It’s important to put this increase in context. When we measure electricity prices relative to how much people earn, we find that the recent rise in prices is pretty modest. In fact, so far it hasn’t even canceled out the big drop in electricity prices in the late 2010s:
But rising costs are rising costs, and it would be a bad idea to wait until things get really bad before we address the problem.
What’s more, this is the exact opposite of the direction that electricity costs ought to be going in. We’re in the middle of a miraculous revolution in energy technology. The cost of solar power and battery storage has absolutely plunged in recent years, to the point where developing countries like India are choosing renewables over coal simply because it’s cheaper to do so. Energy prices should be going down, not up. And in China prices are going down, in fact, despite an explosion in demand. Why is America different?
Everyone agrees that there are a bunch of factors at work. For example, here’s Wired:
There are several dynamics driving the current power price spike. Rising electricity demand, volatile fuel prices, inflation, tariffs, a slowdown in transmission line construction, and long delays in adding new generators to the power grid are all conspiring to create more expensive utility bills.
The Wall Street Journal also cites the increasing cost of rebuilding power infrastructure after natural disasters.
But one factor that lots of people cite is the AI boom. AI runs on computing power (or “compute”, as they say). You need compute to train models, and you also need compute to make AI “think” about the answers it’s giving. Scaling up this latter type — which is called “inference compute” — is now the most important way that AI companies are improving their models over time. And compute requires energy, because you’re running electricity through a bunch of chips in some data center somewhere. Every time you ask AI a question, it’s using some noticeable amount of electricity.
People go back and forth about how much the increased electricity demand from AI is already affecting electricity prices. A recent report from Lawrence Berkeley National Laboratory argues that AI hasn’t affected electricity prices yet, pointing out that states that saw greater electricity demand growth actually had falling prices over the past few years:
But Bloomberg Technology disagrees. Their analysis looks at smaller geographic areas, and considers wholesale rather than retail prices, and argues that data centers are already having big local effects:
A Bloomberg News analysis of wholesale electricity prices for tens of thousands of locations across the country reveals the effects of the AI boom on the power market with unprecedented granularity. The locations and prices were tracked and aggregated monthly by Grid Status, an energy data analytics platform. Bloomberg analyzed this data in relation to data center locations, from DC Byte, and found that electricity now costs as much as 267% more for a single month than it did five years ago in areas located near significant data center activity.
As the country builds a lot more data centers, we can expect this effect to filter through to retail prices, and start squeezing American consumers.
And remember that AI scaling tends to be exponential, meaning that AI is forecast to use a lot more compute in the near future. Here’s the EIA’s projection:
So right now, the rise in electricity costs has actually been modest, as has the effect of AI on those costs. But if AI keeps scaling up, things are going to get very hairy in the next decade. And that’s not even counting the increased electricity usage from switching from internal combustion cars to EVs.
It therefore seems like scaling up electricity generation in the U.S. is an incredibly important and pressing task. We don’t want a future in which AI outcompetes humans for scarce electric power and forces them to live in the dark and cold. Nor do we want a future in which America sacrifices the most crucial high-tech industry simply because we’re unwilling to build power lines, solar panels, and batteries. What we need, now more than ever before, is energy abundance.
And yet despite a few promising initiatives, the policies of the Trump administration are not geared toward providing America with energy abundance. In many cases they’re actually going to prevent us from getting the energy we need. Trump’s whole approach to energy seems like a mess of conflicting values, ideas, and policy initiatives.
A week ago, I wrote a post about the possibility of a bust in the AI sector:
One aspect of the story I didn’t talk about was the web of “circular” deals between AI model makers, compute providers, and chipmakers. Almost every news story or blog post that worries about the threat of an AI bust mentions these. Here are some excerpts from a good overview of the situation by Emily Forgash and Agnee Ghosh in Bloomberg:
Two weeks ago, Nvidia Corp. agreed to invest as much as $100 billion in OpenAI to help the leading AI startup fund a data-center buildout…OpenAI in turn committed to filling those sites with millions of Nvidia chips. The arrangement was promptly criticized for its “circular” nature…This week…OpenAI…inked a partnership with Nvidia rival Advanced Micro Devices Inc. to deploy tens of billions of dollars’ worth of its chips. As part of the tie-up, OpenAI is poised to become one of AMD’s largest shareholders…
OpenAI confirmed it had struck a separate $300 billion deal with Oracle to build out data centers in the US. Oracle, in turn, is spending billions on Nvidia chips for those facilities, sending money back to Nvidia, a company that is emerging as one of OpenAI’s most prominent backers…Nvidia plans to invest up to $2 billion in equity in Elon Musk’s xAI in a financing round tied to Nvidia’s chips…
[Nvidia] agreed to buy $6.3 billion worth of cloud services from CoreWeave, which rents out access to Nvidia chips. OpenAI, meanwhile, received $350 million in equity from CoreWeave ahead of the IPO and recently expanded its cloud deals with the company to as much as $22.4 billion.
Some people worry that this web of relationships either A) could exacerbate a bust, or B) is a sign that a bust is coming. Here’s an example of the former, from that same Bloomberg article:
[S]ome analysts and academics who’ve tracked the tech industry long enough see uncomfortable similarities to the dot-com bubble. “In the late 1990s, circular deals were often centered on advertising and cross-selling between startups, where companies bought each other’s services to inflate perceived growth,” said Paulo Carvao, a senior fellow at the Harvard Kennedy School who researches AI policy and who worked in tech in the late 1990s. “Today’s AI firms have tangible products and customers, but their spending is still outpacing monetization.”
Third, there is what Azeem Azhar calls a “financial ouroboros” now taking place in AI: what looks like revenue generated by sales is, in some cases, really just the same stock of money going in circles between the various AI companies…[G]iven the pace at which capacity is expanding, one can’t dismiss concerns that circular flows of money are a warning sign.
So let’s think about why these circular deals could be a problem.
Basically, the supply chain for AI looks like this:
chipmakers —> cloud providers —> AI companies
Chipmakers like Nvidia make chips and sell them to the cloud providers like Microsoft, who operate the data centers that train and run models for the AI companies like OpenAI. OpenAI pays Microsoft for compute, Microsoft pays Nvidia for the hardware to generate that compute. OpenAI is currently trying to cut out the middleman, creating its own cloud computing capacity.
Many of the circular deals take the form of an upstream company buying the stock of a downstream company, who then uses that cash to buy products from the upstream company. For example, Nvidia buys stock in OpenAI, which buys GPUs from Nvidia.1
As far as I can tell, there are two main fears about this sort of deal. The first is that the deals will artificially inflate companies’ revenue, tricking investors into overvaluing their stock or lending them too much money. The second is that the deals increase systemic risk by tying all of the AI companies’ fortunes to each other.
Let’s start with the first of these risks. The question here is whether AI’s circular deals are an example of round-tripping or vendor financing.
Suppose two startups — let’s call them Aegnor and Beleg2 — secretly agree to inflate each other’s revenue. Aegnor buys ad space on Beleg’s website, and Beleg buys ad space on Aegnor’s website. Both companies’ revenues go up. They’re not making any profits, and they’re not generating any cash flows, because the money is just changing hands back and forth. But if investors are looking for companies with “traction”, they might see Aegnor and Beleg’s topline revenue numbers go up. If they fail to dig any deeper, they might give both companies a bunch of investment money that they didn’t earn. This is called “round-tripping”, and it happened occasionally during the dotcom boom.
Now what I just described is completely illegal, because the companies colluded in secret. But you can also have something a little similar happen by accident, in a perfectly legal way. If there are a bunch of startups whose business model is selling to other startups, you can get some of the “round-tripping” effect without any collusion.
On the other hand, it’s perfectly normal and healthy for, say, General Motors to lend its customers the money they use to buy GM cars. In fact, GM has a financing arm specifically to do this. This is called vendor finance. It’s perfectly legal and commonplace, and most people think there’s nothing wrong with it. The transaction being financed — a customer buying a car — is something we know has value. People really do want cars; GM Financial helps them get those cars.
So the question is: Are the AI industry’s circular deals more like round-tripping, or are they more like vendor finance? I’m inclined to say it’s the latter.
First of all, in deals like Nvidia’s relationship with OpenAI, the revenue is only flowing one way. OpenAI pays Nvidia for chips. OpenAI needs chips to perform its core business, which is selling AI services. And Nvidia’s core business is selling chips, especially to people who provide AI services. Both companies are just doing what they’re set up to do.
Second, these are all big companies that come under intense public scrutiny. Nvidia is a public company, with all of the disclosure requirements and accounting regulations that that entails. While it’s possible that foolish investors might look only at Nvidia’s revenue and not its cash flows when deciding to lend the company money or buy its stock, it seems unlikely that it would make a huge difference.
In fact, if you look at Nvidia’s stock price, it doesn’t seem to have risen as a result of the big OpenAI deal — in fact, it’s about the same as it was in August, and showed no visible “pop” after the deal was announced in late September:
This really looks to me more like a case of vendor finance. OpenAI is a big company, but it’s not a public company, which limits its ability to borrow and to sell its stock. Nvidia is a public company, and right now it’s the most valuable company on the planet, with a market cap around 9x that of OpenAI. That makes it a lot easier for Nvidia to borrow money3 — i.e., its cost of capital is lower. It makes a lot of financial sense for Nvidia to use its low cost of capital to support a more capital-constrained customer. As Azeem Ahar puts it:
Customer demand is fierce—and it’s speeding up. Anthropic’s annualized revenue leapt from $1 billion to $5 billion in under six months, a fivefold jump that signals just how hungry enterprises are for AI capabilities; Google confirmed that demand for its AI services nearly tripled between May and October this year, with token consumption surging across every major product line…
How can OpenAI possibly match this blistering pace? Nvidia and the hyperscalers have the strongest balance sheets and can effectively use those to expedite the buildout of capacity.
Now this doesn’t mean deals like this are safe or a good idea — it only means that these deals are unlikely to represent “funny money”.
What about the second risk of circular deals? The more of these deals there are, the more the fate of each of the companies involved depends crucially on what happens to the other companies in the chain. If OpenAI’s business model fails, Nvidia takes two hits — it’ll lose one of its biggest customers, and it’ll also take a big loss on its equity investment.
Nvidia was already making a huge bet on the AI industry. Almost all of Nvidia’s revenue comes from data centers at this point, rather than from gaming or other lines of business — and data centers are almost all about AI now. If there’s a giant AI bust, Nvidia will take a huge hit, whether or not it has to write down its stock in other AI-related companies. Basically, with these deals, Nvidia is making a leveraged4 bet on something that it’s already making an existential bet on.
And Nvidia isn’t alone here. If AI has a big crash, every single company in the circular deal diagram is going to be either toast or in huge trouble, whether or not they have circular deals in place. The fates of all of these companies are already closely tied together, via their mutual dependence on a single technology. Correlations are already very high here. Tying the companies even more closely together with equity purchases doesn’t really change what will happen if AI itself goes bust.
What about counterparty risk? All of these companies are utterly dependent on AI, but that’s different than being dependent on one single other company. Even if AI succeeds wildly, Open AI might fail as a company. So we don’t want Nvidia to become too dependent on OpenAI, because Nvidia’s position as the top AI chip company is too valuable to risk.
This is a real danger. But it’s not clear that the circular deals make things worse — instead, they might be letting AI companies diversify their dependencies. In the first quarter of this year, more than half of Nvidia’s revenue came from just four big mysterious customers — probably Amazon, Microsoft, and Google, and possibly Meta. Building up demand from OpenAI could help Nvidia be less dependent on those big customers.
In fact, I suspect that this is what’s actually going on with these circular deals — diversification. All of these companies are existentially dependent on AI itself, but they don’t want the added vulnerability of being dependent on specific AI companies. So they’re investing in as many of the other big AI companies as they can, in order to diversify away their company-specific risk — just like when you buy a bunch of stocks in order to make your portfolio safer.
Diversification reduces risk. Leverage increases risk. AI companies are (effectively) borrowing money in order to double down on their risky bet on AI itself. But in doing so, they may also get a bit of diversification. We’ll end up with a system that’s more vulnerable to an AI crash, but less vulnerable to the failure of any specific AI company.5
Essentially, I think the story of these circular deals is really no different than the big macro story about the AI boom itself. AI is the most promising technology out there, and we’re simply betting a lot of our economy on the hope that we can exploit its full potential.
Interestingly, some of the deals go in the opposite direction. OpenAI is buying AMD’s stock and AMD’s chips at the same time. The fact that this is characterized as a “circular deal” similar to the Nvidia-OpenAI deal shows that the term “circular” is being thrown around a bit sloppily.
I can’t actually find it now, but I remember in the mid 2000s, reading some article about the rise of Islamism in Pakistan. I remember one general saying: “All the young guys have beards.” It reminded me of the fateful words attributed to King Louis XV: “Après moi, le déluge.”
I thought of that line when I read about the texts recently leaked from the Young Republicans’ group chats:
Leaders of Young Republican groups throughout the country…referred to Black people as monkeys and “the watermelon people” and mused about putting their political opponents in gas chambers. They talked about raping their enemies and driving them to suicide and lauded Republicans who they believed support slavery.
The leaks led to a backlash from the GOP, with the New York and Kansas chapters of the Young Republicans getting shut down, a Vermont state senator stepping down, and a handful of other participants losing their jobs. (JD Vance didn’t join in, making excuses for the “kids” in the chat group, even though they were in their late 20s or 30s).
It’s good to see that the institutions of the Republican Party still have enough power — and enough of a conscience — to crack down on things like this, at least a little bit. But it’s unlikely that official censure or condemnations will stem the trend toward authoritarianism and racial hatred among the party’s younger members. The leaked chats are not even slightly surprising for anyone who has lurked in online right-wing spaces and discussions over the past few years.
In fact, if anything, what’s surprising is that the mainstream media seems to have been so blindsided by texts that were so tame compared to what gets said on public forums like X and 4chan every day. Do people really not know that this is what young right-wingers are like now? On social media, there is a lot more unabashed Hitlerism than in the Young Republicans’ group chat. Popular right-wing accounts now regularly ridicule the widespread belief that Hitler was evil as a “religion” or a “myth”:
Even the Young Republicans in the offensive chat group talked about the trend, acknowledging that the younger people in the party are fans of the Nazis. The leaked chats’ Hitler talk was mostly in the form of jokes and sarcasm, but Politico reports that “the group chat members spoke freely about…the love of Nazis within their party’s right wing”.
In fact, attempts to rehabilitate Hitler are becoming more common all across the right-wing media ecosystem. A year ago, I wrote about Tucker Carlson’s embrace of a revisionist historian who calls Winston Churchill the true villain of World War 2, claims that Hitler actually wanted peace, and declared that Hitler conquering France was preferable to a modern drag show.
What’s going on here? I think the new trend toward Hitler apologia on the right probably represents the confluence of a few different trends.
First, the World War 2 generation has now mostly passed away. With them has gone both the eyewitness accounts of Nazi horrors, and also some portion of America’s pride in being the country that triumphed over Nazi Germany. There is no longer a large contingent of American voters and respected elders who will be personally offended and horrified if you crack a Hitler joke or engage in revisionist history about WW2.
Also, the rise of the Palestine movement on the left probably contributed to the trend. Although leftists certainly don’t like Hitler, the deep antisemitism of the Palestine movement — which tends to view Jews as presumptive Zionists unless they prove their innocence via anti-Israel activism — has effectively kicked Jews out from under the protective umbrella of progressive pro-minority activism. That gave rightists a green light to unleash their own much more virulent antisemitism without fear of leftist attack.
But most importantly, I think social media rewards extremism and punishes moderation within each party. You can see this dynamic at work in the response to the Young Republicans’ leaked texts. In the old days, when internal party debates mostly happened in private, it was possible to crack down on Nazis and other extremists. Now, thanks to social media, those debates mostly happen in public, where the opposition can see everything and potentially take advantage of any internal divisions. This makes right-wing leaders extremely reluctant to cultivate the appearance of dissension in the ranks, for fear of joining in a “left wing pile on”:
In my experience, this dynamic is similar to the one that prevailed among Democrats in 2020 and 2021. If I criticized things like “defund the police”, progressives who would almost certainly agree with me in private would demand that I not give ammunition to right-wingers and racists by publicly criticizing progressives. Any attempt to restrain the worst excesses of the extremists was seen as giving aid and comfort to the enemy. And so the worst extremists ran rampant — as they are now running rampant on the right.
Social media probably boosts extremism in other ways. Extremist sentiment tends to go viral more easily, and get more clout and attention, than moderate content. Social media apps like X expose Americans to the ideas of people outside their polity — including places like Pakistan, where Hitler is not as deeply reviled — without Americans knowing where those ideas are coming from.1
And by boosting extremists, social media exposes everyone to maximum threat from the other side; when some random clout-chasing pseudonymous progressive calls normal Republicans “Nazis” for restricting immigration, those Republicans may feel like the whole progressive movement is calling them “Nazis”. This may make some of them simply give up on the idea of policing their actual Nazi extremists in order to maintain respectability, because they may feel that this is a lost cause.
Anyway, the upshot here is that the combination of the fading memory of WW2, the indirect influence of the Palestine movement, and the corrosive extremism of social media are allowing the worst voices on the American right to rehabilitate Adolf Hitler. That is extremely bad; Hitler richly deserves his reputation as the most terrible villain of modern history, and the truth of his rule is even more awful than the popular caricature.
Right now, the GOP and the rightist movement are dominated not by Nazi ideology, but by the personal charisma of one towering figure — Donald Trump. Some progressives may be angry at me for saying this, but for all his authoritarian impulses, Trump is not a Nazi or a Hitler figure. He’s a corrupt personalist Peronist who would love to be a dictator if he could, but he bears no allegiance to Nazi racial theories, and the policies he wants bear little resemblance to Hitler’s.
But Trump is a very old man — the oldest President we’ve ever elected, even older than Biden was at this point in his term. He’ll be 82 by the time his term finishes. Humans are mortal, and aging means cognitive decline (which is already starting) and, eventually, death. Trump will shuffle off this mortal coil, and even before he does, he will cease to be coherent or sharp enough to control the GOP or determine the direction of the MAGA movement.
There is simply no one else on the political Right whose charisma comes even close to Trump’s. If you think JD Vance is a Trump-like figure, just watch him ordering some donuts.
In the absence of a towering charismatic figure, the movement that Trump built will have to be held together by ideology. Figures like Vance or Stephen Miller won’t be able to rely on personal charisma to hold together and direct the movement that Trump bequeaths to them, so instead they’ll turn to more typical, pedestrian expedients. That will mean ideology.
Now we are starting to get an idea of what kind of ideology will be necessary to hold together the MAGA coalition. The idea of the Great Replacement — that immigration is a plot to subjugate both the White race and the Republican party — will be absolutely core to that ideology. Now we’re learning that World War 2 revisionism will be important as well. Rightists believe that Hitler’s place as the Great Satan in America’s folk memory gives liberals and leftists a moral trump card, and so they feel like they need to deemphasize the evil of the Nazis in order to level the political playing field.
Antisemitism may also be a way for the MAGA movement to retain some of its surprising levels of support from minorities without the aid of Trump’s brash charisma. Research shows that the most antisemitic demographics in the U.S. are young conservative Black and Hispanic men. Matt Yglesias read this research and summarized it thus:
The epicenter of antisemitic attitudes in the United States, in other words, is the conservative Black and Hispanic population that has often voted Democratic in the past due to identity politics but has trended toward the GOP in recent cycles. Liberal African-Americans are slightly less antisemitic than white conservatives, and Black and Hispanic conservatives are substantially more antisemitic than white conservatives.
With its immigration sweeps and urban crackdowns, MAGA is not in a great position to retain minority votes. But hating on Jews may allow them to hold on to a few foolish guys who don’t realize that Hitler would have sent them to the gas chambers too:
In other words, the MAGA movement is probably going to become more Nazi-adjacent and Nazi-apologetic when Trump’s personal influence fades. That’s obviously extremely scary, because the Nazis well deserve their reputation — it’s hard to think of a worse ideology for an American political movement to embrace. Americans who love freedom will have to fight very hard to resist this ideology.
But at the same time, the struggle against MAGA will get a lot easier when Trump is gone. It’s a lot easier to fight against an army of grim, cruel Hitler apologists than it is to battle a charismatic populist, and basically no demographic group in America is actually majority antisemitic. Things will probably get worse before they get better, but America is unlikely to stand for actual Nazism.
I was waiting to write about the new trade war between China and the U.S., because there was always the chance that it would be resolved within a day by a “deal” in which both China’s leaders and Donald Trump retreat from their threats and business goes on as usual. In fact, that outcome is still very possible, but it’s been over a week now, and the situation hasn’t been resolved, so I guess it’s time to write about it.
Basically, what happened is that China slapped extremely stringent new export controls on rare earth metals, in an attempt to extract concessions from Trump and the United States:
China has unveiled broad new curbs on its exports of rare earths and other critical materials…Overseas exporters of items that use even traces of certain rare earths sourced from China will now need an export license…Certain equipment and technology for processing rare earths and making magnets will also be subject to controls…[China] later announced plans to expand export controls to a range of new products…[these] five more rare earths — holmium, europium, ytterbium, thulium, erbium — plus certain lithium-ion batteries, graphite anodes and synthetic diamonds, as well as some equipment for making those materials.
Trump immediately responded with bellowing bravado, announcing new 100% tariffs on Chinese goods, as well as various new export controls. Trump’s treasury secretary, Scott Bessent, joined in, calling China’s trade negotiators “unhinged”.
But just a few days later, Trump and Bessent were already backing down in the face of China’s threats. Trump admitted that 100% tariffs on China were “not sustainable”, and declared that “[W]e’re doing very well. I think we’re getting along with China.” Bessent offered a “truce” in which the U.S. suspends tariffs on China in exchange for China suspending its threat of export controls.
The most likely outcome, therefore, is that China simply wins this round of the trade war, as it won the last round. In April, Trump announced big tariffs on China. China retaliated by implementing rare earth export controls, causing Trump to back down and reduce tariffs to a low level. But China didn’t reciprocate — it kept its export controls in place, allowing America to keep buying rare earths only through some short-term conditional arrangements. China then used these controls to extract even more concessions from the hapless Americans:
Trump ratcheted up his new tariffs on imports from China to 145% in April…China responded by raising its new duties on shipments from the US to 125%…Before the economic pain of these sky-high levies really started to bite…The US reduced its new levies on Chinese imports to 30%…China agreed to match the US reciprocal rate, lowering its new tariff on American goods to 10%…
American officials were under the impression this would entail a lifting of Chinese restrictions on the export of rare earths…But China’s government continued to limit the issuance of export licenses…A framework deal was then agreed in June, under which China agreed to issue export permits for “controlled items” that “meet the conditions in accordance with the law.” In return, the US agreed to lift countermeasures it had introduced targeting exports on products such as ethane that’s used to make plastic, chip software and jet engines.
This was also entirely consistent with the pattern of Trump’s first term, in which he agreed to suspend planned tariffs on China in exchange for empty promises of agricultural purchases that China never ended up keeping. It fit the common caricature of Trump as a cowardly bully who acts with extreme aggression toward weak opponents, but who retreats from any rival who stands up and hits back.
If the pattern holds this time, then Trump will retreat from his threats of sky-high tariffs, but China will keep its new export controls in place. Lingling Wei and Gavin Bade report that China’s leaders believe they have the American President over a barrel:
In its trade standoff with Washington, Beijing thinks it has found America’s Achilles’ heel: President Trump’s fixation on the stock market…China’s leader, Xi Jinping, is betting that the U.S. economy can’t absorb a prolonged trade conflict with the world’s second-largest economy, according to people close to Beijing’s decision-making. China is holding a firm line because of its conviction, the people said, that an escalating trade war will tank markets, as it did in April after Trump announced his Liberation Day tariffs…The market’s sharp negative reaction on Friday to China’s new rare-earth restrictions and the potential U.S. retaliation served as a reminder of the economic vulnerability Beijing seeks to exploit.
Meanwhile, op-eds in Chinese state media portray the U.S. as weak and irresolute. This is the TACO trade, but it probably isn’t just that. Dictatorships — and China now truly deserves to be called by that name — tend to flatter themselves with the idea that their unity of command gives them a consistency and willpower that democracies, enslaved to their fickle electorates, naturally lack. That assumption proved disastrously false for the Axis and the communist bloc in the 20th century, but with American society divided by various political and social conflicts, it might prove correct this time.
If so, that’s very bad news for the United States — and for the democratic world in general. If China learns that its control of rare earths is a trump card that it can use to extract anything it wants from other industrialized countries, it will push that advantage as far as it can. After surrender on trade issues, the obvious next set of demands is geopolitical — control of Taiwan, dominion over the South China Sea, U.S. troops and ships out of Asia, and so on.
Another possible move for China, instead of making demands, is to simply cut other industrialized countries off from rare earth supplies entirely. If other nations weren’t able to manufacture electronics or other high-tech products, it would allow the Chinese leadership to fulfill their vision of China being the only country that does a significant amount of high-tech manufacturing. This would deindustrialize America, Europe, Japan, and India, reducing them to selling natural resources and services.
A few years ago I predicted that something like this would be part of China’s long-term strategy. The threat of forced deindustrialization was one of my arguments for why the U.S. shouldn’t shy away from Cold War 2:
Ultimately, if China gains control of the global economy, it will choose to use that control to weaken the U.S. Its leaders rightfully see American power as a threat to Chinese power, and would seek to use things like interdiction of seaborne trade and strategic control of resources to reduce that threat by weakening the U.S. economically. It would similarly work to economically weaken key American trading partners like Japan, South Korea, and the EU, in order to reduce their power as well; this would have negative knock-on effects for the U.S. economy.
A poorer U.S. would not be a boon for China economically, but the country’s leaders clearly don’t think in purely economic terms…Minimizing China’s ability to strategically impoverish the U.S. is an important reason to resist Chinese power instead of simply acceding to it.
This prediction is coming to pass even earlier than I expected.
So it will be an incredible shame if Trump does end up backing down from this new trade war. And it will be doubly a shame, because this trade war is eminently winnable. It won’t be easy, but if the U.S. really commits to doing bold and smart policies, we can end our dependence on China for rare earths and other minerals, and thus break the economic yoke that China has placed around our neck — and the whole world’s neck.
Let’s talk a bit about how we could do that.
The U.S. could become independent in rare earths if it really wanted to
Hi, folks! I haven’t done one of these roundups lately, so I’m overdue!
Econ 102 is moving to one episode every two weeks, since Erik is very busy starting his new job at a16z. But never fear, I have some other projects in the works, for those of you who enjoy hearing my voice. In the meantime, here’s one Econ 102 episode, which itself was delayed due to legal issues switching the podcast over to a16z.
Anyway, on to this week’s list of interesting things, which is mainly just me telling people they’re wrong about stuff.
1. Doomerism is so passé
I don’t see nearly as many pessimistic screeds these days as I did a couple of years ago. But every once in a while, one pops up on my screen, and it’s invariably just as maudlin and overwrought as they all are. The latest example is an essay in the New York Times by Andreas Reckwitz from Berlin, entitled “The West is Lost”.
Why is the West lost? Well, fairly predictably, he starts out with climate change:
The most dramatic loss is environmental. Rising temperatures, extreme weather, disappearing habitats and the ruination of entire regions are eroding the conditions of life for humans and nonhumans alike. Even more threatening than present damage is the anticipation of future devastation — what has aptly been termed climate grief. What’s more, mitigation strategies themselves promise losses: a departure from the consumer-oriented lifestyle of the 20th century, once celebrated as the hallmark of modern progress.
Look, climate change is obviously a bad thing, but let’s not exaggerate. The notion that runaway climate change was going to make the Earth uninhabitable was always highly dubious; in 2021, the author of The Uninhabitable Earth wrote an article entitled “After Climate Alarmism”, in which he embraces more balanced, reasonable forecasts. Thanks to the amazing progress in renewable technology, and China’s gung-ho willingness to scale that technology rapidly, the world will probably be spared the worst. And the idea that climate change is going to be solved — or even meaningfully altered — by pious Europeans eschewing modern consumer lifestyles is just numerically illiterate. Technology is solving the problem while German intellectuals wring their hands.
Next up, Reckwitz tells us that we’re experiencing economic devastation:
Economic changes have also brought loss. Entire regions once defined by prosperity — Rust Belt America, the coal fields of northern England, small-town France, eastern Germany — are now locked in decline. The optimism of the mid-20th century, when upward mobility seemed the natural way of things, has proved exceptional rather than typical. It was, it turns out, a historical interlude. Deindustrialization and global competition have fractured societies into winners and losers, with large segments of the middle class seeing their security erode.
I wrote a whole post debunking the notion that globalization hollowed out the American middle class:
In fact, middle-class wages have risen almost as strongly in the “neoliberal” era as they did in the glorious three decades after WW2:
Reckwitz also claims that infrastructure is deteriorating in the U.S. No it isn’t. The American Society of Civil Engineers has upgraded the country’s infrastructure “report card” in recent years. Yes, American infrastructure costs way too much, but the government ponied up the cash.
Now, there are some major problems the West (defined here as America, the Anglosphere, and Europe) has yet to really address. These include low fertility, low trust in government, and housing shortages. America’s politics are certainly not encouraging. But every civilization, at every time in its history, faced major problems. With the possible exception of low fertility — which no one knows how to reverse — all our problems are eminently solvable.
I think many American progressives take it as almost an article of faith that if we exaggerate the severity of our problems, it’ll galvanize society to greater action. But op-eds like this one demonstrate that such pessimistic exaggeration simply paralyzes people into hand-wringing helplessness.
2. Indian entrepreneurs aren’t very clannish
In 2025, I’ve started to hear a lot more anti-Indian sentiment, including in some fairly elite tech circles. I talked about it in a post last month:
And as for self-dealing and ethnic exclusivity, I’m starting to hear this charge thrown around a lot as well, on social media and even in some San Francisco parties.4 The X user “Power Bottom Dad”, a harsh critic of H-1bs, has identified at least two lawsuits in which a big American company was accused of preferentially hiring Indians or discriminating against U.S. citizens. Whether that is a general pattern, as Power Bottom Dad asserts, or a couple of isolated incidents that could be found from time to time among any ethnic group in America, is probably irrelevant to the politics of the issue; what matters is that there is a portion of Americans out there who see Asian immigrants, and particularly Indian immigrants, as a clannish self-dealing cartel.
Well, I recently discovered that the economists Sari Kerr and William Kerr actually wrote a paper in 2021 that challenges some of those stereotypes. Here’s their abstract:
We explore co-ethnic hiring among new ventures using U.S. administrative data. Coethnic hiring is ubiquitous among immigrant groups, averaging about 22.5% and ranging from <2% to >40%. Co-ethnic hiring grows with the size of the local ethnic workforce, greater linguistic distance to English, lower cultural/genetic similarity to U.S. natives, and in harsher policy environments for immigrants. Co-ethnic hiring is remarkably persistent for ventures and for individuals. Co-ethnic hiring is associated with greater venture survival and growth when thick local ethnic employment surrounds the business. Our results are consistent with a blend of hiring due to information advantages within ethnic groups with some taste-based hiring.
So clannishness among immigrant groups is common — which is not surprising, given that people tend to hire out of their personal networks, and immigrants know a lot of other immigrants. But as immigrants go, Indians are below average:
Why only look at new firms? Because A) it’s a lot harder to know who’s responsible for hiring in bigger, older firms, and B) since there’s a lot more discretion in hiring at startups and small businesses than at big companies, you’d expect new firms to show much more pronounced clannishness.
So why are Indians less clannish than Poles, Koreans, etc.? Probably because they usually speak at least pretty good English when they come over. As Kerr and Kerr note, the better that immigrants can speak English, the more easily they can form bonds of trust and cooperation outside of their ethnic community.
Anyway, I don’t expect this research to move the needle for the people who have already convinced themselves that Indians are especially clannish in the workplace. But for everyone else, I hope it adds some much-needed balance and perspective. And I hope it reveals that the impulse to blame social ills on Indians shows every sign of being a classic panic rather than a sober assessment of the effects of immigration.
3. Why are Nobel laureates getting older?
We just had the Nobel Prize announcements, and some people have noticed that in physics, chemistry, and medicine, the laureates seem to be getting older and older over time:
What conclusions should we draw from this trend? One possible interpretation is that recent discoveries haven’t been very groundbreaking, so a lot of the most important scientists did their work decades ago. People have tried to investigate this idea, and the evidence is generally pretty split.
But you can also make the exact opposite interpretation! Perhaps groundbreaking discoveries are actually becoming more common, so that there’s a large “queue” of scientists waiting for their prizes. The Nobel can’t be given posthumously, so older candidates naturally go to the front of the queue, so that they can get their prizes before they die.
Now here’s a subtle idea. A “queue” could also form if there’s a compression in the upper tail of the distribution of scientific discoveries. Suppose that these days, modern science is making more discoveries that are a 9/10 in terms of importance, but fewer 10s and 8s and 7s. That could improve the overall pace of scientific progress, while making the prize-winners seem less stand-out and special.
There are also neutral interpretations, like the “burden of knowledge” idea — the idea that as scientific knowledge accumulates over the decades and centuries, it takes more and more time to get up to speed on cutting-edge ideas. Maybe scientists are making just as many big breakthroughs as before, but now they’re forced to wait until they’re older in order to make them.
So really, we don’t know what’s behind this trend, or whether it’s good or bad.
4. Why do we still have so many radiologists in the age of AI?
A whole lot of people think that AI is going to eliminate many human jobs. Perhaps no one is more confident of this prediction than the engineers who are actually building the AI in question. For example, nine years ago, Geoffrey Hinton — one of the key creators of modern AI — declared we should stop training radiologists, because in five to ten years, they would be replaced:
“I think if you work as a radiologist, you are like the coyote that’s already over the edge of the cliff but hasn’t yet looked down…People should stop training radiologists now. It’s just completely obvious within five years deep learning is going to do better than radiologists…It might be 10 years, but we’ve got plenty of radiologists already.”
ChatGPT hadn’t yet been invented, but the AI systems of 2016 were already pretty good at reading medical imaging. Hinton felt that he could see the writing on the wall.
And yet here we are, nine years later, and the radiology profession is doing just fine. Deena Mousa recently wrote about this in an excellent Works in Progress article:
Radiology is a field optimized for human replacement, where digital inputs, pattern recognition tasks, and clear benchmarks predominate…But demand for human labor is higher than ever. In 2025, American diagnostic radiology residency programs offered a record 1,208 positions across all radiology specialties, a four percent increase from 2024, and the field’s vacancy rates are at all-time highs. In 2025, radiology was the second-highest-paid medical specialty in the country, with an average income of $520,000, over 48 percent higher than the average salary in 2015.
As Mousa notes, some of the reasons for this are pedestrian. AI models don’t perform nearly as well once they get out in the real world and have to go far beyond their training data. And humans don’t trust the AI, so regulators and health insurers sometimes insist on human radiologists.
But there are other reasons for the persistence of the radiology profession that are even more profound — and that should serve as important reminders about how little we know about the economic effect of AI.
First, AI model makers don’t actually know what radiologists do. They know radiologists read scans, but they don’t know all the other stuff they do:
Radiologists are useful for more than reading scans; a study that followed staff radiologists in three different hospitals in 2012 found that only 36 percent of their time was dedicated to direct image interpretation. More time is spent on overseeing imaging examinations, communicating results and recommendations to the treating clinicians and occasionally directly to patients, teaching radiology residents and technologists who conduct the scans, and reviewing imaging orders and changing scanning protocols.This means that, if AI were to get better at interpreting scans, radiologists may simply shift their time toward other tasks. This would reduce the substitution effect of AI.
Because AI engineers don’t really understand all the things radiologists do, they will be slow to design AI systems that address all of these tasks. And even when someone does get around to addressing this problem, it’s not clear when we’ll get AI that’s as good at humans at all of these tasks. Humans may remain in the loop.
Finally, AI increases productivity, which reduces cost, which increases the number of patients who can be served. This increases the demand for radiologists’ labor:
As tasks get faster or cheaper to perform, we may also do more of them. In some cases, especially if lower costs or faster turnaround times open the door to new uses, the increase in demand can outweigh the increase in efficiency, a phenomenon known as Jevons paradox. This has historical precedent in the field: in the early 2000s hospitals swapped film jackets for digital systems. Hospitals that digitized improved radiologist productivity, and time to read an individual scan went down. A study at Vancouver General found that the switch boosted radiologist productivity 27 percent for plain radiography and 98 percent for CT within a year of going filmless. This occurred alongside other advancements in imaging technology that made scans faster to execute. Yet, no radiologists were laid off.
Instead, the overall American utilization rate per 1,000 insured patients for all imaging increased by 60 percent from 2000 to 2008. This is not explained by a commensurate increase in physician visits. Instead, each visit was associated with more imaging on average.
AI engineers, who look mainly at their models’ capabilities, don’t generally think a lot about this overall economic ecosystem. They’ve seen their creations up close, and they know their capabilities better than anyone else, but that doesn’t mean they know what a radiologist does at work. And they don’t know whether AI will simply change how radiologists spend their time at work while also improving their productivity. Hinton might some day be right, but as of right now he was wrong.
5. Is Europe starting to stand up to China, just a little bit?
For a long time, Europe has basically ignored every threat that came out of China. It has ignored the increasing evidence that China is supporting Russia’s war effort in Ukraine, which means that China is waging proxy war against Europe itself. And it has taken only modest measures against the flood of Chinese government-subsidized imports that threatens to deindustrialize and hollow out Europe’s manufacturing industries.
But while Europe is still in too parlous and precarious a state to do much about China’s military threat, it’s at least starting to push back on China’s cutthroat economic competition. For example, some Chinese companies have been trying to get around Europe’s tariffs by building factories in Europe. But Europe is unsatisfied with the prospect of having its people simply work for Chinese bosses; instead, the EU is now insisting that Chinese investments transfer technology to local European companies:
The European Union is poised to toughen its trade stance towards China, with the bloc set to insist that Chinese investors in Europe transfer technology to local firms…Trade chief Maros Sefcovic insisted on Tuesday that the bloc was open to investments from China but only under the right conditions.
These conditions include the creation of “real value added” in the EU, the creation of “real jobs”, and that there “will be real technology and real intellectual property “transferred, as European companies [have] been doing when they’ve been investing in China”…
It remains to be seen how such a policy would work in reality.
It’s not much, but it’s a start. Tech transfer requirements would allow Europe to build a globally competitive EV industry a lot sooner.
Of course if China retains control of those factories, it could still pose a threat if hostilities between the two blocks increases further. But European countries can just nationalize their factories at any time. In fact, the Netherlands just took over a Chinese semiconductor company:
The Dutch government has taken control of Chinese-owned semiconductor maker Nexperia, warning of risks to Europe’s economic security…The Dutch ministry said it invoked the country’s Goods Availability Act because of “recent and acute serious governance shortcomings and actions” at Nexperia, which is based in the Netherlands and has been majority-owned by Chinese technology group Wingtech since 2019.
I’m glad to see Europe discovering that it still has at least the vestiges of a spine. It’ll need a lot more than this in the years to come.
Well, it’s time for my annual Economics Nobel post! If you like, you can also check out my posts for 2024, 2023, 2022, and 2021.
Other than the tired old question of whether the Econ Nobel is a “real” Nobel prize,1 there are basically three things to talk about in these posts:
The research that got the prize
What the prize says about the economics profession
What the prize says about politics and policy in the wider world
So first let’s briefly talk about the research. This year’s prize went to Joel Mokyr, for writing about culture and growth, and Philippe Aghion and Peter Howitt, for making models of technological innovation. For good summaries of what this prize is all about, see:
I’m personally much more familiar with Aghion and Howitt’s work, so let’s start with that. The main idea they won the prize for is a model of how competition drives innovation, which they published in 1992.
The basic idea of this model is that technologies become obsolete as they’re replaced by better technologies. If you’re in academia, this might not be a problem, because your goal is to publish papers. But if you’re in a company that’s trying to turn a profit, the notion of your discovery being superseded should worry you. Suppose you spend a bunch of money and hire a bunch of researchers and invent a cool new product, only to see it superseded a year later by something even better. That’s how Digital Equipment Corporation must have felt when their cool new “minicomputers” were quickly made obsolete by the advent of the personal computer. It’s Joseph Schumpeter’s “creative destruction” at work.
In Aghion and Howitt’s model, this creative destruction deters companies from investing in research.2 It provides a natural brake on the pace of technological innovation, and limits how fast the economy can grow.
In 2005, Aghion and Howitt published an important update to this theory, co-authored with Nick Bloom, Richard Blundell, and Rachel Griffith. The new theory deals with the effect of competition on the rate of innovation. If a market is very competitive, the Aghion and Howitt (1992) theory dominates, and innovation is relatively low. If the market is very uncompetitive, it’s also less innovative, because a monopolist doesn’t feel threatened enough to innovate. But if the market is simply somewhat competitive, then companies will innovate a lot, because whoever wins the competition will get tons of profit from being a temporary monopoly.
Innovation is therefore maximized when companies are “neck and neck”. It’s easy to look at the incredibly expensive AI competition going on right now, and see this kind of “neck and neck” effect at work.
This is all pretty standard macro theory stuff — it imagines a pretty simple economy with just enough complexity to explain the idea that the authors want to think about, and works through the mathematical implications of that theory. And in fact it’s a bit easier to test than many macro theories, because it isn’t actually pure macro — the models’ primary implication is about how individual companies behave, which lets you get some causal evidence.3 Some causal evidence supports the “inverted U” of Aghion et al. (2005), while other studies fail to find it.
What’s harder to think about is how to apply these models. As Lina Khan and other modern antitrust advocates have discovered, there’s not really a big dial labeled “amount of competition in the economy” that you can easily turn. Someday we may be able to do that, but right now, Aghion and Howitt’s most famous theories remain largely descriptive rather than prescriptive.
As a side note, Aghion is one of my favorite economists, and I know his research fairly well (Howitt’s less so, sadly). And these aren’t actually my favorite papers of his! So let me mention some others that I think are important.
He has a 2017 paper with Benjamin Jones and Charles Jones that provides a good way of thinking about how AI will affect economic growth — and pretty prescient, considering it was written five years before ChatGPT even came out. Basically the idea is that AI is still constrained by bottlenecks, and that as AI handles more and more, the bottlenecks become more and more important. This is why Tyler Cowen predicts that AI won’t supercharge growth as much as optimists expect.
Aghion also has a 2018 paper with Bergeaud, Lequien and Melitz, looking at the effects of exports on innovation. As someone who has advocated for export subsidies as a way to boost productivity, I often find myself going back to this paper. The upshot is that top companies become more innovative when they compete (and win) in world markets, but less-competent companies become less innovative due to the creative-destruction effect.
Aghion also isn’t just a theorist; he does important empirical work as well. Aghion et al. (2015) found that China’s industrial policy often increased productivity by boosting competition4 — obviously a topic I’m very interested in. His 2023 paper with Blundell and Van Reenen found that regulation has a real and negative effect on businesses in France (though perhaps not as big an effect as one might expect).
And perhaps most relevantly, Aghion, Antonin, Bunel and Jaravel did a literature review in 2022 on the relationship between automation and jobs. They find that at both the company level and the industry level, automation actually increases jobs, probably by growing the overall size of the market:
In this article, we survey the recent literature and discuss two contrasting views on the impacts of automation on labor demand. A first view predicts that firms that automate reduce employment, even if this may ultimately result in job creations taking advantage of the lower equilibrium wage induced by job destructions. A second approach emphasizes the market size and business stealing effects of automation. Automating firms become more productive, which enables them to lower their quality-adjusted prices, and therefore to increase the demand for their products. The resulting increase in scale translates into higher employment by automating firms, potentially at the expense of their competitors through business stealing. Drawing from our empirical work on French firm-level data and a growing literature covering multiple countries, we provide empirical support for this second view: automation has a positive effect on labor demand at the firm level, which remains positive at the industry level as it is not fully offset by business stealing effects. [emphasis mine]
This contradicts the dire predictions of researchers like Daron Acemoglu who believe that automation is a job-killer. It was written before generative AI came out, so this result could change, but it’s highly encouraging.
Anyway, those are my favorite Aghion papers, and I kind of wish more of them would have been cited in his Nobel, but in general I’m very glad he won the prize. He’s really one of the best researchers we have on the topic of innovation — a key voice guiding us through our strange new era of rapid technological change.
Anyway, on to Mokyr. Mokyr’s key work, especially his book A Culture of Growth, is very different than Aghion and Howitt’s. It’s basically a narrative history of the scientific advances that led to the Industrial Revolution. I’ve never read the book, mostly because I’m inherently skeptical of cultural explanations of growth, but I guess now I have to read it.
Mokyr’s explanation of why the Industrial Revolution happened in Early Modern Europe, as opposed to in China or somewhere else, isn’t completely cultural. Part of his hypothesis is technological — he thinks the printing press allowed scientists and engineers to more easily exchange ideas and build on each other’s innovations. This is actually somewhat testable — for example, Dittmar (2011) showed that the spread of the printing press predicted future economic growth.
Mokyr also cites political factors — in particular, Europe’s fragmentation, which allowed scientists and inventors to shop around for a country that would support them. It might be possible to test this too, using natural geographic factors (which tend to separate regions into multiple countries) to predict where growth would eventually happen.
But Mokyr’s most famous idea — and the one that he put in the title of his book — is that Europe’s top scientists and thinkers had a special culture that allowed them to kick-start economic growth. Basically, Mokyr argues that they believed in the idea of scientific progress — the idea that science and technology are basically good for humanity, and that they naturally build up and improve over time. This attitude, he thinks, was key to Europe’s take-off — and, eventually, to the whole human race’s ability to escape poverty.
It annoys me a bit that this type of work won a Nobel prize. This is not because I disagree with the idea — in fact, I think it’s probably quite right, and despite never having read Mokyr at all, I’ve been writing for years that we need similar attitudes in the modern world:
I love the general outline of Mokyr’s ideas, and I expect I’ll find them very reasonable. But I don’t think the Econ Nobel should be about ideas that simply sound legit — even if I’d personally make policy based on those ideas.
In previous years, the prize had been trending toward greater recognition of empirical economics and applied theory — basically, it had been rewarding economists who made econ more of a science. An award for narrative history and untestable theories about culture takes us in the exact opposite direction.
In fact, I spent most of my Nobel post last year complaining about Acemoglu and Robinson doing something similar, with their theory of institutions and development:
But at least Acemoglu and Robinson tried to test their theory empirically! Yes, the tests aren’t that reliable, but at least there’s still the idea that some sort of empirical, testable science is being done. Mokyr tries to quantify some of the cultural elements he’s talking about, but he doesn’t actually try to test his theories against data. It’s not a bad or worthless exercise, but it’s not scientific at all.
The effect on the economics profession may not have been foremost in the committee’s mind, however. Instead, they may have been thinking about the anti-growth turn in Western culture. Under Trump, America is embracing antivax lunacy and ignoring the power and promise of electric technology, while regular Americans are terrified of AI. Meanwhile, much of Europe is embracing degrowth ideology and shunning air conditioning, while over-regulating information technology.
The West, in other words, may be losing the very secret sauce that powered its rise. We desperately need Mokyr’s culture of growth — the belief in the power of innovation to help regular human beings, and the conviction that progress is cumulative. The Nobel committee may be sending a message to Western civilization to break out of the high-level equilibrium trap into which we may have fallen.
It’s actually only one of two deterrents. The other deterrent is that when the pace of research is really fast, companies know they’ll have to overpay for researchers down the line just to stay in the race, which will reduce their future profits. That sounds hokey until you remember that Meta is now paying AI researchers $250M salaries.
It’s very difficult to figure out cause an effect when you’re dealing with one entire economy, but a lot easier when you have lots of different companies you can look at.
If you accept the “inverted U” theory of competition and innovation, this result implies that before the industrial policies, China’s industries were too dominated by SOEs or other uncompetitive quasi-monopolies. So industrial policy might be a good antitrust tool!
