Latest / The Fiscal Policy Podcast with Fexingo: Government Budgets, Stimulus, and Public Spending / The Real Cost of Government Subsidies for Chip Manufacturing
Transcript
- Lucas: Let's talk about the biggest industrial policy experiment in a generation: the government's push to bring semiconductor manufacturing back to the United States. Luna: You're talking about the CHIPS Act — the $52 billion subsidy package signed into law in 2022. Lucas: Exactly. And now, four years on, we have enough data to ask a pretty uncomfortable question: is this money actually buying what we think it's buying? Luna: Define 'what we think it's buying' — national security? Jobs? Technological leadership? Lucas: All of the above. The stated goals were to reduce reliance on Asian fabs, secure supply chains for defense and critical infrastructure, and create tens of thousands of high-paying construction and manufacturing jobs. The CHIPS Act allocated about $52 billion in direct grants and loan guarantees, plus a 25 percent investment tax credit for capital expenditures on chip-making equipment. Luna: And the Commerce Department has been handing out the money in tranches — Intel got up to $8.5 billion, TSMC got $6.6 billion, Samsung got about $6.4 billion. Lucas: Right. Those are the headline numbers. But when you drill into the per-job cost, the picture gets murky. A December 2025 report from the Government Accountability Office estimated that the direct grants alone work out to roughly $200,000 per projected job. When you layer in the tax credit, we're talking closer to $300,000 per job. Luna: That's more than three times the average annual salary for a semiconductor technician, which is around $80,000. Lucas: Which raises the question: could the government have achieved the same employment outcome at lower cost? Maybe. But defenders of the program would say you're missing the point — it's not just about jobs, it's about rebuilding a strategic industry. Luna: Let's talk about that strategic argument. The CHIPS Act was sold as a national security necessity. In 2021, 75 percent of advanced logic chips were made in Taiwan, and the Pentagon was rightly nervous about a potential blockade. Lucas: So the idea was to create 'secure' domestic fabrication capacity. But here's the nuance — most of the CHIPS Act subsidies are going to foreign-owned companies like TSMC and Samsung. TSMC's Arizona plant will be owned by a Taiwanese corporation. Samsung's Texas facility is Korean. The intellectual property and key decision-making remain offshore. Luna: So we're effectively paying foreign firms to build factories on U.S. soil, but we don't control the technology inside them. Lucas: That's the tension. The Commerce Department has inserted 'guardrails' — companies can't expand fabrication capacity in China using the money, and they must share certain financial data. But critics argue those guardrails are weak. A 2024 report from the Center for Strategic and International Studies found that only 15 percent of the total grants had explicit requirements for domestic R&D investment. Luna: And what about the timeline? The original CHIPS Act envisioned fabs coming online by 2024 or 2025. We're now in May 2026 and only a handful of facilities are producing at volume. Lucas: Delays have been endemic. Intel's Ohio megafab, which was supposed to break ground in 2023, didn't start construction until late 2024, and production isn't expected until 2028. TSMC's Arizona plant faced labor shortages and permitting issues — its first batch of chips was delayed by about a year, into late 2025. Luna: So the government is spending billions on facilities that aren't producing yet, while the industry continues to rely on existing fabs in Taiwan and South Korea. Lucas: And the market hasn't stood still. During the delay, TSMC and Samsung both announced massive expansions in their home countries — partly funded by their own governments. Taiwan's own chip subsidy program is around $15 billion, and South Korea's is about $20 billion. So we're in a global subsidy race, and the U.S. is just one participant. Luna: Which brings us to the fiscal cost-benefit: is $52 billion in grants plus a tax credit that the Treasury estimates will cost $100 billion over ten years a good deal for taxpayers? Lucas: Let's run the numbers. The Congressional Budget Office estimated in early 2026 that the total federal cost of the CHIPS Act — grants, loans, and tax credits — will be roughly $150 billion over a decade. By comparison, the entire Department of Energy budget in 2025 was about $50 billion. So this is a significant fiscal commitment. Luna: And the projected benefits are often stated in terms of GDP growth or tax revenue from new factories. The Semiconductor Industry Association claims every dollar of chip subsidies generates $24 in economic output. That multiplier seems aggressive. Lucas: It is. Academic studies of similar industrial policies — like the 2009 stimulus for green manufacturing — found multipliers closer to 2 or 3. The $24 figure assumes the subsidies change long-run productivity, which is speculative. More modest estimates suggest the program might break even if the new fabs generate enough corporate tax revenue over their lifetimes. But that's a 20- to 30-year horizon. Luna: And there's an opportunity cost. That $150 billion could have funded universal pre-K for a decade or doubled NIH research spending. Lucas: Exactly. The real debate isn't whether chip manufacturing is a good thing — it's whether the government should be picking winners with this level of spending, or whether tax reform, regulatory streamlining, and immigration policy for skilled engineers would have been more cost-effective. Luna: What about the state-level competition? Places like Ohio, Arizona, and Texas gave additional incentives — tax abatements, infrastructure improvements, training grants. Intel got a $2 billion incentives package from Ohio alone. Lucas: And that's a whole other fiscal layer. A 2025 study from the Tax Foundation found that state and local incentives for chip fabs averaged about $150,000 per job, on top of federal subsidies. So the total government cost per job could be pushing half a million dollars. Luna: At that price, the government could literally pay each worker their full salary for five years and still come out ahead. Lucas: Right. But supporters would argue you're buying an ecosystem — not just jobs, but supplier networks, R&D spillovers, and a skilled workforce that attracts other industries. The question is whether those spillovers materialize. Early evidence from Arizona suggests some suppliers are setting up nearby, but it's too soon to tell if it's transformative. Luna: So where does this leave us in May 2026? The CHIPS Act is still in its early innings. Most of the money has been allocated but not fully spent. Lucas: The Commerce Department has obligated about 70 percent of the direct grants. Actual disbursements are slower — maybe 20 percent has hit company bank accounts. The tax credit is being claimed as companies file returns, but the full revenue impact won't be clear until projects are complete. Luna: And the political landscape has shifted. The current administration has talked about 'reshoring' and 'strategic sovereignty,' but there's growing bipartisan skepticism about the cost. Some members of Congress are pushing for stricter clawback provisions if companies don't meet job targets. Lucas: Which gets to the core of fiscal policy evaluation: how do you measure success? If the CHIPS Act leads to a secure domestic supply of advanced chips by 2030, but at a cost of $200 billion, is that worth it? That's a political judgment, not an economic one. Luna: And it's a judgment that other countries are making too. The EU's Chips Act is about $47 billion. Japan is spending $15 billion. China is pouring tens of billions into its own industry. The U.S. isn't alone in this race. Lucas: So the real lesson might be that industrial policy is inherently expensive and uncertain. The CHIPS Act may end up being a reasonable insurance policy against supply disruptions, but it's not a guaranteed return on investment. And for the fiscal watchdog, that's a hard pill to swallow. Luna: Insurance policies are never a bargain until you need them. I guess we'll know more in five years. Lucas: Yeah. And by then, we'll have spent another few hundred billion on the next industrial policy — maybe for batteries, maybe for clean energy. The pattern keeps repeating.