Eyepopping Factory Construction Boom in the US: Semiconductors, Auto Industry, and Everyone Else

A massive corporate rethink is underway. Inflation in construction costs explains only a small portion of the spending boom.

By Wolf Richter for WOLF STREET.

Companies plowed $18.4 billion in April into the construction of manufacturing plants in the US, a seasonally adjusted annual rate of construction spending of a record $212 billion, according to the Census Bureau today. This was up by 140% to 200% from the range in 2015 through mid-2021.

The spike in factory-construction activity began in the second half of 2021. The CHIPS Act, signed into law in August 2022, seems to have turbocharged it, though the money hasn’t even started flowing yet.

Beyond semiconductors, a large number and a great variety of manufacturing plants have been announced over the past two years, and we’ll get to some that were announced just in recent weeks. They’re part of what the National Manufacturing Association has termed, “general reshoring.”

During the record year of 2023, spending on factory construction spiked by 71% from 2022, and by 138% from 2021, to $196 billion. And 2024 is on track to set a new record as the eyepopping boom continues.

This boom is a result of a massive corporate and government rethink after the supply-chain and transportation chaos during the pandemic, the increasingly edgy relationship between the US and China, the fundamentally scary dependence by US companies on production in China, and the nerve-racking dependence on semiconductor production in Taiwan.

Inflation explains only a small portion of the spending boom.

The Producer Price Index for construction costs of nonresidential buildings exploded in mid-2021 through 2022, but then hit a ceiling and has remained roughly unchanged at this sky-high level since December 2022 (red in the chart below).

From September 2021 through December 2022, the index spiked by 28%. Over the same period, investment in factory construction jumped by 56%. Since then, the PPI has remained flat to down, while investment in factory construction jumped by another 64%.

On a year-over-year basis, the PPI for nonresidential construction has been negative for the first four months of 2024 (green):

Construction costs are just part of the investment in a factory.

In general, today’s largely automated manufacturing processes require costly equipment that can dwarf the construction costs of the building. But the costs of the manufacturing equipment and related costs are not included in the data of construction spending here.

In terms of the semiconductor industry: the total cost of a chip plant might reach $20 billion, but the construction costs are only a smallish part of it. The costs of the equipment and the costs of getting it all to operate eat up a far bigger part of the investment.

Massive investments in semiconductor plants.

The first grants and loans under the CHIPS act were awarded only in March 2024, spearheaded by the announcement that Intel would get up to $23 billion in grants and loans, plus up to $25 billion in tax credits. This was followed by announcements that TSMC (the world’s largest contract chip maker), Samsung, and Micron each would get over $6 billion in grants, and that Global Foundries would get $1.5 billion in grants. Smaller announcements followed.

But the deals will take time to finalize, and disbursement will be spread out over phases.

The total costs of building a fab, to equip it, and to get it up and running are huge. TSMC has announced over $65 billion in investments, including $40 billion for two fabs that are now under construction near Phoenix. Despite assorted delays and issues, TSMC said that its first fab “is on track to begin production leveraging 4nm technology” in the first half of 2025.

Intel has rolled out $100 billion in investment plans, including $43 billion for facilities in Ohio, New Mexico, Oregon, and Arizona. Texas Instruments is investing $30 billion in Texas. Samsung is investing about $45 billion in Texas. Micron is talking about $125 billion, with new plants in New York and Idaho. These numbers are gigantic, but only a smallish part of it all will be construction costs of the building and will be tracked here.

New factories beyond semiconductors.

Chip plants are written all over this boom in factory construction because the plans are huge and costly, heavily government subsidized, and get all the publicity. But there’s more going on.

Here is a sampling of announcements beyond semiconductors over just the past few weeks:

Toyota announced a $1.4-billion investment in its Princeton, Indiana, facility to assemble its new three-row battery electric SUV, bringing total investment in Indiana to $8 billion. “This investment will not only provide plant infrastructure to build the all-new BEV, it will add a new battery pack assembly line using lithium-ion batteries supplied by Toyota Battery Manufacturing North Carolina, a $13.9 billion facility slated to begin production in 2025,” Toyota said on April 25.

Kohler Co., maker of kitchen and bath products, inaugurated a 1-million-square-foot manufacturing facility in Casa Grande, Arizona.

Crystal Window & Door Systems announced a new manufacturing facility in Johnston County, North Carolina, to specialize in aluminum and vinyl extrusion, and window and door fabrication.

ION Storage Systems commissioned a solid-state battery manufacturing plant in the US, in Beltsville, Maryland.

GAF Energy inaugurated its 450,000-square-foot Timberline Solar manufacturing facility in Georgetown, Texas.

GF Casting Solutions AG, a division of Georg Fischer AG, headquartered in Switzerland, announced a new manufacturing facility in Augusta, Georgia, to produce large structural cast-aluminum components for the automotive industry.

Boviet Solar announced a 1-million-square-foot manufacturing facility for solar panels in Greenville, North Carolina.

Green New Energy Materials, a battery component manufacturer, announced a $140 million lithium-ion battery separator manufacturing facility in Denver, North Carolina.

Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the beer and iced-tea mug to find out how:

Would you like to be notified via email when WOLF STREET publishes a new article? Sign up here.

  103 comments for “Eyepopping Factory Construction Boom in the US: Semiconductors, Auto Industry, and Everyone Else

  1. Debt-Free-Bubba says:

    Howdy Lone Wolf Other places are typing recession worries, housing glut due to interest rates, factory and construction spending both fall.

    • Wolf Richter says:

      They’ve been saying the same stuff for two years, due to whatever. Recession articles get lots of clicks, that’s the real reason.

      Here’s total construction spending. The peak of the spike was in April 2023, a year ago, and there’s still no recession:

      • Debt-Free-Bubba says:

        Howdy Your website makes you a Lone Wolf for sure.

        • joedidee says:

          CHIPS only has $285,000,000,000 of free money for LARGE CORPORATIONS who outsourced after NAFTA/WTO
          gotta love getting PAID TO COME BACK HOME
          Inflation Production Act going to kick it into high gear

          when will it end?? in 2025 right after election of course
          time to pay the bill folks

        • MarMar says:

          “when will it end?? in 2025 right after election of course
          time to pay the bill folks”

          To the contrary, as Wolf explicitly said, it’s only really just getting started.

        • Geriatric says:

          You are quite correct, truth always hurts. Industrial building construction is a late economic boom stage marker, due to the 4 to 5 year lead time between conception and completion. The inevitable following correction occurs for different reasons and speeds each time, but the results are always similarly harsh. This is the essence of capitalism, survival of the fittest. In this cycle the problem is likely to be the fact that USA production costs will always be greater than Chinese slave type communist ones and are Americans willing to pay a lot more for all their electronic gadgets? Particularly with persistent 5%+ inflation? Huge exports of expensive USA made goods to the rest of the world are regrettably past, even if they are made by AI robots.

      • Ciprian says:

        Wait for the earnings growth coming from these factories and companies. A new decade of growth, recession what? Yields high for some time to come.

        • joedidee says:

          new HI-TECH-AI factories to employ under 50 workers
          instead of 5,000
          so few make it while rest get to become takers

        • JS says:

          That’s just not true/feasible. As someone who runs a manufacturing facility, I can tell that you have your workers (normally three shifts), your period cost employees (senior leadership, front line leadership, technical staff, engineers/facilities, administration (health and safety, sometimes finance/accounting staff)), plus maintenance, warehouse staff, material movers, etc plus your cleaning staff (normally outsourced), security/front desk support, cafeteria workers, etc.

          Not all of these people will be located in that location, but as your capacity rises, there’s always tons of ancillary positions that get added at corporate offices (and remote) to support the new manufacturing facility.

      • Jack says:

        Recessions come when construction ends, all the new builds go on line, and there is oversupply.

        Construction jobs are replaced by much fewer operational jobs. Prices fall due to spike in supply.

        When this happens there could be a recession. Until then, guns are blarring.

        • Kent says:

          I tend to think recessions are more difficult to get to in the current age. Construction and auto manufacturing were huge portions of the US economy in the ’60’s and ’70’s. Not so much anymore. And if manufacturing reaches it limits, companies can always import more from China. Maybe we will have a recession an nobody will notice.

  2. dougzero says:

    Another for the pile: Wolfspeed is building a chip mfg place in NC.

    • Northernlights says:

      Unfortunately I think the Wolfspeed business plan requires free money as they haven’t turned a profit since they got rid of the LED business. Maybe the latest set of grants will float them for a while longer, but their partnerships with Jag and GM are not very inspiring. At least GM won’t need product anytime soon.

      All of these plans to build factories look great on paper until someone actually has to deliver physical products. Fuji is expanding in Wake county, but they actually know what they’re doing.

      I swear NC has more than it’s share of big smoke and mirrors projects, like the mysterious Apple dreadnought and VinFast.

      • El Katz says:

        Vinfast is – once again – delaying their production facility in NC.

        From GM Authority:

        “Struggling Vietnamese EV startup VinFast has reportedly delayed the construction of its North Carolina production facility. The plant, initially announced in March 2022, broke ground last July with an expected completion date by July of the 2024 calendar year. However, this timeline was later pushed back to 2025, and now, the plant launch timeline may be facing yet another delay.
        According to a recent report from Automotive News, which cites Reuters, VinFast could postpone the facility’s launch as the company continues to suffer financial losses.”

  3. Harvey Mushman says:

    “Eyepopping Factory Construction Boom in the US:”

    Music to my ears!!!

  4. Ole CG Olesen says:

    The Boom is exlusively due to the destruction of competitive EUROPEAN Industrial Manufactoring …. only STUPIDOS in the European Comission have not realized this yet .. or .. they have been BRIBED / INTIMIDATED .. what do I know .. in order to act as SUICIDAL as they do .. to the DETRIMONY of EUROPE … BASTA !

    • John k says:

      Yes, us is raiding eu, and particularly German industrial capacity. Nordstream destruction guaranteed eu high energy costs while boosting us energy firm profits. And now us persuades eu to ‘de-risk’ china trade. Why are eu leaders allowing this? I assume US dollars are involved.

    • DubTea says:

      It’s the cheap natural gas/energy we produce that entices their manufacturing to off-shore the EU. We’re not stealing they are expanding.

  5. Tom V. says:

    Despite all the talk of restrictive interest rates, quantitative tightening and general malaise around recession prognostications, a lot of companies and people sure are spending.

  6. RickV says:

    Great summary as always. Now if only they can find enough trained workers to staff the plants. Does this mean more upward pressure on salaries and wages?

    • Wolf Richter says:

      They’re already having trouble finding enough workers to build the plants, LOL.

    • John k says:

      More upward pressure would be good. Wages haven’t kept up with inflation.

      • Whitey Joe Young says:

        I recently read more than one article saying wages have risen more than inflation, but people still only notice how much gas costs, insurance rates, etc.

  7. Eric says:

    I could digest inflation data from different industries better, if you could provide wage inflation data of the industry, too.

    • Wolf Richter says:

      If you can’t digest it, don’t eat it, my rule of thumb.

      More seriously, this was about factory construction spending. And I gave you the cost increases for nonresidential construction, which include the costs of labor.

  8. Brant Lee says:

    Will robotics and AI along with plentiful energy eventually bring most industries back to the U.S.? Even then, does anyone want to work here anymore?

    • Harvey Mushman says:

      “Even then, does anyone want to work here anymore?”

      If you want to buy a home, furnish it, put a car or two in the garage, eat, raise a family, etc etc.

      You have to work.

    • Ciprian says:

      What about the new young generation that is being trained now? It’s happening slowly.

      • Midwest Ralph says:

        I can see it here in the midwest. Factory space where I work is being doubled, but it is still tough to find the workers.

        • JS says:

          It’s forcing those of us in the industry to go back to the “glory days” of apprentice programs, giving low level employees opportunities via training programs, educational programs, etc. and hiring non-traditional employees (high functioning autistic employees, older employees (on 2nd career), etc.

          Painful for sure (I can attest), but in the long run the right decision for our country. If you’re looking to pay peanuts and treat your employees poorly, you will struggle but that’s a positive result IMO.

        • old ghost says:

          Here in my little corner of fly-over country, a factory built in 1912 has a new metal roof. I am not sure when the old roof collapsed, maybe a year or two back, but it was a sight.

          No name on the building now, so I don’t know what they are planning to use it for (maybe warehousing?).

          No way it will be used as residential. So probably a small part of the big number in Wolfie’s construction spending.

    • Glen bar says:

      Your first sentence made some sense, and it’s really the only way that the United States can out-manufacture China in the sector that’s being manufactured. It may be cheap for China to make and sell 10 million units; probably because you’ve got the equivalent of half the population of Metro New York City and their little lemur fingers carrying out all that assembly, but build the factory here, eat the high capital starting expense in automation, and 5 years into production will make all those little lemur fingers as expensive as a Ph.D’s salary.

      In other words, building it here with robots will be hella less expensive than building it there with all those little fingers of theirs. Our expense will be the $158/month electric bill. Their expense will be $12M in annual labor costs.

      Now your second sentence on the other hand… It can use some work. Americans enjoy the highest per capita GDP of anyone, anywhere for medium and large economies. We build walls to keep em out, not to keep em in…

      • Brant Lee says:

        LOL, I admit to being a bit sarcastic. It will always be the same in the U.S. if you want to work, you can make it well.

  9. Franz G says:

    there will be demand for semiconductors for many years to come, but i still think the nvidia ai craze is just that, a craze.

    market cap of 2.8 trillion usd? p/e of 68? how ridiculous is that?

    • Gattopardo says:

      Quadruple earnings and its 16 P/E. That’s the bet.

      • Franz G says:

        while we’re dreaming, why not increase earnings by a factor of 10, and then the market cap can grow to $6.5 trillion usd and still only have a p/e of 16

        • Kent says:

          Cisco’s revenues were $1.3 billion in 1994 and $57 billion in 2023. Revenues don’t equal earnings, but its a pretty good proxy. Cisco was the critical infrastructure for the internet, nvidia is for AI. We are in about the same stage with AI today as we were with the net in 1994.

        • Franz G says:

          kent, cisco trades at significantly below its 2000 peak. maybe a poor example?

      • Glen bar says:

        The p/e of 60 is trailing 12 months. Look at their revs and net income last quarter–$25B and $12B respectively and it’s easy to see that there p/e will be at or around 12 this time next year.

        • Franz G says:

          yeah, if you assume that not only do their margins stay where they are, but that the ai craze leading other bigtech companies to buy tons of chips and server capacity, continues.

          i’ve been studying economies for a long time, and in my experience, margins like that don’t continue indefinitely absent any illegal or otherwise anticompetitive behavior.

    • Eng says:

      There’s certainly an AI craze with some usefulness but not a whole lot despite 2-3 years of continuous growth. However they are getting close, to something profitable, so they need more compute to test out that theory.

      The future of AI is far greater then what is currently out there. Sam Altman will claim that only his models are the best and at the end stage of what AI is capable of but that’s false. OpenAI and other platforms are a stepping stone for what’s next.

      All of the above requires new and more silicon. Whether it works or not is beside the point, they all need silicon and Nvidia is ideal for that.

  10. Ethan in NoVA says:

    I thought I saw that the chip factory handouts had strings attached regarding DEI hires and other stuff. This was causing issues with TSMCs choice of hiring the people they wanted?

    Do we need the capacity really?

    • bunter says:

      The capacity comes with the territory. You can’t fire up a foundry on demand it has to be running continuously. So, you’d better have a stream of designs to feed it.

      Also, small geometry like 4nm, uses equipment and stock materials that are geared for continuous operation.

      Ultimately, if we wish to remain a high tech society then we best have this infrastructure onshore.

  11. TulipMania says:


    How does this compare to the late 1990s Internet bubble?

    Massive factory and building spending then as well, although the current factory construction seems more sustainable than that.

    • Wolf Richter says:

      The biggest thing back then was laying fiberoptic cables across the globe, and install the hardware to make it work, so that vast amounts of bits could move through these huge data-pipelines. This was at the time when the www in in URLs stood for “worldwide wait” and when 1 megabit per second was considered huge broadband that almost no one had access to, and when most people were still on dialup internet (65 kilobits per second). There was a mega-bubble in these fiber and equipment companies (including Cisco, Lucent and a gazillion startups), and it blew up spectacularly, and investors got wiped out, and lots of these companies vanished, but what came out of it was the internet as we know it today where you can get 10+ gigabit/s broadband on a cellphone, which is 10,000 times as fast as 1 megabit.

      Today, this effort to encourage semiconductor production on US soil is different. It’s an industrial policy to protect long-term US economic and military interests. To some extent this is a matter of national security. No one wants to be dependent on China for the core of what makes everything tick.

      In terms of the broader effort to manufacture on US soil, it seems everyone learned a big lesson during the supply chain chaos in 2020 and 2021. People saw that total offshoring and relying on China to hopefully save a few percentage points is just not the way to run a business. They found out that it made them very vulnerable.

  12. Toys R Yota says:

    What’s your take on the Toyota investment who seemed to be EV denialist until the CEO was removed. There seems to be lots of flux in the space, but it seems to me that Toy Yoda might be late to the EV game.

    • Wolf Richter says:

      yes, they’re years behind in the EV game; and so they’re dissing EVs out of one side of their mouth to slow down the EV runaway train to make it easier to catch up, while they spend billions a year to catch up.

      • Goldendome says:

        Toyota may be behind in EV but are years ahead of many in hybrid vehicles. Toy has hybrids (I believe) in all model cars & trucks.
        Hybrid technology is wonderful allowing people to drive much longer distances without having to worry about charging while using much less gasoline. I believe Gm & Ford have greatly cut back EV production for lack of demand. I’ve owned a hybrid for 10 years—best car I’ve ever owned.

        • Wolf Richter says:

          We own a Ford hybrid, LOL. Ford has hybrid powertrain options available in all its models, including the F-150.

          Ford’s EV sales in May jumped by 65% from a year ago. EV sales are not pulling back.

      • Goren says:

        Yeah, I guess that is why Toyota just recently announced record sales and profits.

        Just state you hate Toyota for whatever reason and be done with it.

  13. Bobby1971 says:

    It will be interesting once the factories are up and running, how many new tarrifs will be needed to keep these factories competitive to cheaper global imports. Prelude to trade wars. Will the total of US factory workers even make a dent to match those of 1950?

    • Wolf Richter says:

      People that invest in factories do this math — what it takes to be competitive — all the time. There are many costs involved in producing overseas, including: transportation, delays, loss of quality control, and loss of intellectual property (standard for production in China). Those costs began to amass during covid, and companies saw how vulnerable they were, and how costly it was to produce overseas.

    • RH says:

      It is not just costs. Read the news. The interdiction of more commerce at choke points is likely in future. Aso the charging of huge refueling, storage, docking, insurance, and other fees.

      A certain exporting country is in deep financial trouble, so it may later not be able to protect the ships of its exporters one day. Materials and supply lines will be also be politically,restricted within blocks, e.g., rare earths.

    • Tom15 says:

      We run a mom and pop business.
      Equipment we operate needs to be updated.
      Us manufacturer cost $20k. Chinese $2.5k.
      Chump change for the big guys.
      Performance: Chinese longevity: ?????

      Gonna need higher tarrifs.

  14. Dan Farrand says:

    Maybe if re-shoring keeps up for 5 or 10 years it will start to matter. It may also be that we will see higher prices and worse quality.

    Whose going to work in these factories ? For some of it, the Mexicans that are being imported may be suitable, but I’m not sure the Mexicans coming now are coming for the “opportunity”, but we will just have to wait and see.

    Perhaps the Junior College’s will figure it out and develop a 2 year “mechanics” degree program to equip people with basic engineering skills. Enough to be able to run NC machinery. But regardless, people will need to be paid more than they pay in China or Viet Name or Mexico. So it will either be massive capital outlays for massive automation or higher prices behind tariff walls or both.

    • Wolf Richter says:

      Factory workers today are far different from decades ago. 1. These highly automated factories have far fewer workers. 2. Lots of factory workers have engineering degrees, and in chip plants, many have advanced STEM degrees. And those that don’t are highly trained and qualified. There are some exceptions, such as meatpackers — but even there, automation is taking over, and jobs require more skills.

      In a modern car, the labor cost component is very small, less than 10% of the cost of assembling the car, thanks to automation.

      • FaradayRotation says:

        Please listen to wolf closely on this point- he is right, and I know he is right because I work at one of these modern factories.

        A proper modern factory is a palace of automation. They totally different animals than the smokestack-deathtraps of yesteryear. They are much safer, much more pleasant working environments, better for the environment, and vastly more efficient in every respect. Meanwhile, the people operating them are infinitely better educated, better trained, better paid, and far more respected for their brain power and skills.

        The jobs at these places are generally solid and well-paying for hardworking and creative people of every walk of life (doofuses, goofballs, and layabouts excepted – though occasional goofing around is permitted). Many engineering positions require at least a masters, often a PhD. Many of those same brainiac engineers get regularly barked at by technicians with 20-30 years of experience and training on the process line. So while these jobs are anything but cushy, you will have a lot more satisfaction working at one of these places than your great-grandpa ever did working the process line before automation, that I can promise!

        I think Wolf’s labor cost estimate is maybe a little low for USA (if including other industries), but I’m sure the difference doesn’t pencil out in the end. I think businesses have discovered that modern automation has really changed the financial calculus even as recently as ten or fifteen years ago – western countries can still be bastions of production if they execute correctly. So while the people cost is undoubtedly aggravating to the MBAs and tycoons, those with a working brain know that the real eye-watering money pits are those darn automated thingies which are always breaking, and if they aren’t, the engineers are always experimenting on them, the vendors always have the hottest addons, and the techs are always fiddling to keep them running. But if the robots run and they are run well by a good team, the factory will literally spray out money like a fire hose at many times the rate a pre-automation or middlingly-automated factory would.

        There’s irony in that giving robots jobs that only a genuinely desperate human would ever do, we create tons of jobs doing things that humans love doing (despite our loud complaining all the while).

        Anyways, like Wolf said, these jobs are *difficult* to fill. You have to have both a lot of capacity for abstract thinking, programming, and planning as much as knowledge of nuts, bolts, and grease. The US doesn’t have enough people trained up because no one has really done this before outside of select niches within certain industries. But here we are! And you *can’t* just import these folks and plug them in – they have to be retrained to work in a highly-automated environment.

        While I’m on this binge: the thinking on supply chains has shifted fundamentally, because the risk of buying key feedstock from a temperamental neighbor is being fully comprehended by the business-gurus at the top. Their calculus says says the risk is much, much bigger than they thought it was and now they want to mitigate the risks. Remember: a modern factory has an enormous appetite and entire business units and departments dedicated to feeding the machines. Every second the tools are idle or the line is down, a factory business loses gobs of money through every pore and orifice. COVID showed them exactly how much money they can and will lose if they have supply chains which don’t tolerate even lukewarm geopolitical risk.

        • Louie says:


          Excellent piece of writing! Kudos to you!

        • danf51 says:

          Then why does a single 155 artillery round in the new factory opening by General Dynamics cost $4000 while Russia produces a round for $1000 ?

          It remains to be seen what re-shoring will mean. I will believe it when I start seeing appliances made in US factories.

        • Harvey Mushman says:

          I don’t usually read long posts, but I read this one. Very interesting!

          Haas Automation Inc. is right down the road from me. I regularly see huge automation machines being trucked by my building.

    • Dave Chapman says:

      Connecticut used to have Electronics Technician High School.
      I hired one of the graduates. He was pretty good.
      Do not underestimate 2-year technical training programs.

      I used to work with a guy who came from a family of loggers. He joined the Air Force, and they taught him coding. This guy was one of the best Software Engineers I have ever seen. The military has some astoundingly good 3-month and 6-month technical training programs.

      And, yes, we have a very large number of solid Computer Engineering, Manufacturing Engineering, and Systems Engineering programs at colleges all over the USA.

      • Anthony A. says:

        I went to a technical high school in Connecticut and a technical college there too. Then I got my MBA after working for 10 years in heavy industry. Great schools in that state. I had a solid career and am retired now.

  15. Natron says:

    On the automation front I checked out the ROBO ETF and it seems to be tracking the SP500 so far mol.
    I half expected it to be participating more in the rally… any ideas on why it’s lagging if that’s what it’s doing?

  16. Hubberts Curve says:

    It is my personal opinion that these latest generation chip plants are no longer profitable, so the purpose of the Chips money is to subsidize the cost for an otherwise money losing facility.
    Back in the late 90’s and early 2000’s Intels latest Fab’s famously cost about $1 Billion. They produced chips that when they first came on the market that sold for $800- $1000 each retail.
    Now , as you mentioned wolf, a new chip fab costs $20 Billion or more yet the latest Intel chip sells for $699 retail. This massive increase in capital expenditure does not really get any huge increase in productivity , or reduction in inputs either has the process’s are massively more complicated and getting good yields is even more difficult.
    If I had a business where my capital costs for a new factory were 20 times higher than a few years ago, and I was producing items for sale with that business that were 20% cheaper than before I would be scratching my head and looking for a hefty government subsidy too.

    • grant says:

      Volume and gross margin also matter, it’s not as simple as “$699 retail chips today produce 20% less profit than $900 retail chips did 20 years ago”

      • Hubberts Curve says:

        in 2000, the steppers used to make the state of the art chips of the time cost about $10 million and could process between 100 and 200 wafers per hour. The newest ASML Twinscan EUV lithography machine costs $300 million and can process 180 wafers per hour. So the Volume per capital expensive Tool has not changed, but the capital cost of the machines is 30 times higher and the price of the end product is lower. I doubt the gross margin is magically higher today.

        • AuHound says:

          How much has the wafer size increased since 2000? Remember, a doubling in the diameter increases chips/wafer 4x.

          Google search:
          Wafer Properties
          1-inch (25 mm) (mid-late 60s)
          2-inch (51 mm). Thickness 275 µm.
          3-inch (76 mm). Thickness 375 µm.
          4-inch (100 mm). …
          5-inch (130 mm) or 125 mm (4.9 inch). …
          150 mm (5.9 inch, usually referred to as “6 inch”). …
          200 mm (7.9 inch, usually referred to as “8 inch”). …
          300 mm (11.8 inch, usually referred to as “12 inch”).

    • JimL says:

      It is strange that you say the latest chip plants are no longer profitable because chip makers can be some of the most profitable businesses in the world.

      That said, it isn’t because of profitability that these plants are being subsidized, it is because of variance and risk.

      Building a semiconductor fabrication plant is all about fix costs versus marginal costs. There are humongous fixed costs to building a fabrication plant, but the marginal cost of production is absurdly miniscule. The difference in materials and labor in producing 1.5 million units versus 1.8 units is a rounding error in semiconductor manufacturing.

      So with huge upfront costs, but insignificant marginal costs, it is all about hiw much demand there will be for the product. If demand comes in ever so slightly lower than expected, a chip manufacturer will lose a ton of money because he cannot cover the huge fixed costs of building the plant. However if demand is ever so slightly higher than expected, a manufacturer will make a ridiculous amount of money.

      Those $699 chips you mention probably have above a 90% marginal cost. So once the fixed costs are covered, 90%+ of the money falls to the bottom line profits. However the company had to bet tens of billions of dollars in building the fab to even get into the game.

      What complicates the bet is that it has to be made many years prior.it takes many years to build a fab, furthermore construction starts before the technology that will be used in the plant is even ready for full scale mass production.

      Unless a company has a pseudo monopoly like Intel had in the 90’s, it is an insanely risky bet for companies to take on the risk to build a cutting edge semiconductor manufacturing plant. So much money has to be spent on unproven and unknown processes. So much can go wrong.

      However when it goes right, the fans are the closest thing to being legal monetary printing presses.

      That insane risk is why it has reached the point that nation states need to get involved on the building of these plants. Companies (even multinational monsters) don’t have the resources to make the bets needed to remain cutting edge.

      • JimL says:

        It should also be noted that the U.S. taxpayer is getting some benefit from these insane bets. They are benefiting from having access to the most cutting edge chip manufacturing.

      • Hubberts Curve says:

        I think you right before we started reaching the limits of Moores Law. Back then ,once the capital costs of the Chip Fab, Design Costs ( chip design), Masks, Etc where payed for the gross margin per chip was big.
        But now as we approach the limits of physics both the capital costs, shake out costs, and marginal costs have multiplied exponentially. The costs of running a 7NM fab ( and below) are enormous. The costs for electricity, water, wastewater disposal, rare gases, chemicals, and other supplies is eye watering.
        The D!X Plant and all of surrounding Intel ops that supply it with services, engineering, chip design, etc which include Hawthorne Farm, Jones Farm, and Aloha employee 20,000 people and that does include the hundreds of suppliers with huge numbers of employees.
        Thats a heck of a monthly nut to cover before you even begin to pay back the initial capital costs of the fab.

  17. Tresho says:

    Meanwhile, back in Indiana: “Eli Lilly plans to double its investments in a new plant in Indiana, according to a report from German Manager Magazine. Eli Lilly wants to invest an additional $5.3 billion into the construction of the factory. This puts the total investment, according to the report, at nine billion dollars – the highest amount for a single facility in the history of Eli Lilly. The reason for the massive increase in financing volume is the ongoing demand for the company’s products, especially the weight-loss injection Zepbound.”

    • Ciprian says:

      This goes to enrich the investors. Clear as day. Until supply and demand is balanced again.

    • Anthony A. says:

      Just think, someday all Americans will be skinny again!

  18. BS ini says:

    Wonder if LNG is included in manufacturing plants they are big spends

  19. Citizen AllenM says:

    I am amazed that we are continuing to onshore super high tech. And to think only about the profit loss statement. Instead, it is entirely about control of the highest level chips, and that is more important than desktop crap. AI doesn’t run at the gummint specs on a bunch of 9900s. Simply put, we now want a ton of the super chips, and they will be made here and kept here. Other production is also here because of location. Shipping higher end stuff from ip theft zones is also problematic. Notice that none of the super high end pharma stuff is coming from China? I wonder why. Cars might as well be considered mid level products, and notice they are already put together in Mexico. America owns so much intellectual property, and that advantage is not being lost.

    What is now apparent is the desire to cut loose our lessor client states, and have our trade empire only matter for us to send the highest products to limited distribution areas (like weapons), and just be happy to have cheap resources to push the consumer happiness for our voters.

  20. Goodbye Mr. Chips says:

    South Korea is going to invest ~$470 B into chip production. They are building 1000 acre farm about 40 km outside of Seoul. It used to be domestic rice production to guard against starvation sanctions. Now, it is chip production to guard against economic starvation.
    Never count out the crazy Koreans.

  21. Mike says:

    It boggles my mind that nobody in Arizona cares that we’re about to become the microchip manufacturer of the nation.

  22. R2D2 says:

    The US spent 38 years happily gifting away its IP, jobs and factories to brutal communist dictatorship China (1978-2016). (But refused to do the same to the brutal communist dictatorship USSR). The US is now furiously back-pedaling, and will spend the next 38 years trying to unwind the China folly. It’s one heckuva mess!

  23. Bear Hunter says:

    Has goverment grants and other free money ever made anything number one?

    Second tier and outdated before the doors open.

    • Dave Chapman says:

      I am in the business, so people ask me about the CHIPS act.
      “80% of the money will be stolen” is what I tell them.

      We are pretty much on track, so far.

      • Wolf Richter says:

        ““80% of the money will be stolen” is what I tell them. We are pretty much on track, so far.”

        Well, no, not yet. None of the money has been handed out yet.

        Also, the companies have already invested many billions of dollars in fabs, and the CHIPS grants and loans will just replace funds they’d already spent.

    • JimL says:


      When the U.S. was investing heavily in NASA, the U.S. was by far number one in space. Since the U.S. has been investing ridiculous amounts in the military, it has easily had the number one military in the world. When the U.S. government was investing heavily in higher education we had by far the best higher education system in the world.

      Knee jerk anti-government spending is silly. There is smart spending and stupid spending. It is better to be thoughtful about it rather than knee-jerk ideological against it.

  24. In China We Trust says:

    US companies have spent decades hiding their wealth in assets overseas to avoid taxes. Agree when Covid hit in March 2020, the most powerful country in the world was brought to its knees, as the realization that the simplest of everyday items could not be readily available or sequestered. 4 years later I don’t think we’re in any better shape, if another manufactured crisis hit our nation tomorrow. The irony of jumping in bed with Communism, while preaching global democracy will be a story to tell for the ages. My local Kroger in Colorado had fresh broccoli today, but the produce manager could not explain why it was shipped from China.

    • Escierto says:

      This is why I have always refused to shop at Wal-Mart. Just look for a product on sale there that is NOT manufactured in China. You can’t find one. Wal-Mart is as much a front for selling Chinese goods as Shein or Temu.

  25. GringoGreg says:

    Will be producing sheet that will cost many times more than China and dozens of other places around the globe can churn-out! You can buy an Chinese EV in many countries for half the price, or even less than half, of a equivalent EV manufactured in America. Just one glaring example of 1000s of much lower priced products.

  26. GringoGreg says:

    Mercantilism is now America’s future which means inflation, expensive sub par products and declining standard of living.

  27. JimL says:

    “This boom is a result of a massive corporate and government rethink after the supply-chain and transportation chaos during the pandemic, the increasingly edgy relationship between the US and China, the fundamentally scary dependence by US companies on production in China, and the nerve-racking dependence on semiconductor production in Taiwan.”

    It is just human nature. When things are calm and stable, it is natural for companies to simply look for the lowest price/cost. Reliability is naturally assumed. When catastrophe hits, it wakes people up to the fact that the lowest price isn’t always the best. There are other factors like reliability, timeliness, and integrity that can matter as much or more than price/cost.

    Before the pandemic, companies were shipping production offshore for a few percent savings in costs because A.) Their competitors were doing it and they needed to compete. And B.) The reliability and timeliness of moving production offshore, often outside of the company’s control, was assumed to be no different than keeping it close.

    The pandemic made companies reassess the tradeoff between low costs and reliability. It was more than the pandemic however. I worked with a small manufacturer who had a plant in south eastern Europe than took in raw materials and did some pre-production work before shipping the semi-finished product to the U.S. for automated finish. When the ship blocked the Suez Canal and cut off all traffic through there, it literally shut down the European factory. Their margins were so tight that the extra costs of shipping raw materials around Africa was not economically feasible.

    The company almost went under.

    They have since decided to diversify their supply chain even though it costs more.

  28. Hardigatti says:

    It will take more than money to successfully onshore manufacturing again. 30+ years of offshoring have drained much of the US know how. STEM degrees are not valued in the US. We’ll have to import them.

    • Franz G says:

      stem degrees are not valued because we decided that wall street wizards who sell crap financial products should make 10 times what engineers do.

      no wonder the best and brightest don’t want stem.

      • Drive-by Poster says:

        Makes me think of how Boeing failed as MD management who didn’t have strong engineering culture took over and put near term profits ahead of design and build for the ages. Now we have planes going out where no one inspected how the wings are attached to the body.

  29. MM says:

    Wolf, end of the first paragraph, swap that comma for a period.

  30. Drive-by Poster says:

    So where to invest to make money on the factory construction boom? Already holding Nucor Steel, but what else?

  31. Michael Donahue says:

    These were the best comments ever!

  32. Michael Donahue says:

    These were the best comments ever.

  33. Michael Donahue says:

    These were the best comments ever. I didn’t already say this!

  34. Glen says:

    Took awhile for giving up the manufacturing sector chasing low cost labor and loose regulations to catch up. I remember back in the 90s when that was a huge concern. Not sure how this reshoring will go and not clear it is being done for all the right reasons. But won’t know for a long time as it will take a long time to see fruits, if any. Hard to put genie back in bottle once out and while a lot of construction jobs not clear how many permanent ones along with people to fill them. In the end it seems likely this will all just contribute to more inflation.

  35. Excellent article as usual! A few points:
    1) The CHIPS and ludicrously named “Inflation Reduction Act” (IRA) have all sorts of Dem strings attached, such as union labor requirements and climate change mitigation.
    2) How many of these plants would get built without either direct or indirect government support?
    3) It will be interesting to see how many of these plants get built AND are operating profitably relative to their foreign competitors. Call me skeptical…

  36. Robert says:

    Obvious reason for factory boom is the Ukraine war, e.g., even in Arkansas, ranked #42 for defense contracts, there is a boom in factory construction, rebuilding, etc.

    • Wolf Richter says:

      Good lordy. The biggest portion of the boom are advanced semiconductor fabs, EV factories, and EV battery and components factories, but also electrical equipment, etc.

Comments are closed.