that saves Apple billions of dollars. It also constrains capacity, because a substantial portion of TSMC’s production will already be locked up to Apple.
While the deal provides immense financial tailwinds for both firms, it accelerates the offshoring of U.S. production to Taiwan and considerably harms rival chipmakers.
More tech companies have followed Apple’s lead and replicated this fabless model in recent years. Amazon, for example, bought an Israeli semiconductor startup, removing a promising new player from the market. As the paper puts it, “Apple will make or break the Chips Act.”
THE PAPER LAYS OUT several categories of concrete reforms to supplement CHIPS Act funding, some of which are already within the Commerce Department’s authority for implementation. One overarching theme is that Commerce needs to work in concert with antitrust regulators and trade authorities in a whole-of-government approach.
The paper sets a central target for Commerce’s reshoring efforts: incubate at least four independent leading-edge logic chipmakers to ensure a “thicker,” more resilient market that can break concentration. That would likely entail strengthening Intel, Samsung, and potentially AMD as challengers to go up against TSMC.
“CPO should instead prioritize larger amounts of funding to the second-tier foundries,” the authors recommend.
But to ensure the commercial environment is sustainable for competitors, policymakers need to take additional actions on both the supply and demand sides.
For one, U.S. chipmakers need to have reliable long-term demand for their products, especially in the highly volatile mature-node market.
Government intervention could be one avenue as a buyer in the market to manage supply and maintain stable pricing levels. As the authors note, this arrangement is similar to how the government has intervened for decades to stabilize critical commodity markets, such as agriculture and oil, which are prone to booms and busts.
The authors also recommend legislation to force fabless firms to dual-source from multiple foundries to open up competitive opportunities for U.S. chipmakers. That could break Apple and TSMC’s chokehold over the market. Another potential remedy to dissuade the fabless model is to require open IP licensing practices, as federal authorities used to do, and guard against patent abuse. Limits on stock buybacks based on a sliding scale linked to investment could be beneficial as well, which Commerce can impose upon CHIPS Act funding recipients.
To further dissuade offshoring, the authors recommend a range of actions to both provide incentives for U.S. production on the supply side and purchase of American-made chips on the demand side.
Stronger tariffs and most-favored-nation rates are one way to give domestic producers an advantage, along with stronger enforcements against foreign dumping practices.
Another related solution is to impose a heavy fee on semiconductor purchasers, if these firms don’t source at least 30 percent of their aggregate chip components from the U.S. As carrots, there could also be tax incentives for these buyers to use domestic sourcing, similar to the Inflation Reduction Act. Both of those proposals would require legislation to pass through Congress, however.
Restoring America’s prowess in the semiconductor industry will entail a constellation of reforms that go beyond just the funding from the CHIPS Act. Anything short of that will not substantially reverse business dynamics.
“The CHIPS Act will be for naught if the leading edge fabless firms’ market power is left unchecked,” the authors conclude.
