What this covers
This is a conversation with Todd, who leads the office administering the CHIPS Act's manufacturing incentives, about how the money is actually being deployed. It begins with the problem: over 35 years, the U.S. share of global fab capacity fell from about 40% to about 10%, taking expertise and supplier ecosystems with it. The program totals over $50 billion, split between $39 billion in manufacturing incentives and $11 billion for R&D, with tax credits and up to $75 billion in loans and guarantees alongside. Todd describes an office of nearly 200 people, 40 to 50 of them investment professionals from finance and consulting, that seeks out projects and negotiates iteratively instead of simply processing applications.
The range runs from deal mechanics to industrial detail. Todd explains how government puts in 5% to 15% of project capital, a structure he ties to Private Equity Alignment of Interest. He also covers milestone-based disbursement and the demand evidence applicants must show, such as General Motors' long-term agreement with GlobalFoundries. Foreign firms qualify too, and he cites at least 14 TSMC suppliers and 24 Samsung suppliers planning U.S. operations. The scale of a leading-edge fab gets concrete treatment, with 600,000 cubic meters of concrete and cleanliness 10,000 times that of an operating room. Workforce is named as possibly the decisive factor of the decade.
The contested turns come late. Todd argues for building resilience now, like buying insurance in calm weather, though the industry is not supply constrained. He concedes many projects relocate production rather than add capacity. He also flags the risk that Chinese equipment spending oversupplies mature node chips. Success, he says, is a V-shaped recovery toward about 14% share by 2030 under SIA projections, versus about 8% unaided.
Todd argues the CHIPS Act should be run like a private equity investment, using milestone-based, co-invested awards to build a durable U.S. semiconductor ecosystem, with success judged by long-run capacity share rather than near-term returns.
- Government money is small relative to industry capex (about $39B against roughly $650B of U.S.-directed investment), so it must crowd in private capital and align incentives
- Awards are disbursed over construction and commercial milestones and require state/local incentives and customer demand evidence
- The real test is whether U.S. share of global capacity reverses its decline by around 2030
Awards must stay commercially viable through milestone payments and investor-style demand diligence, not subsidized overcapacity.
- Equipment spending in China, focused on current and mature technologies, could oversupply those nodes at certain points, creating an overcapacity risk that Todd considers part of the fundamental analysis.
“that current and mature industry has the potential, particularly at certain nodes to be oversupplied.”
- Funds are not paid out at signing but disbursed over construction and commercial milestones, such as completion of construction stages or qualification of equipment, which protects taxpayers and keeps projects commercially viable.
“It's not dollars out the door day one, it's dependent on milestones, construction progress, et cetera.”
- The program aims to avoid distorting markets by creating sustainable commercial opportunities rather than subsidized capacity that lacks economic viability.
“We're also not trying to distort a market. We're trying to create sustainable commercial opportunities across the board.”
- As part of diligence, the office seeks to understand the true underlying demand for each technology or fab, rather than accepting announced capacity at face value.
“We will, as part of diligence, understand what the true demand is, underlying a given technology or fab.”
- Awards include some flexibility so companies can adjust when projects do not play out as planned, because semiconductor markets are cyclical and demand forecasts are uncertain.
“We do work into our agreements, certain amount of flex to recognize that things are not going to play out exactly the way we plan”
- The office builds relationships with companies and their customers to understand demand signals, as a typical investor would, asking whether a project can be commercially viable over time.
“We have built relationships with their customers. We need to understand the demand signals in the market.”
- Todd evaluates projects as a typical investor would, asking whether they will be commercially viable over time and structuring awards around that, rather than seeking financial returns for the government.
“That's the way we approach it as a typical investor would. Will this be able to be commercially viable over time?”
- The program's design explicitly includes engagement with customers, investors, and other private capital sources, which Todd describes as atypical for government programs.
“that nature of being able to engage regularly with applicants engaging externally with other stakeholders”
- The CHIPS team evaluates long-term commercial viability by building relationships with companies and their customers and by understanding demand signals, then structures awards so viability is required.
“We need to understand the demand signals in the market. We need to understand what customers are asking for”
- Awards are sized by evaluating each project's economics and expected demand for the facility's output, reflecting the team's view that cyclical semiconductor demand must be forecast before committing capital.
“we try to size our awards appropriately, we understand what the returns on their projects are going to be”
Geographic concentration and fragile supply chains make onshore semiconductor resiliency urgent to build now.
- U.S. semiconductor supply became fragile because the industry's supply chain relies heavily on unique geographic choke points, particularly Taiwan, which created a lack of resiliency that was exposed during the pandemic.
“our supply chain is really reliant, to a large extent, on unique geographical choke points, particularly in places like Taiwan, but not only in places like Taiwan that created a lack of resiliency.”
- The semiconductor industry is cyclical despite high long-run growth trends, and segments like memory and current/mature chips have recently turned down, building inventory and easing supply constraints.
“the semiconductor industry is a cyclical industry despite high cyclical growth trends. And over the last couple of years, that industry, parts of that industry, memory, current and mature, have turned down and inventory has built.”
- Building semiconductor resiliency should happen during the current softer part of the cycle, just as one buys insurance when the weather is calm rather than during a storm, even though the industry is not currently supply constrained.
“that is exactly the time where we need to build the resiliency. It's like you don't want to be buying the insurance and putting a new roof on and hurricane shutters on when the wind is blowing, you want to do that during the sunny times.”
- The pandemic showed the fragility of chip supply when cars could not be completed because of cheap components costing around $0.50 to $2, such as microprocessors that control windshield wipers.
“you couldn't get your car because of some $0.50 or $2 microprocessor or some other type of chip that dealt with window windshield wipers on a car.”
- Over the last 35 years, the U.S. share of global semiconductor fabrication capacity fell from about 40% to about 10%, eroding the expertise and ecosystem needed to build fabs.
“we've gone from about 40% of the world's semiconductor fabrication capacity to, based on the latest SIA report, about 10%.”
- A single ASML EUV lithography machine contains a huge number of component parts drawn from many first-, second- and third-tier suppliers, making the semiconductor supply chain extremely complicated to rebuild onshore.
“If you look at one ASML, EUV machine, I mean, the number of component parts in that system itself and the number of first and second and third-tier suppliers”
- The AI boom creates an opportunity to build the entire AI semiconductor ecosystem onshore in the U.S., which Todd calls a major national and economic security priority.
“there is a real opportunity that we can help on the back of AI create the AI ecosystem from soup to nuts in the U.S.”
- All three leading-edge logic companies have committed to their versions of 2-nanometer production in the U.S., and Micron plans leading-edge high-bandwidth memory here, even though the U.S. produces 0% of high-bandwidth memory today.
“we have all three of the leading edge logic companies committed to their version of 2 nanometer”
- Many CHIPS-funded projects shift production from one region to the U.S. rather than adding incremental capacity, since companies like TSMC would otherwise have built most planned fabs in Taiwan.
“In many cases, it's shifting, something that might have been done in one region for another.”
The CHIPS program operates as a private-equity-style investment office aligning company and government interests.
- The CHIPS Act provides over $50 billion, of which $39 billion is manufacturing incentives administered by Todd's office and $11 billion is R&D funding, supplemented by investment tax credits, loans and loan guarantees up to $75 billion, and state and local incentives.
“CHIPS Act was $50-plus billion. Of that, $39 billion is manufacturing incentives, that's where I sit, and then there's $11 billion that's R&D.”
- Direct CHIPS incentives are expected to cover roughly 5% to 15% of a project's overall capital expenditure, as an expectation rather than a hard rule.
“those incentives people should expect will be approximately 5% to 15% of the overall capital expenditure spent on a project. That is not an absolute, hard-and-fast rule”
- Even $39 billion is not a large amount of capital in this industry, since a single leading-edge fab can cost $20 to $25 billion, and roughly $650 billion of projected investment (28% of about $2.3 trillion over a decade) is expected in the U.S.
“a single leading-edge fab could cost $20 billion to $25 billion, it is not a significant amount of capital.”
- Todd's office is designed as an investment office using private equity logic, where the government seeks economic and national security returns rather than financial returns, and awards are structured so that companies have skin in the game.
“We're not looking for dollars returns. We're looking for economic and national security returns on our investments.”
- The program departs from typical government grant processes by engaging applicants iteratively and regularly, rather than accepting applications, disappearing for months and returning with a decision.
“What we didn't want to do, a lot of government programs are put your application in, we'll go away for multiple months, we'll come back, and we'll tell you what we decide.”
- Because the government puts 5% to 15% of capital next to companies that add significant money of their own, the structure aligns interests so that capital goes into projects for the right reasons rather than distorted ones.
“We're putting 5% to 15% up. There's some investment tax credit. There's significant money coming from the companies themselves. That aspect has an ability of aligning interest”
- Building a leading-edge fab is an extraordinarily complex undertaking, with about 600,000 cubic meters of concrete (twice the Burj Khalifa's roughly 330,000), around 5,600 miles of cabling, and cleanliness about 10,000 times that of a hospital operating room.
“the amount of concrete in a given fab is 2x what the amount of concrete that was in the Burj Khalifa, 600,000 cubic meters of concrete versus around 330,000.”
The office should proactively pursue strategic goals rather than react to whatever applications arrive.
- The program's vision for success, released in early 2023, set four pillars plus a supply chain fifth: at least two leading-edge logic clusters, a significant advanced packaging start, a commercially competitive leading-edge memory business, and solving key mature-node bottlenecks exposed by the pandemic.
“we wanted to create at least two leading-edge logic clusters in this country.”
- The program does not reward U.S. companies over foreign ones; it funds companies, domestic or foreign, that commit to long-term investment, supply chain build-out, R&D and workforce in the U.S.
“My job is not to reward a U.S. company versus a separate company.”
- The CHIPS office built a team of close to 200 people, including 40 to 50 investment professionals recruited mostly from sophisticated investment, consulting and financial backgrounds, which Todd considers the critical linchpin of the program.
“We now have close to 200 people from 0. And importantly, the investment team... we have 40 to 50 people”
- Because only certain companies can perform certain technologies, the government's role is not to pick winners and losers but to enable multiple firms to grow onshore and in allied ecosystems.
“our job is not to pick winners and losers. Our job is to enable multiple flowers to grow in effect”
- Rather than reacting to whatever applications arrive, the office proactively seeks out projects and pushes applicants to do more, so that the program pursues the government's strategic goals rather than the applicants' own.
“We're trying to figure out what do we really need and we go proactively out and try to find it. We try to push our applicants to do more”
- Success will be judged in 5 to 10 years by whether U.S. share of global chip capacity shows a clear V-shaped recovery, rising from about 10% today to around 14% by 2030 under SIA projections, versus falling to about 8% if left unaided.
“Left unaided, we probably would go down over the next eight or so years to 8%. And SIA's projections is based on what they're seeing is we will go up to 14%.”
- The CHIPS Act has been among the most bipartisan programs of the administration, with support growing as China tensions and AI became national security priorities, so the election is not expected to change the office's behavior.
“this program... might have been the most bipartisan programs of this administration”
Workforce development is the decisive determinant of lasting success and requires decades of local, sustained effort.
- There is no silver bullet for semiconductor workforce shortages; progress requires local, daily work on relationships, curricula, and widening the talent funnel to show students that semiconductor careers are compelling alternatives to software.
“I honestly wish I could tell you that there's a silver bullet here. There is not.”
- Workforce is one of six key success indicators in applications, requiring applicants to present a workforce plan, and is evaluated in a merit review by the office's workforce team.
“as part of our application, one of our six key success indicators is workforce.”
- The program has set aside specific dollars for workforce development in regions such as upstate New York, Taylor Texas and Phoenix-area Arizona, alongside state and local economic development groups.
“we have set aside very specific dollars to go directly into workforce funding to help support the development of the workforce ecosystem in upstate New York or in Taylor, Texas or in Phoenix”
- Workforce is one of, if not the most, critical determinant of success over the next decade, because fabs need PhDs and materials scientists, lab technicians, electricians and welders, and building that talent pipeline takes decades.
“workforce is one of, if not the most critical components to whether we're actually going to be successful over the next decade because we can build a lot of fabs, but we need to have the people.”
- The ultimate success of the CHIPS Act is for the ecosystem of suppliers, research, universities and workforce to become self-perpetuating after initial infrastructure subsidies overcome the startup hump.
“so that these become self-perpetuating entities. That is the ultimate success in the program.”
- The workforce team both evaluates applicants' workforce plans in merit review and then works directly with awarded companies to organize training pathways and intermediaries after preliminary terms are set.
“that team does work, particularly now post the preliminary memorandum of terms directly with each of these applicants”
- A workforce center of excellence within the forthcoming NSDC public-private partnership is intended to spread best practices nationwide, making talent development a long-term effort over decades rather than a one-time program.
“there will live a workforce center of excellence that is meant to take best practices across the country”
- Recreating U.S. fab capability requires rebuilding expertise, infrastructure, and requirements for power, water, workforce, and suppliers, a multiyear decade-long process rather than a one- or two-year fix.
“that's not a one-year or two-year process, it's a multiyear, decade-long process”
Customers, not only government, should pay for supply security through long-term commitments.
- Applicants must show evidence of demand such as order books, and the program encourages long-term customer agreements with real skin in the game, such as General Motors' long-term agreement with GlobalFoundries for its Malta facility.
“We encourage that because we do feel that it is really important for customers to show with real skin in the game that this matters.”
- Customers naturally want low prices and abundant supply, so resilient supply chains require customers to pay for supply security and prioritize diversification, not just the government.
“Customers, of course, want all the demand that they need and they want it for the same price. That's what they want. And our message is we're doing our part, we need to see customers also step up”
- Many auto companies have become much more focused on their semiconductor supply chains than they were a couple of years ago, treating resiliency as insurance.
“Many of the auto companies have really got religion about this and are really focused on their supply chains in a way they certainly weren't a couple of years ago.”
Early results show large investment and supplier ecosystems forming around the anchor fabs.
- Announced awards to Micron, Intel, Samsung and TSMC total up to $27.5 billion and are materially larger than initial plans, with Samsung expanding from one fab to two fabs plus packaging and R&D facilities.
“Samsung announced two fabs as opposed to one fab, plus an advanced packaging facility, plus an R&D facility, plus expanding their Austin site.”
- The program targets 20% of leading-edge logic production in the U.S. by the end of the decade, up from 0% today, with over 90% of leading-edge logic currently made in Taiwan and the rest in Korea.
“We put out a goal by the end of the decade to have 20% of leading-edge logic in this country by the end of the decade. By the way, that's 0% today.”
- Of the up to $29.5 billion in incentives announced through preliminary memoranda of terms, the projects represent over $300 billion of aggregate investment, roughly 10 dollars of other capital for every government dollar.
“we've announced up to $29.5 billion of incentives through our preliminary memorandum of terms. And that is over $300 billion of aggregate investment that would come along with that. So approximately 10:1”
- A dedicated environmental and NEPA team helps companies anticipate environmental challenges, since these fabs face environmental review challenges as they are among the largest facilities built.
“We have a whole team around environmental and NEPA that is prioritizing how we help these companies get ahead of the game there”
- Companies face upfront constraints from scarce GPUs, energy, and gas-powered turbines needed for new power plants, while the U.S. has a comparative advantage in plentiful and cheap energy relative to other regions.
“you can't get enough GPUs. You can't get enough energy. You can't get enough gas-powered turbines to create new power plants.”
- At least 14 direct TSMC suppliers and 24 strategic Samsung suppliers have plans to construct or expand U.S. operations because of these anchor fabs, showing that ecosystems are forming around the projects.
“at least 14 direct suppliers of TSMC have plans to construct or expand because of TSMC. In the case of Samsung, I think there's 24 strategic suppliers”
- In roughly 14 months since the first funding opportunity, the office received about 650 statements of interest, about 200 pre-applications or full applications, and 160 concept plans for the smaller supply chain opportunity.
“we have received, I think we've announced most of this, about 650 statements of interest, about 200 pre-applications or full applications.”
Program success depends on external coordination, export controls, and rising AI demand.
- Micron probably will not reach 10% of DRAM memory production by the end of the decade, though the U.S. will get close.
“probably won't get to 10% by the end of the decade in DRAM memory, but we'll get close.”
- The CHIPS program is only one component of a broader set of government and market tools, so its success depends on coordination with other government parts and key customers.
“We are one part of a broader set of government and market tools, and we try to engage with those other parts”
- Export controls on China can have unintended consequences, and the program must consider how restrictions alter equipment flows and where capacity ends up.
“Given the guardrails and the export controls that we have in place, you have to think about unintended or even intended consequences”
- Industrial policy of this kind has largely not been done in the U.S. for many years, so the program's large, highly capitalized partners are not in crisis and the program takes responsibility seriously.
“sort of industrial policy that really the U.S., if ever, maybe never has done, but certainly not for many, many years.”
- The U.S. will be the one region with leading-edge capacity from almost every major player, since no other country has more than two of the major leading-edge players meaningfully active.
“There is no other country that has more than two of those players that is meaningfully active.”
- When the CHIPS Act passed, artificial general intelligence was not part of ordinary vocabulary, but within a short time it became a constant topic, which Todd cites as a sign of AI's rapid rise as a demand driver.
“AGI, artificial general intelligence, was not even like a word in the lexicon, right? Now you can't go two minutes without talking about it.”
The office was built from scratch with senior private-sector recruits and adapted its approach to AI demand.
- The surge in AI demand was not anticipated when the CHIPS Act passed, but it has not changed the program's approach; it has only made leading-edge logic and memory more important.
“It has just made what we're doing even more important. It means the need for leading-edge logic and memory chips is even more important.”
- When the CHIPS Act passed in August 2022, there were no people, processes or rules to execute it, so the office had to build the entire program from scratch after Todd and a colleague joined the week after Labor Day 2022.
“There are no people. There are no processes. There are no sets of rules.”
- The CHIPS office went from no staff after joining in September 2022 to close to 200 people, with about 40 to 50 in the investment team.
“We now have close to 200 people from 0”
- Recruiting senior people with two-plus decades of experience from Goldman Sachs, Synopsys, Intel, private equity, venture capital, and McKinsey was the first priority, and the office set processes and rules by February 2023.
“pulling that talent together was job number one. Job number two was setting a set of processes and rules”
Announced funding stays uncommitted pending diligence, because taxpayer dollars demand maximum outcomes.
- About $29.5 billion of the $39 billion manufacturing funds has been announced across nine companies under preliminary memoranda of terms, plus $25 billion in loans, which is over 80% of available dollars, with some reserved for DoD requirements.
“We're over 80% of the dollars that have been announced.”
- Preliminary memoranda of terms are subject to final confirmatory due diligence and negotiation of ultimate award documents before disbursement begins, so announced amounts are not yet committed cash.
“what we've announced on these preliminary memorandum of terms are subject to final confirmatory due diligence, the negotiation of the ultimate award documents.”
- Companies must already have a state or local incentive in hand before applying for CHIPS awards, so the federal office does not pick geographies.
“companies have to come to us already with a state and local incentive in hand.”
- Because every dollar is taxpayer money, the office has a duty to get as much economic and national security outcome as possible for the smallest possible amount of money.
“we owe it to the taxpayer to do this to get as much as we can for as small amount of money as possible”