How CHIPS Executed on an Ambitious but Vague Mandate

Inside the implementation decisions that mattered most

“Victorious warriors win first, and then go to war.” Sun Tzu

Policy work sits at the top of Washington’s prestige hierarchy. Shaping ideas and legislation is held in high esteem; by contrast, deciding how to actually implement a vision is undervalued. And execution talent — people who can convert policy into real programs — is in short supply.

To build the CHIPS program, we first needed to translate our statutory mandate into a coherent, implementable strategy. Intentional program design wasn’t the entirety of execution, but success would have been impossible without it.

The CHIPS Act did not specify execution strategy

The CHIPS Act was signed into law on August 9, 2022. When Mike and I joined the team right after Labor Day, we were still a ramshackle start-up troupe of mostly “detailed” talent borrowed from other bureaus or agencies. And we had $39 billion burning a hole in our pocket, with high expectations from both the White House and one of the most sophisticated and politically savvy industries in the world.

Section 9902 of the legislation offered a mandate and a high-level mission, instructing the Secretary of Commerce to create a financial incentive program to strengthen national security by building domestic semiconductor capacity. The specifics — strategy, segmentation, prioritization, and sizing — were left almost entirely to the implementing team to determine. For example:

• CHIPS authorized $39 billion across grants, cooperative agreements, other transactions, loans, and loan guarantees (up to $6 billion of which could be used as subsidy cost to support up to $75 billion of loans and loan guarantees) — but it did not specify the balance of each or when a particular tool or strategy might be better suited to the task.

• The legislation identified eligible uses of funds (e.g., construction, expansion, or modernization of commercial semiconductor fabs and related supply chain facilities) and required at least $2 billion to be allocated to legacy or mature node capacity essential to national security, but did not recommend specific allocations and prioritization beyond that (i.e., how much should go to leading edge logic, and how to prioritize funding memory vs. logic, or supply chain vs. fabs).

• CHIPS laid out broad prioritization criteria — projects needed to strengthen supply chains, advance national security, promote economic competitiveness, and demonstrate financial viability — but did not provide further selection logic or sizing direction.

Our job was to turn this enabling legislation into an executable industrial policy program. There were hundreds of design, process, and organizational structure decisions to make: What should we do first? How will we size individual awards? Should we prioritize loans? What is most important for national security (and how do we even define that)? Building a capable program required quickly identifying which decisions would prove most significant.

Deciding how to execute

Four early decisions shaped how we executed the CHIPS program and ultimately set us up for success. Rather than a reactive, “let’s see what comes in” approach, we set our intentions and expectations early and codified them in our Notice of Funding Opportunity (NOFO).

1. Using rolling applications

One of the first design decisions we had to make was whether to create one universal deadline for all applications or to evaluate them on a rolling basis. Many grant programs choose the former — it’s easier to organize a review process around a fixed date, compare all applications against each other, and announce all awards together.

Program statute often sets a short-term date by which all money must be allocated, making the universal deadline a more efficient option. But this model gives an edge to the most organized and government-savvy companies and to projects that are furthest along in development (which would likely materialize even without our funding); it’s harder for nascent ideas and technologies to compete with more detailed, mature ideas in this process. A batched rolling process also constrains resources and timelines, making it challenging to work with applicants to shape projects.

CHIPS did not define a short-term target date. This was to our benefit — we knew that the industry would evolve over the coming years (ChatGPT was not even launched when the legislation was passed) and that many important industry players — especially foreign companies — were more likely to come to the table over time rather than respond to a tight timeline. We wanted to allow and encourage companies to apply as soon as they were ready. We also wanted the ability to proactively shape the end result — rather than fund what was already in a company’s plan, we wanted to push them to pursue more ambitious projects, which would take more time to develop. And we wanted to leave room for emerging technologies, such advanced packaging and glass substrates, which we expected would need additional time to translate into well-defined, fundable projects. Rolling applications just made more sense.

But a rolling process creates its own set of challenges. We needed to manage resources without set timelines and to equip our team to proactively push reluctant companies. And, most critically, we had to make sound decisions without the full picture of what might come in.

2. Proactively slicing up the pie

$39 billion sounds like a lot of money — until you put it into context. TSMC alone was expected to spend over $40 billion in capital expenditures in 2025. Global semiconductor revenues in 2024 exceeded $600 billion; industry capital expenditures were over $150 billion. We were focused on funding capacity buildout through 2030; $39 billion spread over those six years is merely 1% of current global annual revenue and 4% of global capex. We clearly wouldn’t be able to do everything we wanted to, and we needed our dollars to go as far as possible.

Opting for a rolling application process complicated things. With a fixed application deadline, you can review the entire set of applications and select the strongest ones, allocating funding accordingly. But the combination of a rolling program and limited capital in a highly dynamic industry meant we had to make high-stakes decisions without knowing our complete portfolio. Our four leading edge applicants alone requested $70+ billion in total funding — nearly twice our total budget. Without a plan we risked overallocating to early applications and not being able to fund the reluctant applicants or emerging technologies that came along later in the process.

To mitigate that risk and guide our outreach and decisions, we proactively outlined a target allocation of capital across the various industry segments. Doing so required documenting what we could realistically accomplish and what our ideal portfolio would look like. We broke this down further: How many leading edge fabs would we ideally fund? How much memory capacity? How many packaging facilities? How much for the upstream supply chain? Luckily, we had the investment tax credit at our disposal, which applied to all semiconductor projects and we thought would be sufficient incentive for some projects. We needed to focus our dollars on the critical projects that the tax credit alone would not catalyze.

We had articulated our high-level goals and priorities in our Vision for Success, but we needed much more specificity to truly guide our overall effort. For example, the document committed us to at least two new leading edge logic clusters, but we still needed to define what a cluster was, how many fabs it would consist of, and how much to invest while still leaving enough money for our other priorities.

Our strategy and investment teams dug into the economics of each subsegment, turning to outside consultants to gain deeper industry insight. We started with supply and demand — we broke down the industry by node, modeled all current and announced supply, and evaluated multiple detailed demand projections. We studied past industry cycles and grappled with questions about how AI could shift demand growth and how oversupply in China might impact different legacy nodes. This helped us determine how many new fabs would be needed, which made a clearer case for convincing companies to build them in the US instead of elsewhere.

But that was only the beginning. We then built cost models for different types of fabs and used proprietary industry data to estimate cost differentials with other regions. We had to understand the scale of investment needed to accomplish our detailed goals, what it would cost in the US compared to other geographies, and the returns that companies would require to commit to building in America. And these economics varied significantly by sector. A typical leading edge fab costs north of $20 billion; TSMC, the industry leader, generates operating margins over 45% company-wide. A mature node fab would cost about half of that amount, but with much lower margins (GlobalFoundries operating margins are in the mid-teens). Contrast those with advanced packaging facilities, which may require only $2 billion of investment, but where overall company margins have historically been in the single digits.

Quantifying the expected costs of the priorities outlined in our Vision for Success helped us determine funding needs across the ecosystem. In estimating the range of new leading edge fabs required to meet global demand through 2030, we determined that roughly 8-10 of them would need to be in the US to reach our target 20% market share.¹ While possible, it would require $200+ billion, so we needed to be judicious about how much of our funding to direct at that goal.

Our internal roadmap sharpened our sense of what success could look like and how much capital to allocate to each sector. It also allowed us to shape our applicant pipeline rather than react to it, and to evaluate applications with a clear goal in mind. We were not bound to hard targets — we knew we would have to adjust based on the actual applications we’d receive and how the industry evolved. But the roadmap clarified our goals and established discipline and focus. We iterated on our allocation targets as we received new information, reevaluating our initial analysis with the investment committee and the broader team and adjusting the portfolio accordingly.

3. Sizing awards

With our target portfolio allocation established, we had to decide how to size individual awards and what to communicate upfront. We believed that setting expectations early would boost our negotiating leverage and help anchor our approach to applicant engagement.

Our mantra on sizing awards was simple: give applicants just enough subsidy — and not a penny more. We wanted to catalyze private investment, not displace it. And we needed a consistent framework by which to evaluate applications.

First, we established a strategy for award sizing. We borrowed a page from corporate finance and private equity, implementing the internal rate of return (IRR) vs. hurdle rate methodology taught in every finance course in the world. This strategy — which many companies and investors use to make investment decisions — compares long-term annualized returns on invested capital against a company’s historical cost of capital. This calculus was familiar to companies and grounded our negotiations in numbers rather than anecdotes or vibes. It wasn’t a pure science — factors like national security importance, a company’s other options, and project risk all factored in — but it created a strong starting point. As the NOFO put it:

“…the degree to which the request for CHIPS Incentives is reasonable and necessary to make the project viable … [will be] based on cash flow modeling, IRR analysis, sensitivity analysis, and other applicable analyses, and whether the projected IRR in the cash flow model is reasonable based on the expected risks and returns of the project, historical projects of similar nature, or other relevant market benchmarks.”

We didn’t set a specific IRR target. Instead, we crafted a framework aimed at producing a return attractive enough for firms to build in the US. We paired it with another critical clause:

“Most direct funding awards are generally expected to range between 5-15% of project capital expenditures.”

Fifteen words in a 129-page, 75,000-word document — yet perhaps the most consequential and risky. We could have chosen to omit this range; there was a chance that these relatively low percentages might discourage companies, especially those on the fence. And what if we couldn’t hold the line in negotiations — were we setting ourselves up to fail?

But we didn’t pluck that range from thin air; our team — including our excellent Chief Risk Officer, Andy Kuritzkes, and a team from Oliver Wyman — conducted a detailed analysis to determine what “just enough” subsidy to generate attractive returns might be. To establish a realistic range, they modeled sample returns for different types of projects based on historical project investment returns, global cost differentials, and detailed operational models at different subsidy levels.

Our analysis affirmed that the 25% tax credit was a meaningful baseline incentive and, in some cases, sufficient to catalyze investment (especially when paired with state or local incentives). Our funds would therefore be better used for important projects that required an incremental push. The 5-15% range also meant that public support would still be below half of total capex (25% tax credit, plus our 5-15%, plus state and local contributions, which are typically quite modest). The companies themselves and other private capital would be the majority investors with real skin in the game, aligning their incentives with ours. And since we tied support to upfront capex and not ongoing operational costs, companies were incentivized to find creative ways to shrink ongoing costs over time. Lastly, establishing a range also conceded that financial viability varied across subsectors; an advanced packaging facility with thinner margins would likely require a higher percentage of support than a leading edge fab.

We could have forgone the complicated IRR methodology and set a consistent target percentage for all awards. This would have considerably simplified our process and created less oversight risk, but at the expense of flexibility and the ability to shape ultimate outcomes through differentiated incentives. Establishing and communicating our approach early set expectations and bounded negotiations. And it positioned us as a serious counterparty grounded in analytical rigor, redirecting the conversation from the government affairs teams of these companies to their operations and finance executives (a subtle, but important distinction).

Defining our funding range also helped with our portfolio math — we knew that if we funded at our 10% midpoint on average, our $39 billion could catalyze nearly $400 billion in total investment. Importantly, it also created an analytical framework and discipline for our internal teams and investment committee that proved valuable when debating and defending sizing decisions.

4. Closing information asymmetry

I had a lot of anxiety about taking on this job. We were handicapped by enormous disadvantages in negotiations: the companies on the other side of the table knew far more about their cost structures, had armies of well-paid analysts, could hire the best advisors, and had substantial political influence. How were we not going to get eaten alive?

We needed to level the playing field. To signal our seriousness, we hired top talent. We also developed our understanding of the industry through iterative applicant engagement and built trusting and open relationships with customers and investors. And we borrowed another page from private equity, striving to align interests and force transparency.

Our sizing approach played a part here, since it required companies to disclose data and contribute their own considerable investment. Even more significant, though, was our decision to add an upside-sharing provision to the NOFO:

“Recipients of more than $150 million in direct funding must share with the U.S. Government a portion of any cash flows or returns that exceed the applicant’s projections above an established threshold. The terms will be set on a case-by-case basis and are intended to incentivize accurate projections and ensure that taxpayers benefit from outsized financial returns” [emphasis added].

This clause was not required and drew significant industry criticism.² And it sparked intense debate within our team and with the White House — did we need to complicate the program if this was truly about national security outcomes?

While our upside-sharing mechanism was designed to benefit taxpayers, we didn’t expect to get much money back. The clause’s biggest near-term benefit was that it reduced information asymmetries and incentivized companies to provide accurate projections. Upside sharing discouraged companies from gaming the system by giving us a conservative financial case to justify more subsidy — outsized profit would just come back to the government.

Our award sizing methodology undoubtedly created complexity, but it made clear that this was not your typical government grant program. Together, our sizing and upside-sharing clauses reduced the company’s edge and aligned their interests with ours.

CHIPS gave us discretion

The CHIPS legislation defined the program’s high level goals, but not how to go about achieving them. Our program design was geared toward keeping us on our front foot — effective execution boiled down to tactical early decisions that drove internal focus and gave us more leverage in negotiations. By establishing a proactive allocation framework, using rolling applications, and defining a transparent sizing and upside-sharing methodology, we set expectations for companies, provided clear direction to our team, and leveled power and information asymmetries.

These were carefully considered decisions on how to execute on a broad mandate. Our experience affirmed why the corporate world believes “execution eats strategy for breakfast” — government needs to internalize this maxim and place more emphasis and value on program design and execution.

1. While not outlined in the statute or our Vision for Success, Secretary Raimondo announced a goal of achieving 20% of market share by 2030 in early 2024.

2. The Trump administration has expanded on this model and taken direct equity stakes in projects. We detailed our thoughts on equity as a tool in an op-ed last year; we’ll have more to say on it later in this series.