In June 2021, US Census Bureau data recorded that private manufacturing construction in the United States was being put in place at an annualised rate of roughly $79 billion. By June 2024, that figure had reached $236 billion — a tripling in three years, without precedent in the modern record of American manufacturing investment. The acceleration was not evenly distributed across industries. It was overwhelmingly concentrated in computer, electronics, and electrical manufacturing — the specific sectors targeted by the CHIPS and Science Act and the Inflation Reduction Act, both of which became law in August 2022. By the Biden administration’s accounting in January 2025, the two laws, together with the earlier Infrastructure Investment and Jobs Act, had catalysed roughly one trillion dollars of announced private investment, including $449 billion in electronics and semiconductors, $215 billion in clean power, $184 billion in electric vehicles and batteries, and $93 billion in clean energy manufacturing and infrastructure.
Nearly four years after enactment, the record on what that legislation has and has not produced is substantial enough to read carefully. The reading is more interesting than either the policy’s defenders or its critics have generally allowed. The laws did a real thing, measurable in concrete poured and contracts signed. They did not, cleanly, do the thing they were most often described as doing. The gap between those two propositions is the subject of this piece.
WHAT THE POLICIES DID, AND DID NOT, DO
The CHIPS and Science Act authorised $52.7 billion in direct funding, including $39 billion of manufacturing incentives administered through the Department of Commerce, alongside a twenty-five per cent investment tax credit on new semiconductor manufacturing facilities. The Inflation Reduction Act allocated roughly $369 billion over a decade in tax credits, loans, and grants targeted at clean-energy supply and demand — production credits for solar, wind, battery, and hydrogen manufacturers; demand-side credits for purchasers of electric vehicles and clean energy systems. The combined direct fiscal commitment, by the Congressional Budget Office’s scoring, sat in the low hundreds of billions of dollars over ten years.
The observed private investment response vastly exceeded the direct fiscal commitment. By the Semiconductor Industry Association’s December 2025 accounting, the CHIPS Act alone had been associated with roughly $630 billion of private investment across 140 projects in twenty-eight states, with a predicted 500,000 jobs in construction and operation. Policy analysts tracking IRA-linked clean-energy investment reached similar orders of magnitude. The ratio of private investment to direct public subsidy was, by most accountings, on the order of five to ten to one — an unusually high leverage ratio for industrial policy, attributable in part to the tax-credit structure and in part to the underlying demand conditions the policies were designed around.
This is the real thing the policies did. They produced, at observable scale, a concentration of private manufacturing investment in the specific sectors the legislation identified. No sensible reading of the data disputes that.
WHAT THE EVIDENCE SUPPORTS, AND WHAT IT QUALIFIES
Two important qualifications follow.
The first is methodological. The Congressional Budget Office’s May 2022 projection of private non-residential fixed investment growth between Q3 2022 and Q4 2023 — a baseline produced just months before the laws were enacted — was 4.0 per cent. The actual growth rate over that period was 3.9 per cent. Analysts at the Tax Foundation pointed out in early 2024 that this counterfactual, imperfect though it is, does not suggest that CHIPS and the IRA increased aggregate investment — only that they redirected it. Interest rates rose faster than CBO had projected, putting downward pressure on investment in sectors not targeted by the policies. Oil prices rose, pushing upward on energy investment. The sectoral composition of investment shifted dramatically toward computer, electronics, and electrical manufacturing; the aggregate roughly tracked the pre-policy baseline.
This reading is not unique to the sceptical analysts. The Federal Reserve Bank of Boston’s November 2024 analysis, which is broadly favourable to the policies’ employment effects, estimated that CHIPS-and-IRA-related investment would create roughly 6,000 additional jobs monthly in 2024 and 4,000 monthly in 2025 — equivalent to one to two per cent of the national monthly average of new jobs in 2023. The effect is real and measurable. It is not, at aggregate level, transformational.
The policies did a real thing, measurable in concrete poured and contracts signed. They did not, cleanly, do the thing they were most often described as doing. The gap between those two propositions is the subject of this piece.
The second qualification is operational. The announced facilities have been slower to build than the announcements suggested. TSMC’s Arizona fab, the largest single investment in the CHIPS Act cohort at roughly $65 billion, was initially scheduled for first-production ramp in 2024. Labour shortages — specifically, a shortage of construction and installation technicians with the specialised skills required for semiconductor-grade cleanroom installation — pushed first production into 2025, and the second fab from 2026 to 2027. This pattern is not unique to TSMC. The SIA’s own analysis anticipates a US shortfall of 90,000 semiconductor technicians and engineers by 2030 if current training levels persist. The capital commitments have been made. The human infrastructure required to operate them has not kept pace.
A similar pattern, with different specifics, is visible in the clean-energy side of the IRA portfolio. The announced battery-plant and solar-manufacturing pipeline is large. Execution has been concentrated in specific regions, predominantly in the South and Southwest, where incentive structures and labour markets have been most favourable. Several announced projects have been paused, rescoped, or cancelled, particularly where Chinese solar overcapacity has undercut the domestic economics that the IRA credits were designed to support. These cancellations have been a minority of the announced pipeline, but they are real, and they illustrate a point that tends to get lost in headline accounting: a private investment announcement is not an operating facility, and the distance between the two is longer than the political news cycle accommodates.
WHAT THE RECORD ALREADY SHOWS
Four years is a short horizon for assessing industrial policy. The twenty-year verdict is what will matter, and the twenty-year verdict is a different question — one about whether the facilities currently under construction become operating assets with enduring economic activity, whether the workforce required to operate them materialises, whether downstream demand holds up through the next several business cycles, and whether the regulatory and fiscal environment remains stable enough for the long-lived capital assets to earn their returns.
What the four-year record already shows, though, is substantial. The United States has concentrated a remarkable amount of private capital into a specific class of industrial asset — semiconductor fabs, battery plants, EV factories, solar-module and clean-energy manufacturing lines — in a compressed span of time. This concentration has altered the geographic distribution of new American industrial employment, with disproportionate gains in states whose prior manufacturing base had eroded. It has produced a physical industrial build-out that did not exist before, and whose assets will outlive the political coalition that created them. These are not trivial outcomes. They are the outcomes the policies were designed to produce.
What the record does not show, and what the policies were also often described as producing, is a general increase in American industrial capacity or in aggregate private investment. The sectoral concentration came at the cost of investment elsewhere in the economy — investment in traditional manufacturing, in services, in the broader housing and commercial-construction base, all of which face higher borrowing costs partly because of the same interest-rate environment the policies were launched into. This is not a failure of the policies. It is a feature of their design: they targeted specific sectors, and they got specific-sector investment. The broader claim — that industrial policy was a general expansion of American industrial capacity — is not what the evidence shows.
WHAT THE NEXT STAGE WILL DETERMINE
The 2025 change of administration has introduced a further variable. The second Trump administration has kept the CHIPS Act’s basic structure intact — the semiconductor agenda retains unusual bipartisan support, and cancellation of signed awards would have capital-markets implications the administration has not chosen to incur — but has selectively wound down or restructured specific Biden-era programmes, most visibly the Natcast semiconductor research consortium. The IRA’s clean-energy provisions have faced greater uncertainty; specific tax credits have been narrowed or limited, and private-sector planning has adjusted accordingly. The political durability of the 2022 legislation was always going to be tested across subsequent administrations; the test is now in progress.
The reading that holds up across the four-year record, and that will probably hold up across the longer horizon, is narrower than either the policies’ advocates or their critics have generally advanced. Industrial policy of the CHIPS-and-IRA kind can, demonstrably, move large amounts of private capital into specific sectors on short time horizons. What it cannot, demonstrably, do is expand the aggregate supply of the factors — workers, training infrastructure, capital at scale in unfavoured sectors — that determine whether the concentrated investment eventually earns its cost. That second step is a multi-decade project, and it does not move on the timelines of legislation. The record four years in is that the policies performed their first task well. The second task, which is the task on which the long-term verdict actually depends, is the task the record does not yet address.
That is the honest summary, and it is the summary likely to be cited ten years from now. The concrete is poured. The buildings are rising. Whether they become operating assets that produce enduring economic value — or whether a meaningful share become write-downs in the next cycle — will be settled by decisions that have not yet been made, by workers who have not yet been trained, and by demand conditions that have not yet formed. The legislation has performed the part of its work that legislation can perform. The part that legislation cannot perform is, as always, the decisive part.
PRIMARY SOURCES
— US Census Bureau. Construction Spending: Manufacturing. Monthly time series of value of private manufacturing construction put in place. https://www.census.gov/construction/c30/historical_data.html
— Nahm J. Manufacturing Gains from Green Energy and Semiconductor Spending since the CHIPS and Inflation Reduction Acts. Federal Reserve Bank of Boston, Current Policy Perspectives, November 2024. https://www.bostonfed.org/publications/current-policy-perspectives/2024/manufacturing-gains-from-green-energy-and-semiconductor-spending.aspx
— Semiconductor Industry Association. 2025 State of the U.S. Semiconductor Industry Report. SIA, 2025. https://www.semiconductors.org/state-of-the-u-s-semiconductor-industry/
— Tax Foundation. Supply-Side Economics vs. Industrial Policy: TCJA, IRA, CHIPS Act. March 2024. (Counterfactual analysis of aggregate investment effects.) https://taxfoundation.org/research/all/federal/supply-side-economics-industrial-policy/
