Your flash reading for the September U.S. manufacturing PMI from S&P Global came in at an extraordinary 57.0, versus 53.9 last month and a forecast of 53.6. “All five components helped boost the PMI. Production growth revived after having waned over the prior three months, reaching its fastest since April 2022, as new orders growth also accelerated to the fastest in nearly four-and-a-half years. Jobs growth hit the highest since February 2021.” Anything above 50 suggests expansion of domestic manufacturing and this marks the 13th consecutive such reading. (The complementary Institute for Supply Management PMI has likewise signalled expansion since the start of the year and was likewise strong in its most recent August reading: 54.6, with “New Orders Growing; Production Growing; Employment Growing.”) It’s somewhat hard to believe how strong domestic industry looks given the headwinds from energy prices, but hopefully the progress continues.
Of course, one year ago in September, free traders were harping on data for the first few months after Liberation Day, confident that the initial data proved them right. I had just debated blogger Noah Smith, who insisted that PMI readings were the way to keep tabs on the sector’s health:
The purchasing managers index, the PMI is also below 50, which signals contraction in the industry, and this is Bloomberg reports on this Wall Street Journal reports on this, everyone reports on this as a reliable first indicator of how manufacturing is doing. It’s not perfect, but over the long term it’s pretty correlated with government statistics on how manufacturing is doing.
A few days later, the Cato Institute’s Scott Lincicome tweeted out the month’s ISM PMI and Noah mocked, “But Oren Cass told me that economics isn’t a science and that tariffs would revitalize American manufacturing! How could the economists be right after all? HOW???”
Scott and Noah don’t tweet about PMIs anymore, though, somewhat amusingly, Noah continues with his narrative arc regardless of what the data says. Last September, “U.S. manufacturing is getting f[*]d.” November, “Trump said he’d revive manufacturing, but his tariffs are strangling it.” December, “Manufacturing is dying.” January, “Trump is really f[*]g over American manufacturing.” April, “No, America is not in a ‘stealth manufacturing boom.’ All the numbers still look bad.” May, “Trump was not hired by China to destroy American manufacturing. But if he had been, what would he be doing differently right now?” August, “Trump’s tariffs are clobbering most of U.S. manufacturing (except data center related stuff).”
For what it’s worth, that August tweet does cite the July ISM PMI. But not the data. Only a few isolated comments from respondents. The data was a 55.6 reading with “New Orders Growing; Production Growing; Employment Growing.”
One cannot help remembering Iraq’s Military Spokesman:
“There are no American infidels in Baghdad. Never!”
Of course, another lesson from the triumphant capture of Baghdad is that early success is no guarantee of achieving ultimate objectives. We’ve got plenty of headwinds, the broader economy could turn down, and, as we always emphasize at American Compass, reindustrialization requires not only stable and credible tariffs, but also industrial policy and workforce policy and so on. But the past year should put to bed for good the idea once peddled so confidently that aggressive tariffs weaken the manufacturing sector. And we should remember going forward that people who stick with their narrative regardless of trends in their own preferred data may more be concerned with their narrative than with the data.
AND NOW FOR SOME ARTIFICIAL INTELLIGENCE
I had an argument recently with a prominent (and, notably, left-of-center) economist who thought it would be great if robots proved better than doctors and thus replaced humans entirely. I suggested that actually, something important would likely be lost, and we needed to guard against that. He replied that if human doctors still contributed something valuable, people would be free to choose that in the market. I questioned whether individual choices in the market really dictated available options in fields like health care.
This was on my mind as I read: U.S. Health Officials Move Quickly to Deploy Medical A.I. Despite Concerns (New York Times). The story has a lot of good examples about the kinds of choices we are going to have to make as a nation, through the political process, about when to trade human values for AI efficiency. Pay attention to the actual use cases described, which are never that the individual of means will be better served, but rather that the tools will offer ways to provide care to those who otherwise cannot afford it.
Venture capitalist Vinod Khosla says his son’s company “could offer an AI primary care provider that could give underserved people rapid access to care.” Ross Harper, CEO of Limbic, says deploying his AI therapist “is a moral issue to get this out into the hands of people who can’t access care. … We have to remind ourselves: What are we comparing this against? We’re comparing it against no treatment, a long wait list.”
On one hand, as a static analysis, it is correct that AI-only treatment might be an improvement over an alternative in which patients have no access to any care. On the other hand, it’s not hard to follow the mindset to its logical conclusion, which is that we don’t have to provide access to other care because we can offer AI-only treatments, and then to start searching for where else access to care can be cut because an AI-only treatment will suffice. This will start in government programs, and then move on to insurers. The “bronze” plan will be AI-only.
This dynamic will be pervasive across society. If we speak only in the language of efficiency and cost reduction, AI may often win out. It will be crucial to think instead in terms of where we want to see our human labor concentrated and where we don’t care if it has a role.
Speaking of bad health care incentives, the Times also has a good story on how the Battle of Hospital A.I. vs. Insurer A.I. Is Pushing Medical Costs Higher. This is another flavor of the problem I highlighted in my recent essay, where “ever more A.I.-generated job applications are being processed ever faster by A.I. screeners.” As in so many areas of the economy, we have to remember that incentives to deploy AI for rent-seeking can be higher than incentives to deploy AI productively, and will then trigger demand for counter-deployment, all of which can increase economic activity while reducing our well-being.
Jensen Huang Goes on a Podcast. Jensen went on Ezra Klein’s podcast and, as usual when Jensen goes on a podcast, made quite a hash of things. In his view, if the labs say “there is no way to contain our experiments, there’s just no way; when we test our A.I. models, it will get out, and it will damage the world — then I think the answer is that we have to shut the labs down.” This, of course, is what the labs are saying. So, um…
What he says next is funnier, though. The reason “we have to shut the labs down” is “because the cost to humanity, the damage is too great. The shareholder, the liabilities — it could be civil liabilities, it could be criminal liabilities. I mean, the liability’s incredible.” This gets the liability argument precisely backward. As noted last week, David Sacks has been advancing the argument that legal liability provides the appropriate check on the labs; because there is liability, labs have the right incentives not to do dangerous things. He quoted this guy named Jensen Huang saying, “The market forces are already there. We don’t need any new laws. We don’t need new regulations.” Now Jensen argues that the extraordinary liability for shareholders is exactly the reason we have to shut things down.
Demonstrating a fundamental misunderstanding of how people learn math and use it in their lives, he said, “the multiplication table is starting to be forgotten. ... Basic math is being forgotten. Does it matter?” And then answered his own question. “I don’t think it does. I don’t think it does.”
And of course, he talked up China. He called it “terrific” that we are using a lot of Chinese open models in the United States. His framing on chips was interesting: “The question is, ultimately, who are we depriving? Are we depriving China of a chip for their industry, or are we depriving the United States of a market to compete in?” Why the “or”? Obviously, the chip restrictions would do both. The difference is that, while we decide whether China gets the chip for its industry, China decides whether the United States will have its market to compete in. And we already know what they will decide.
Huang says the United States should have “a greater ambition for the world to be built on the American tech stack. Just as we have a greater ambition that the world is built on the U.S. dollar and that more people speak English — that they use the American version of the internet.” Insofar as we’ve had success with China welcoming the dollar, English, and the American version of the internet, I guess he does have a point.
SPEAKING OF CHINA, THERE WAS A SUMMIT THIS WEEK
The Wall Street Journal has a good rundown of Five Takeaways From Trump’s Washington Summit With Xi. (Is that a photo of Jensen Huang at the head table?) It looks like nothing much was agreed, which, as I wrote for the Financial Times about the last summit in May, is about the best we can hope for.
In the meantime, we continue to build up our own industrial capacity and decouple supply chains. Politico has a good story on Brendan Carr’s FCC, The Unlikely Agency Serving as Trump’s China Bad Cop.
Coming up next, a Chinese vehicles showdown with Rand Paul. Senator Bernie Moreno, sponsor of the Connected Vehicle Security Act, which would effectively ban Chinese cars and parts from the American market, says he will seek unanimous consent for the legislation on the Senate floor. The bill passed unanimously out of the Senate Commerce Committee and has more than 50 co-sponsors in the Senate and more than 100 in the House.
Moreno paused his plan on Thursday, likely because Senator Rand Paul (the likely objector) may be open to talking things over. Senator Paul’s home state of Kentucky produces more vehicles per capita than any other state in the country, so we shall see how far his passion for the “free market” extends to letting the Chinese Communist Party bankrupt his constituents’ employers with subsidized products.
And a Good Luck with That Award goes to EnerVenue, an “American” battery startup that chose to abandon its plan for manufacturing in Kentucky and set up shop in China instead. (Reuters: US battery startup that chose China over Kentucky opens first factory as Trump, Xi meet.) They’ve also accepted a major equity investment from a fund controlled by Hong Kong’s (i.e., China’s) government. In what way is this company American?
CEO Henning Rath says that with HQ in the U.S. and manufacturing in China, he is “combining the best of both worlds and building bridges even in a geopolitical difficult environment.” That may be the best of both worlds for Rath, but it is not the best of both worlds for the United States, which is precisely why U.S. policy must make it a bad one for his company.
Bonus deep cut: The South China Morning Post’s story on EnerVenue leads with a photo of a Kuka robotic arm, presumably in the new factory. Kuka, Kuka, why does that name ring a bell? Ah yes, because per the New York Times yesterday, In a Race With China, German Robots Are Outnumbered 14 to 1:
Once a symbol of German industrial innovation, the robot maker Kuka is now a source of regret. A decade ago, European regulators gave a Chinese appliance manufacturer, Midea, the green light to acquire Kuka, one of Germany’s leading robotics companies. Today Germany is failing to keep pace in the global race to deploy factory robots, which have been crucial to China’s transformation into the world’s manufacturing powerhouse.
It’s not hard to understand how the game is played, only to understand how we allow it to continue.
BY “WE,” I MOSTLY MEAN EUROPE
Of course, the United States isn’t allowing it to continue, though our response continues to be hesitant and halting, even though the status quo’s risks far exceed the risks of overcorrecting. In Europe, meanwhile, “Volkswagen ejected from European blue-chip index in blow to crisis-hit carmaker. VW’s market capitalisation of €38bn is just a fraction of its annual sales of €322bn.” Investors seem not to believe the continent is going to figure this out at all.
Someone who is figuring it out? The former head of the WTO. I recently highlighted Paul Krugman’s confession of error on free trade, and his remark that, “I’ve been shocked not only by my own change of mind, but by some of my colleagues.” One of those colleagues, it would appear, is former WTO head Pascal Lamy. Speaking to the South China Morning Post, Lamy still can’t help getting in his dig at the United States for irrationally pursuing protectionism, but this comes just moments before he admits Europe should go the same way:
The US has embarked on a protectionist crusade, which stems from the deep belief by US President Donald Trump that tariffs are the Swiss army knife of the 21st century economy, which is not so. The ones who pay the price for that at the end of the day are US consumers. … You are absolutely right, we Europeans are reaching a moment of truth with trade and China. Either we can convince China to rebalance the system and decrease its exports or we will have to go protectionist.
Only if you’ve been working in this world a long time can you appreciate the completeness of the transformation in thinking when the words “we will have to go protectionist” come out of Mr. Lamy’s mouth.
THIS IS REINDUSTRIALIZATION
Two very encouraging developments on the reindustrialization front, giving some dimension to the promising topline data. First: Lockheed Martin gets first batch of Patriot interceptor parts from General Motors (Defense News). “GM Defense delivered the initial batch of housing components for the Patriot interceptors called the PAC-3 MSE — on Aug. 28, just 22 days after the Aug. 6 contract signing, Lockheed said, for parts that traditionally take months or years to produce.” Three weeks from contract to delivery, from a commercial manufacturer. Reintegration of the defense-industrial base back into the full industrial base is crucial for both national security and a resurgent manufacturing ecosystem. Hopefully a sign of much more to come.
YCombinator partner Vivian Midha Shen reports that “atoms are in”: Hard-tech startups went from about 15% to almost 50% of the incubator’s Spring 2026 portfolio. That’s a very encouraging sign, and also a reminder that the path to rebuilding the American industrial base is not “tariffs on liberation day, taa daa millions of manufacturing jobs three months later,” but rather a shift in both economic incentives and social commitments that channel talent and capital for the long run. Suffice to say, the typical general equilibrium model used by economists does not have a variable for “excitement about building again.”
IN WHAT IS BECOMING AN ALARMINGLY REGULAR FINAL ITEM…
More bizarre developments from north of the border. Who knew: Even as Canada pursues fuller integration into the European community, ten EU member nations still have not ratified the decade-old free trade agreement that took provisional effect in 2017. How illiberal!
But nothing gets by Prime Minister Mark Carney. He’s even spending time gaming out a military invasion by the United States, saying “it would be irresponsible not to” (Politico). I’m pleased to report that I have also gamed out possible scenarios for Mark Carney invading my own house, and think I’ll likely be OK.
Seriously, though, Nick Phillips has a terrific piece at Commonplace on Canada Dreaming, which I’ll quote at length:
Carney’s popularity has surged as he single-handedly turns this perplexing trade dispute into a crisis and manufactures a grand trial for Canadian nationhood.
It seems that the real driver of these events is the psychic need for such a trial. Canada hasn’t had one in a long time, and prosperous liberal societies aren’t very good at generating them even as we crave thymos, the Greek term for the spiritedness that comes from struggle and purpose. Listening to a recent Francis Fukuyama interview about how liberal societies might generate thymos, one can’t help think of Carney: “Well, there are some possible futures where [liberal societies] actually do face real threats, where the society really does have to mobilize. I think we felt something like that in the 1930s with the Depression and then fighting World War II, and that was very inspiring to many people. It was also very devastating. And the question is, can you recreate that sense of national purpose without having a war or pandemic or revolution? I think you can. I think that there are a lot of projects out there in the world that could be the source of a greater sense of common purpose.”
In his 2021 book Values, Carney identified such a project: fighting climate change. Now, Carney appears to have found a new, more politically potent project: reorienting Canada away from the United States and realigning toward the EU…
Read the whole thing, and enjoy the weekend!
Your "left of center" economist believes that markets will provide anything the consumer wants in health care, which may be the most regulated and distorted market in all of America. That's funny, especially since, if he's really a Lefty, he's totally on board with using govt to alter people's economic and social choices in almost every other field. I wonder how he squares that circle? Perhaps the libertarian bent of his econ training (a problem I know a great deal about personally) has overridden his collectivist instincts on some issues, but the cognitive dissonance must be overwhelming at times.