Industrial Policy: When Is Business the Government's Business?

The four Cs of when government should intervene

For the first time since World War II, the US is practicing industrial policy at scale. We have always exercised some version of it — shaping markets through tariffs, subsidies, tax credits, and R&D support (from railroads and land grants to DARPA and NASA) — but in recent years the scale has changed dramatically. The Biden administration’s CHIPS & Science Act and Inflation Reduction Act were explicit efforts to shape and influence the semiconductor, clean energy, and EV markets. The Trump administration has moved even closer to what some might call “state capitalism,” with the government taking direct stakes in key firms (such as Intel and MP Materials), receiving golden shares, and floating the idea of a sovereign wealth fund.

The expansion of Washington’s role in the economy raises fundamental questions about the scope, goals, and execution of industrial policy. Are these interventions a costly distortion of markets, or a necessary correction when markets break down? Should the government ever “pick winners”? When, if ever, is industrial policy justified?

A heuristic for industrial policy

I have spent my career in the center of American capitalism and believe deeply in the power of markets, competition, and comparative advantage. But I’ve also seen the economy hobbled by uneven playing fields, (rationally) ignored externalities, and brittle supply chains. In high-stakes situations where market forces fail to incorporate national security imperatives, targeted industrial policy can work. But interventions should be rare, limited to when four tests are met:

• The industry is strategically critical.

• The industry is compromised in a way that creates vulnerability and systemic risk.

• The industry has calcified in a way that perpetuates resiliency failures, and cannot (or will not) self-correct.

• The problem is correctable by a bounded government intervention that creates a sustainable market structure.

Critical

Given the current tendency to label everything a national security emergency, we need to be precise about whether an industry is truly strategic and critical. Industrial policy is appropriate when the industry has direct bearing on (a) US military or intelligence operations, (b) the physical health and safety of Americans, or (c) key technologies of future global competition. Interventions should also focus on industries that produce foundational inputs to others — those whose failure would ripple across the economy. Rare earths, magnets, semiconductors, and even batteries fit that description, as they all underlie multiple end products.

Compromised

But importance alone isn’t enough to warrant intervention. Another condition for industrial policy is that the industry is structurally compromised in a way that makes the system inherently brittle or creates a potential chokepoint. That might mean concentration in a sensitive geography, dependence on a single company, or conditions that lock in the status quo. If an industry is critical but widely diversified across multiple companies and allied geographies, intervention is unnecessary.

Calcified

Markets can be highly efficient, yet dangerously and systemically vulnerable.

In classical economics, a market failure occurs when the free market fails to allocate resources efficiently, leading to suboptimal outcomes in pricing or output. Critical industries are exposed to a different failure mode — markets often fail to account for or price in national security risk. Over time, supply chain vulnerability gets built into the industry structure. Industrial policy should target these calcified structures in critical industries that markets alone will not fix.

Markets rarely price in geopolitical tail risk or the collective value of supply chain resilience; if a risk is low probability but high impact, investors tend to discount it entirely. Take Apple, which has total dependence on TSMC fabs in Taiwan. A single catastrophic event — an earthquake, or a Chinese invasion — could halt all iPhone production. Yet Apple’s stock price shows no discernible discount for that existential exposure. The market effectively assumes such an event would cripple the global economy anyway, so it prices it as zero.

There are also public good and coordination failures that prevent markets from self-correcting. The national security benefits of supply chain resilience accrue to society, not any single firm. No one company can justify diversifying its supply chain if doing so raises costs and erodes competitiveness, especially while rivals take no action. As a result, rational companies perpetuate a collective vulnerability.

Correctable

For industrial policy to make sense, there must also be a plausible corrective path that leads to a self-sustaining outcome.

Executing industrial policy is notoriously hard. It invites political capture, bureaucratic error, unintended consequences, and market distortions. Temporary subsidies often metastasize into permanence. Resources are finite; every dollar spent on industrial policy is a dollar not spent elsewhere, so we must spend it well.

Still, there are situations when targeted, time-bound intervention can break a calcified market structure and crowd in private capital. The goal is to use public leverage to reach a tipping point, after which scale, learning, and ecosystem effects create a new equilibrium that can stand on its own.

Leading-edge logic semiconductors: a bullseye for industrial policy

One critical objective for the CHIPS Act was to stand up at least two scaled, leading-edge (LE) logic ecosystems in the US — each with multiple fabs and the infrastructure, supplier network, and workforce to sustain them. This vision was laid out in our Vision for Success in February 2023; one year later, we set an ambitious goal to manufacture 20% of the world’s LE logic semiconductors in the US by 2030. At the time, the US produced 0%. Achieving it would literally redraw the map of this crucial industry.

Leading-edge logic satisfied all of our prerequisites for industrial policy.

First, the sector was critically important. Leading-edge logic chips — sub-5nm transistors fabricated with extreme precision — power the most advanced systems on earth. They are the compute engines for AI accelerators and data centers; the brains of our smartphones and telecom infrastructure; and critical components in satellites, cyber defense, and certain advanced weapons systems. They also function as general-purpose technologies, propelling advances in drug discovery, autonomous driving, and quantum computing. They are the foundation of our technological future and therefore a clear national-security priority.

To make matters worse, the industry was vulnerable and compromised by a dangerous geopolitical chokepoint. Before CHIPS, over 90% of the chips used in US production were sourced from a single company (TSMC) in a single place (Taiwan). And that geography is under fundamental geopolitical threat from the PRC. Korea’s Samsung supplied the remainder, but lagged behind at the most advanced nodes. This combination was key: the product was strategically critical and concentrated in one of the world’s most geopolitically volatile fault lines. If those same chips were available from a range of firms across multiple allied geographies, intervention might not have been necessary. But the geographic and corporate concentration compromised the whole industry and created an unacceptable single point of failure.

Markets had calcified around the failure and could not correct it. Was this simply the “natural” result of comparative advantage — i.e., did Taiwan have fundamental resource, labor, or cost advantages that could not be overcome? No, the structure wasn’t natural. It was path-dependent and could have evolved in other ways. The US dominated the leading edge from the 1990s to the early 2000s, but lost ground as compounding factors took hold: Taiwan carved out subsidies and established a national focus on the industry; US financial markets rewarded asset-lite business models and offshoring; escalating complexity and cost drove out most competitors; and, importantly, Intel made some well-documented missteps.

Observers often allege that chips are manufactured abroad because labor is cheaper in Taiwan. But labor arbitrage is not decisive today. One report suggests that in 1991, labor was over 40% of costs industry-wide; today, extreme automation and increased capital intensity has driven that closer to 13% at TSMC. These changes to the industry, along with scale and ecosystem effects (and TSMC’s strong management), reinforced each other until the market structure became calcified — locked into an “unnatural” state that private capital alone could not break.

All of TSMC’s major customers (Apple, NVIDIA, AMD, Qualcomm, Microsoft, etc.) recognize the danger of this dependency on TSMC. However, none were willing or able to change the structure on their own. No company will willingly raise its own costs or lose queue priority at TSMC for the sake of collective resilience. Financial markets likewise ignore the existential geopolitical tail risk. You’ll search in vain for the Taiwan risk discount in any of these valuations. In short, the market failed to account for resilience.

We believed we could sustainably correct this vulnerability. The underlying economics weren’t hopeless, just skewed. By closing the up-front construction-cost gap, we hoped to change the calculus on new US capacity, unlock private capital, and let market forces coalesce around a resilient and competitive equilibrium in the US. We had a few factors in mind:

• Cost structure: The largest cost differential between the US and Taiwan was in construction, not operations. Once built, ongoing operating costs should converge – labor costs are higher in the US, but energy costs can narrow the delta. If we close that initial gap to Taiwan and Korea, US fabs can be close to cost-competitive over time.

• Ecosystem effects: Taiwan’s advantage is a dense ecosystem of suppliers, packaging facilities, and skilled labor. As new US fabs get built (and with targeted supplier incentives and investment in workforce pipelines), clusters will form, costs will fall, and self-reinforcing scale will emerge.

• Strategic value: Even if a small cost gap remains, value can outweigh price. Proximity to customers, integration with US research institutions and their talent, and greater supply chain resiliency all carry tangible worth. The US needs to be value competitive, not cost competitive.

A helpful analogy: Texas Instruments (TI) manufactures most of its mature analog chips — which sell for <$1 — in the US. Despite fierce price competition with its many Asian and Chinese competitors, TI remains globally competitive through scale, automation, and ecosystem depth. If the US can compete in the oversupplied, low-price point analog world, it should be able to compete at the leading edge too where pricing power is higher and there are fewer competitors.

Projected demand growth also gave us confidence in our ability to shift the market. AI and advanced compute workloads are driving an explosive need for leading-edge capacity — projected to grow 69% by 2028. To meet this demand, many new fabs will be built somewhere — policy can influence where.

In a flat market, adding capacity is zero-sum and likely destabilizing and self-defeating. But in a growing market, shifting supply geography can be positive-sum. The existing supply/demand balance and projected growth matter — and for LE logic, they create positive tailwinds.

Industrial policy as disciplined statecraft

Narrowly applied industrial policy can strengthen resilience and reshape industry outcomes in ways the private sector alone will not.

LE logic satisfied the four C’s of industrial policy: the industry was critical, compromised, calcified, and correctable. But discipline is everything. This framework can help identify other industry segments to target, but we should err on the side of caution. Every market intervention comes with downsides, economic costs, and unintended consequences. Execution — deciding which tools to use, in what combination, for how long, and with what outcomes and accountability — is a delicate art.

When used sparingly and executed with rigor, industrial policy can be an instrument of national resilience. But industrial policy is a corrective, and not a replacement for markets. In rare cases when markets alone cannot secure the future we need, disciplined industrial policy is the best tool to restore balance.