To date, the CHIPS program has achieved many of its objectives: the United States added more computer and electronics manufacturing construction in 2024 alone than in the two decades prior to passage. But success does not shield a program from redesign, and the Trump administration has made some significant changes, including converting Intel’s grant into an equity stake and shuttering Natcast, the operator of the National Semiconductor Technology Center. This instability has led some to suggest that future industrial policy programs should rely only on tax credits rather than discretionary grants because tax credits are not as vulnerable to post hoc adjustment1
There are real downsides to discretionary industrial policy, including political vulnerability, corruption, and bad decisions earnestly made. But it would be a mistake to abandon the model, which has capabilities that tax credits lack: it builds state capacity and expertise, provides the flexibility needed to design the right approach for new challenges, creates leverage that helps maximize the impact of each dollar, and makes success someone’s job.
A discretionary funding program builds expertise
The range of tasks associated with making discretionary investments — determining strategic priorities, deciding which projects and companies merit funding, and negotiating with industry — demands deep knowledge of the industry and commercial wherewithal. The CHIPS Program Office ultimately brought on roughly 180 people with deep semiconductor, investment, national-security, workforce, environmental permitting, and operational expertise. This was a high-caliber team, with expertise that did not exist anywhere else in government.
Tax credits demand a different skill set. Congress decides which types of projects should be supported, and much of Treasury’s job is policing edge cases in a way that is faithful to Congress’s direction while protecting taxpayer resources. The policy function of Treasury’s Office of Tax Policy is staffed primarily by lawyers who are skilled at making these tradeoffs across a range of issue areas; it would be impossible for Treasury to build a deep bench of industry-specific expertise across all the domains that tax credits touch. And by design, Treasury is responsible for crafting eligibility criteria rather than picking the ultimate recipients of tax credits. There are a lot of merits to this model, but it’s unlikely to compel industry experts to join the government.
CHIPS included both a semiconductor investment tax credit and discretionary funding, and the expertise Commerce attracted into government ended up helping Treasury’s implementation of the tax credit too. Treasury had to define a wide range of technical terms about semiconductor fabrication, packaging, and equipment production, while also helping make determinations about which semiconductor activities in China introduced national security risk (which would render a company ineligible for the tax credit). While the final decisions remained Treasury’s, Commerce was happy to lend staff to assist, and Treasury happily borrowed them; without a discretionary funding program, there wouldn’t have been staff to borrow.
One particular decision reveals the importance of industry expertise. In the law, Congress had limited the tax credit to only apply to projects which began construction before the end of 2026. The IRS has standard tests to determine when construction begins and allows for continued eligibility as long as construction is “continuous.” In the case of chip manufacturing, it’s common for chipmakers to build multiple fabs on one campus. Whether the credit’s rules would treat that standard practice as “continuous construction” would determine the power of the tax incentive. One reading would have been to limit the “continuous construction” provision to construction of a single fab. For CHIPS, that made little sense — multiple fabs on a single site often operate as an integrated whole, with shared tooling and an automated transportation system that carries wafers around the site. Moreover, Congress had recently declared building US fabs a national security priority, and building multiple fabs on a single site would meaningfully reduce costs and therefore incentivize chipmakers to continue (and even expand) their US operations after disbursement of all CHIPS program incentives. The CHIPS team’s expertise helped establish the right terms for eligibility, and the final regulations included a safe harbor for continuous construction that extended to multiple fabs on a single site.
Policymakers crafting industrial policy initiatives should weigh whether success will require expertise and if the program model they pursue is equipped to bring that talent on.
Discretionary funding is flexible, enabling prioritization
In the CHIPS statute, Congress charged the Commerce Department with advancing “the economic and national security interests of the United States,” but how it went about doing so was largely up to the Department. That left the CHIPS Program Office with dozens of decisions about how to concentrate resources, including which kinds of fabs to subsidize, how to balance allocations between foreign and domestic firms, which suppliers should share in the awards, and how much support each project should get, to name just a few. By contrast, the CHIPS tax credit is straightforward: anyone who builds, expands, or modernizes a factory for semiconductors or semiconductor tools is eligible for a flat 25% tax credit, later bumped to 35% by the One Big Beautiful Bill Act.
The predictability of a tax credit makes it powerful in driving investment, and CHIPS would not have been as successful without the tax credit as a baseline incentive. But the tax credit also comes with limitations. For example, early on in the program, it was clear that advanced packaging would need to be a program priority. It wouldn’t matter if we attracted leading-edge fabs to the US if those chips still had to be packaged offshore. As a result, packaging became a central focus, and the CHIPS team actively encouraged applications in the space. Those efforts bore fruit when Amkor, one of a handful of independent chip packaging companies, applied for funding to build in Arizona. Although Amkor is headquartered in the US, almost all of Amkor’s production activities are located in Asia.
Advanced packaging facilities are very different from a typical fab and operate on much lower margins. To land Amkor’s investment, the CHIPS team offered a meaningfully higher incentive than its typical award — a reflection of both the project’s strategic importance and its economic fundamentals. The flexibility of the grant program allowed the government to attract an investment that the tax credit alone would not have been able to.
Tax credits can also be an expensive way to drive outcomes. Because a tax credit is indiscriminate in its application, it can apply to a broad range of projects, even lower priority ones or ones that would have been built anyway. Moreover, tax credits are rigid, and Congress rarely has a complete understanding of how to prioritize projects to achieve strategic objectives at the time it drafts a law. This isn’t a criticism — it took the CHIPS team months of dedicated work to put out its first proposal for how to allocate grants, and that proposal changed as the team learned more through direct negotiations with companies. But tax credits rarely have that flexibility. Absent amendment, they can’t adapt to new developments or information.
Discretionary funding creates leverage
The Amkor story points to another feature of discretionary industrial policy: it gives the government leverage.
In total, the four leading-edge applicants alone requested over $70 billion of our available $39 billion of funding. Commerce used that imbalance to press companies for better outcomes, pushing for more investment, faster construction timelines, onshoring of R&D activities, and bringing production of more advanced nodes to the US. The appropriate use of leverage depended on the Department’s evaluation of national security goals and the company’s appetite. In the case of Amkor, our pressure on chipmakers — Amkor’s customers — to offer packaging solutions led them to engage Amkor as a potential partner, which motivated the company to propose a project in the US.
Flexibility also let Commerce set the milestones necessary to ensure projects actually advanced the national security interests they were designed to. As Todd Fisher wrote, in some cases the team negotiated commitments for days of inventory on-hand or required US-based R&D spending to help build up the domestic semiconductor ecosystem.
A discretionary funding program makes success someone’s job
At any given moment, the government is enforcing dozens of policies which will affect the success of an industry. With semiconductors, for example, the Labor Department decides which training programs to prioritize, and advises states on how to most effectively develop a skilled and trained workforce, both for construction and fab operations; the State Department decides whether — and how easily — foreign companies can temporarily bring their overseas staff to the United States; and the Environmental Protection Agency reviews many new chemicals needed for semiconductor fabrication under the Toxic Substances Control Act, for which the review backlog can be years long. Projects often need permits from the Army Corps of Engineers because they affect US waterways, which the Corps reviews under the Clean Water Act. And under the Endangered Species Act, actions that could affect endangered species must be run past the Fish and Wildlife Service. These agencies each have their own mission and mandates unrelated to the industrial policy initiative. Even in the best case scenario, it’s hard to coordinate these policies while still advancing program objectives.
Discretionary grant programs designate a single person whose entire job is to solve the challenge at hand. That leader acts as a general manager and is accountable for the whole of the problem — their professional reputation rests on the outcome. Identifying such a point person to fight for success in interagency disputes is critical. Tax credits do not assign similar leadership.
The CHIPS Act made Secretary Raimondo personally responsible for delivering US fab construction. At various points, she and our team worked on accelerating environmental reviews, unblocking permitting challenges, creating customer demand for US chips, training workers to build and run US fabs, and aligning subsidy programs in other countries. Because Commerce was accountable for outcomes, it took an active role in shaping interagency discussions. Not all issues were resolved in Commerce’s favor, which may well be appropriate given the range of substantive issues at stake, but Commerce’s institutional perspective made a big difference.
Commerce was also accountable for issues for which there was no agency in charge. For example, the CHIPS Act was accompanied by similar programs in the European Union, South Korea, Taiwan, Japan, and, with a somewhat different structure, China. It’s easy to forget now, in the midst of the AI boom, but there was real concern that countries would oversubsidize the industry and build too many fabs. No agency was specifically charged with solving this problem, so it became ours.
Another advantage of having a clear leader was that Secretary Raimondo’s ownership was legible to the semiconductor companies, giving CEOs both a single person to negotiate with and a counterparty who could help remove blockers on important projects.
Successful industrial policy calls for both tax credits and discretionary funding
Most industrial policy efforts will benefit from both discretionary funding and tax credits.
Tax credits avoid implementation challenges like hiring and environmental review and are better insulated from political capture or corruption. Per the Clean Energy Tax Blueprint report on the Inflation Reduction Act, tax credits can’t solve every problem, but they do work well when the problem is price, not capability, and when the market is sufficiently mature to respond. That description captures a lot of markets, including many future candidates for industrial policy. And IFP’s Saif Khan points to two further advantages of credits: they’re enshrined in law, helping to stabilize policy and deter changes by a future administration (though the One Big Beautiful Bill Act rollback of IRA tax credits somewhat cuts against this), and they’re easier for a capacity-strapped government to execute.
Discretionary programs also carry real downside risks. Shifting policy priorities, particularly around a presidential election, diminishes the ability to attract private investment, because the government can’t credibly commit to long-term policies. When poorly run, discretionary programs can be captured by industry or become a tool for favoritism or even corruption. Moreover, discretionary programs are fundamentally a bigger delegation to the executive branch, and so depend more heavily on the quality of the team there.
But it would be a huge mistake to jettison discretionary funding entirely. Major industrial policy programs have better odds of success with the state capacity and flexibility that comes from discretionary funding. This is doubly true in a sector like semiconductors that requires a lot of technical capability where the government lacks up-to-date market information. A tax credit alone is unlikely to have the requisite firepower.
Despite all the risks, the affordances created by discretionary programs make them a necessary part of future industrial policy.
Sam Marullo is a Senior Advisor at IFP. He previously served as the first Policy Director for the CHIPS Program Office and as Counselor to the Secretary of Commerce.
One useful mental model for “discretionary funding” is federal grants, where organizations apply for funding and the federal government picks the winners (and often the grant amount as well), but there are a variety of creative funding mechanisms that suit different purposes. In this piece, discretionary funding refers to not just grants, but the full set of federal financial assistance tools, which also includes cooperative agreements, loans, and other transaction agreements.
