2025-01-30 16:37:30
A Chinese company called DeepSeek recently released a series of large language models that are just about as good as the ones made by American companies like OpenAI and Anthropic, but cost less to make. Within much of the media and the tech community, this has been treated as an epochal event — a “Sputnik moment” for the AI race between the U.S. and China.
As with the L.A. wildfires three weeks ago, DeepSeek has led to an explosion of hand-wringing, shouting, bad takes, and other social media histrionics. Some people breathlessly proclaimed an imminent U.S. stock market collapse:
In reality, the Nasdaq is up about 25% for the year, and is at a higher level than it was two weeks ago:
Nvidia stock did take a hit from the news. Nvidia sells GPUs, the chips that are used to train AI models. DeepSeek shows that AI models can be trained using fewer GPUs; this caused many investors to predict that Nvidia’s chips will be less needed going forward. Because Nvidia is one of the world’s most valuable companies, any significant drop will look big in dollar terms. But the company’s stock is only down to where it was last October; over the past year, as of this writing, it’s still up over 90%:
In fact, as many people have already pointed out, there’s a simple reason why more efficient ways of creating LLMs could actually increase demand for Nvidia’s chips. When something gets cheaper, people buy more of it. So if LLMs become cheaper to produce, people will rent LLMs for more tasks. This will require purchasing more chips — especially since newer models use lots of compute for inference (i.e., “thinking” about the answer to each question).1
Whether demand for Nvidia’s chips goes up or down depends on whether the increase in efficiency outweighs the resulting increase in demand for LLMs. It’s certainly true that previous increases in efficiency — which have been enormous — don’t seem to have harmed Nvidia’s valuation much.
Other people declared that DeepSeek would supercharge the global economy. A top macroeconomist declared that the Chinese LLM release would prove more important than the invention of electricity or the internal combustion engine:
Obviously it’s not possible to prove that statement wrong, but it seems highly unlikely, given that the model’s performance is about the same as that of other models — including other Chinese-made LLMs. The idea that DeepSeek is revolutionary is based on the fact that it can do the same thing for much cheaper. But while the company does seem to have made some real innovations here, Anthropic CEO Dario Amodei points out that they’re roughly in line with previous cost decreases. LLMs are a sector where costs have just been falling very fast — there’s no reason to think that DeepSeek represents such a unique structural break from that trend that we should compare it to the invention of fire or the wheel.
Finally, there were lots and lots of takes about U.S.-China competition. DeepSeek probably does have some relevance there, but the apocalyptic tone of much of the commentary here was overblown. There was a flurry of articles with titles like “Does China's DeepSeek Mean U.S. AI Is Sunk?”, while one prominent AI commentator declared that Facebook had sold out America by making its AI open-source:
This is a bit silly,2 because DeepSeek itself is also open-source; if Facebook undermined America’s “advantage” in AI, then DeepSeek itself is undermining China’s. A more reasonable conclusion, as I’ll argue, is that there isn’t really that much national advantage here to undermine — keeping this sort of model proprietary is just not really as useful or feasible as in other areas of technology.
In addition to all of these overblown takes, there were lots of serious debates and speculation about DeepSeek. These included questions about what DeepSeek’s key technological innovations are, which chips it actually used to train its models, how much that training actually cost, whether and how DeepSeek evaded U.S. export controls, and so on. Not being an AI expert, I’m not qualified to evaluate those interesting debates.
But even without answering those tough questions, I think that there are some pretty obvious lessons a calm, rational outside observer can learn here, just from the fact that a random Chinese company managed to produce a frontier-level open-source LLM. As I see it, the key early takeaways are:
LLMs don’t have very much of a “moat” — a lot of people are going to be able to make very good AI of this type, no matter what anyone does.
The idea that America can legislate “AI safety” by slowing down progress in the field is now doomed.
Competing with China by denying them the intangible parts of LLMs — algorithmic secrets and model weights — is not going to work.
Export controls actually are effective, but China will try to use the hype over DeepSeek to give Trump the political cover to cancel export controls.
2025-01-28 16:05:22
The Progress Studies folks that I hang out with a lot tend to complain that Americans have begun to fear the future instead of embracing it. I generally agree. But being the type of guy who looks for technological and explanations for every cultural phenomenon, I think it’s not very surprising that people in a stupendously rich, comfortable society like the modern United States would tend to be afraid of change.
Fundamentally, change means disruption, and disruption means risk. If you’re a car dealer, Tesla’s model of direct sales is a threat to your livelihood. If you work for a company that makes its money selling solutions to homework problems, LLMs are probably going to put you out of a job. Every single change to business models, government institutions, and technology threatens whoever came out on top in the previous equilibrium.
And despite their constant bellyaching, most Americans did come out on top in the previous equilibrium. The median American is far richer than the median resident of almost any other nation, and enjoys a material standard of living that others can only gawk at. If that’s where you are, any big technological, political, or institutional change carries at least some risk of knocking you from your privileged perch.
On top of that, America is getting older as a country:
Being older makes it harder to adapt to changes in the technical-social equilibrium. If you’re 18 and just starting out at the bottom, big changes are almost pure opportunity; if you’re 50 and you have a mortgage and kids in college and a high salary at a big corporation, you probably want to keep things stable.
And in fact, a common stereotype of Americans is that they demand to have more than their parents had, for less effort and risk:
But while there’s a grain of truth here, I think it’s overstated. Given all the economic incentives to resist change, I’m actually surprised that Americans embrace the future as much as they do!
Consider AI. On one hand, more Americans feel concerned rather than excited about the technology:
But these concerns don’t seem to be strong enough to influence policy very much. Whereas Europe has decided to try to regulate AI instead of building anything itself, the U.S. is charging ahead — Trump has signed an executive order doing away with “AI safety” concerns at the federal level, and Gavin Newsom vetoed an AI safety bill in California.
Or take housing. Of course I talk all the time about how big of a problem NIMBYism is in America, and how a lot of politically powerful homeowners have created a rent crisis by trying to protect their quiet empty streets from townhouses and duplexes. That is true. But relative to changes in population, America has actually been building a decent number of homes since the pandemic:
Some parts of the country are far outpacing the average. The skyline of Austin, Texas looks almost totally unrecognizable from a decade ago, with a forest of new tall buildings:
And the city has built so much housing that rents have plummeted:
For a third encouraging example, look at entrepreneurship. After decades of declining business formation, there has been a huge surge of Americans starting companies since the pandemic, especially in high-tech fields:
Business formation isn’t the only measure of increased dynamism, either. More Americans are looking to switch jobs than in the 2010s.
This is all very encouraging. It suggests that many Americans are rising to meet the challenge of a risky, uncertain future, rather than hunkering down and protecting their wealth. Perhaps the “roaring 20s” are really seeing a shift back toward the bold optimism that the Progress Studies people have been hoping for.
But if so, the shift is still very incomplete. There are still a lot of important ways in which Americans — especially the leadership and the politically engaged class — are trying to resist the future instead of riding the wave.
The most egregious recent example of future-fear comes from our newly elected President, Donald Trump. Not content with freezing approvals for wind power on federal land and water, Trump has now paused solar power as well:
The Trump administration is pausing approvals for new renewable energy projects on public lands and in public waters…The Interior Department quietly issued an order Monday that blocks activities that enable renewable development on federally-owned lands or offshore…For 60 days, the government will not issue any leases, rights of way, contracts or “any other agreement required to allow for renewable energy development.”…It comes as Trump has launched an assault on wind energy in particular, issuing an executive order that pauses new approvals for wind energy. But applying the pause to renewables broadly is an escalation — pausing solar energy action as well.
I don’t want to paint this as a catastrophe yet. First of all, it’s only on public land; most renewable energy is built on private land. And second, it’s not clear whether this is a temporary move that will lapse after two months’ time, or the beginning of a more permanent assault on solar and wind energy. But from everything Trump is saying in interviews, it seems clear that he just doesn’t like solar energy at all:
What’s going on here is, fairly obviously, some combination of simple old-guy aesthetic NIMBYism (“those massive solar fields”) and right-wing culture wars. The latter has been caused by decades of Americans seeing energy technology as being fundamentally about climate, rather than about energy itself. In a post back in December, I wrote the following:
The fact is, climate change mitigation is incredibly — almost totemically — unpopular within the GOP and the conservative movement. There is a strong knee-jerk opposition to climate activists, who are generally seen as communists hiding their radical anti-capitalist agenda behind the facade of a global emergency.3 Anything the climate people want, the Republicans will say “no” to.
So when we rhetorically and mentally put electrical technology in the climate bucket, it immediately becomes a non-starter with around half the country…It would be fairly pathetic if America became a second-rate power in the 21st century simply because we collectively associated the revolutionary technologies of the day with hippies and tree-huggers.
You can easily see this in polls. Even as solar power has gotten radically cheaper, Republicans have become more hostile to the technology:
Solar power is the cheapest form of energy now, even with zero subsidies. One state that realizes this is Texas, which is building more solar than any other state:
In fact, at some times of day, renewables now provide the bulk of Texas’ power:
The fact that Texas is building the most solar and wind should be proof that these technologies simply work. The rest of the U.S. isn’t far behind, though:
If you point this out on social media, of course, a bunch of people will immediately mob you and say “Nuclear is better than solar!”. That’s wrong, of course. Nuclear’s compactness makes it useful for some niche applications, but its high costs and lack of a learning curve means that when land is available for solar, nuclear simply isn’t competitive at all. You can tell this because China, which is building far more nuclear plants than any other country on the planet, is building far more solar power than nuclear.
So anyway, Trump is trying to stand in the way of a cheaper, better power source replacing an old, inferior power source. It’s part of a general pattern of Republicans being afraid of the physical technologies of the future — other examples include electric vehicles and vaccines.
If Republicans tend to fear the future of hardware, then Democrats fear the future of software.
The Biden administration consistently went after big tech companies in its antitrust push — a puzzling decision, given that the price of software is hardly what’s breaking most people’s bank accounts. But the “Neo-Brandeisian” school of antitrust, of which Biden’s FTC chair Lina Khan was one of the creators, believes that harms to consumers are not the main reason for breaking up big companies and preventing mergers. The Neo-Brandeisians believe that the main problem with big companies is that they accumulate too much power — not just market power, but also political power.
Essentially, Democrats went after Big Tech because they thought tech companies had gotten too politically influential. Ironically, this political antipathy ended up being one factor pushing the tech industry to the right and resulting in the election of Trump.
Democrats have also been behind the “AI safety” push, which sees AI primarily in terms of its dangers instead of its opportunities. Biden was the one who issued the executive order on AI safety that Trump just repealed. And Scott Wiener, a California Democrat whose ideas I almost always endorse and respect, was responsible for a bill that would have restricted AI had it not been vetoed by the governor. Those restrictions were, in my assessment, pretty pointless, especially because China would have just proceeded with largely unfettered AI research anyway — as it ended up doing with DeepSeek.
Why did some Democrats want to hobble AI? Again, the most likely explanation is that they were concerned with limiting the power of the tech businesspeople who created the AI. This need to limit the power of the technologist class was the core message of Daron Acemoglu and Simon Johnson’s book Power and Progress, which embodied ideas that seem to have been influential on the left:
Ultimately, the Democrats’ desire to hobble the software industry doesn’t seem too different from the Republicans’ attempts to hobble the green energy industry. In both cases, fear that the technology will empower the wrong people — hippies and commies in the case of hardware, “bro-ligarchs” in the case of software — ends up translating into antipathy toward the technologies themselves. This stands in contrast to the mid-20th century, when Americans, recently unified by World War 2, simply assumed that new technologies like vaccines, airlines, and TV would be created for their own benefit.
I think this is teaching us something important about why countries start to fear the future. It’s not just pure economic incentives or personal risk. When societies become politically and socially divided, people begin to fear that new technologies will empower their political enemies. When people start to believe that their enemies will rule the future, a natural impulse is to stand athwart history yelling “Stop!”.
So if we want to return to a society that eagerly embraces the new, reducing our political and social divisions might be a prerequisite. Americans need to know that when the future comes, we’ll all get to go together.
2025-01-27 09:19:17
This was a fun chat! It was great to reconnect with Paul, who of course I’m a huge fan of. If you aren’t already subscribed to his Substack, check it out.
Anyway, if you like to read discussions instead of listening to them, there’s a pretty good AI-generated transcript embedded in the video. Just click the little notepad icon in the bottom right of the video, next to the “CC” button, and the transcript will pop up on the right. There’s also a “Transcript” button at the top of this post.
Thanks to everyone who submitted questions! We didn’t get to nearly all the things people requested (or that I wanted to get to), but I guess that just means we’ll have to do it again sometime. My key takeaways from this discussion were:
Paul left the NYT because of too many layers of editing. This basically fits with my thesis that op-ed writers need a lot less editing than journalists do, and that good op-ed writers function best with no editing at all.
I’m starting to suspect that the best thing for the print media is to unbundle journalism/reporting from analysis/op-eds — basically, to have news services like the AP and opinion publications like The Atlantic. I’ll think more about that in a future post.
To succeed on Substack, writing very frequently is the best approach. (Note: This was the advice Ben Thompson and Byrne Hobart gave me when I first started out on Substack, and they were right.) This system favors people like Paul and myself who naturally have a lot to say and who usually don’t bother making things too polished.
Paul is absolutely right that the legacy media is in big trouble. The New York Times and other big newspapers could always rely on being seen as the voice of authority in previous decades, but they no longer are. Most people explain this by saying that the legacy media forfeited the public trust by making a series of big mistakes, but my own preferred explanation (as usual) is technoloical: For a long time that was just because they had a natural monopoly on things like classified ads and physical paper delivery. When those natural monopolies went away due to the internet, there was still a long lag where people considered the NYT the voice of authority because they always had before. But eventually, new entrants challenged the authority of legacy media on the substance, and often won — getting better scoops in the journalistic space, or doing better, more in-depth analysis in the op-ed space.
The vibe of the second Trump term is shaping up to be “The world is going to hell but let’s try to have fun with it.” This is in stark contrast to the vibe of the first Trump term, which was all about “resistance” and grim dire struggle. The shift is not necessarily because Trump’s opponents (including both Paul and myself) are less concerned now than before, but because they realized that the old “resistance” strategy didn’t really work. Trumpism and all its attendant dysfunctions and screw-ups will be here for a long time to come.
I’ve often made analogies between the 2020s and the 1970s, but Paul — who actually worked in the Reagan administration — thinks they’re much different. In particular, Trump’s people are much more contemptuous of rules and norms than Reagan’s were. Paul also thinks Trump’s people are much less competent, though I argue that this is mostly at the top level, and that the people at the subcabinet level are actually looking a lot more competent than in Trump’s first term, mostly because of the influx of tech and tech-adjacent talent.
Trump’s mass deportation of illegal immigrant workers in the agriculture industry could raise food prices, making Americans angry.
I’m pleasantly surprised by how well Javier Milei seems to be doing in Argentina. Paul thinks Milei will screw up by trying to peg Argentina’s currency to the dollar.
Paul is confident about the underlying strength of the American economy, especially our productivity and dynamism. This will help to mitigate whatever inflationary shenanigans Trump does. I agree.
On trade balances, Paul doesn’t buy the common theory that the U.S.’ ownership of the reserve currency is a big reason for our trade deficit. He points out that Britain runs a trade deficit, even though they don’t have the reserve currency. Paul isn’t quite sure why we seem to have structurally unbalanced trade, but he thinks that it’s probably not going to change unless the world goes back to capital controls.
On manufacturing, Paul agrees with me that strategic/military/geopolitical concerns are the main reason we need industrial policy and reindustrialization. Neither of us expects manufacturing to go back to 30% of the U.S. economy, or to provide huge amounts of good blue-collar assembly-line jobs to the American working class. But it’s still very important.
In terms of policies to promote reindustrialization, Paul favors subsidies over tariffs, for a simple reason — tariffs are basically just subsidies plus sales taxes. He’s right. I still think targeted tariffs are important in order to quickly block Chinese attempts to forcibly destroy certain Western industries with floods of subsidized exports. But I agree that subsidies — and other industrial policies — should be the main tool of reindustrialization.
In the past, Paul has declared that the U.S. is “an insurance company with an army”, although it turns out he didn’t invent that line. But now, he agrees with me that we’re really just an insurance company. Defense spending has collapsed as a percentage of government spending, and of the economy, since the mid-20th century when Eisenhower warned of a military-industrial complex, and it’s far lower even than during the late Cold War. Progressive thinking on defense spending is really evolving fast, and this is a good thing.
Anyway, I really enjoyed this chat. Next time I promise to mount my phone on a stand so it doesn’t move around so much!
2025-01-25 09:24:10
Reshoring American industry has become a bipartisan policy objective — it has always been part of Donald Trump’s agenda, and Biden cared a lot about it as well. The idea has always been met with skepticism from a number of directions. Many economists and free traders are skeptical of anything that involves tariffs and/or industrial policies. And partisanship being what it is in America, both Republicans and Democrats have naturally doubted the others’ ability to follow through on their promises. But in addition, I often encounter a natural skepticism about America’s ability to do manufacturing at all.
Americans can be forgiven for having this impression. Most of our lived experience has either been the Rust Belt era of the 1980s, or the supercharged offshoring of the 1990s, 2000s, and 2010s. America has not had a factory building boom in a very long time. On top of that, most people who take economics in America learn only one theory of international trade, which is the theory of comparative advantage — basically, the idea that countries specialize in whatever they’re best at. It’s easy to believe that America specializes in software and services instead of making physical stuff, because that’s the way things were trending for decades.
If you believe that, then you probably think that reshoring manufacturing will always be an uphill battle, if not downright impossible. Sure, with enough tariffs and subsidies we could force Americans to buy more expensive stuff made in America, but this will make us all poorer. Why not just leave manufacturing to the East Asians, and maybe the Germans, and focus on what we seem to be good at?
And yet the simple way to think about trade isn’t necessarily the right one. There’s another theory that says that since America has lots of capital and technology, we can do a lot of automated manufacturing. And there’s yet another theory that says that because the world loves variety, the U.S. can manufacture close variants of the things the Asians and Europeans make. The deindustrialization of the U.S. since the turn of the century might be more about an overvalued exchange rate, deliberate Chinese competition, and U.S. industrial policies that favored the financial sector over the manufacturing sector.
The common belief that Americans just aren’t good at making stuff seems contradicted by areas in which we are startlingly good at making stuff — for example, SpaceX, which is pumping out the world’s best rockets from U.S. factories in stunningly high volumes. The American South has also become a hub of high-quality auto manufacturing, with the help of Japanese and South Korean investment:
If that’s true, then reshoring has a chance. The uncompetitive dollar will remain a major problem, but Chinese competition can be blocked with tariffs and other trade barriers, and U.S. industrial policies can shift from a pro-finance to a pro-manufacturing orientation. In fact, this approach is already bearing fruit in a number of strategic industries.
Take solar power, for instance. For years, the main story in that industry was the collapse in U.S. manufacturing, and the overwhelming dominance of China. In an article in Bloomberg last September, David Fickling lamented:
The fall of America as a solar superpower is a tragedy of errors where myopic corporate leadership, timid financing, oligopolistic complacency and policy chaos allowed the US and Europe to neglect their own clean-tech industries. That left a yawning gap that was filled by Chinese start-ups, sprouting like saplings in a forest clearing.
But even before that story hit the presses, things had already begun to change. In December, the Solar Energy Industry Association reported that U.S. solar manufacturing capabilities are on the rise:
In 2017, the U.S. ranked 14th in the world for solar panel manufacturing capacity. Starting in 2018 and then accelerating in 2022, additional factories started springing up left and right throughout the country, with a focus in the South…Major investments poured into building factories and expanding existing facilities. Today, the U.S. has leapfrogged competitors and ranks 3rd in manufacture of solar panels, passing large solar manufacturing countries like Malaysia, Thailand, Vietnam, and Turkey…A new report by SEIA and Wood Mackenzie found that the industry had reached a critical threshold:
After a record Q3, U.S. solar manufacturing has reached a critical threshold. At full capacity, American solar module factories can now produce enough to meet nearly all demand for solar in the U.S.
As more solar deployment happens, more manufacturing will come online…Companies are investing billions of dollars to produce American-made solar panels in states like Georgia, Ohio, Texas, Washington, South Carolina, and Alabama…[T]here are more factories on the way, either announced or under construction.
Obviously the U.S. is still way behind China, but this rising trend of production and self-sufficiency is very different from the typical story you hear. As the article notes, the reshoring of solar began in the late 2010s, under Trump, and may have had something to do with Trump’s tariffs on solar panels. A second round of tariffs, courtesy of Biden, went into effect near the end of 2024, and definitely seemed to have an effect on solar imports:
But it was Biden’s Inflation Reduction Act that really kicked solar reshoring into high gear:
For another example, look at semiconductors. I’ve written a lot about how the CHIPS Act has galvanized U.S. production in this most strategic of all industries, including major investments from Taiwan and elsewhere. This is from a recent report by the CHIPS Program Office:
The United States has seen more investment in electronics manufacturing over the last four years than in the previous three decades combined. Planned investments are now nearly $450 billion, marking the largest wave of semiconductor manufacturing expansion in U.S. history. This includes the two largest domestic investments in semiconductor manufacturing by U.S. companies in history (Intel and Micron), as well as the two largest foreign direct investments in new projects by any company in history (TSMC and Samsung)…Perhaps most significantly, for the first time, all five of the world’s leading-edge logic and dynamic random-access memory (DRAM) manufacturers (Intel, Micron, Samsung, SK hynix, and TSMC) are building and expanding in the United States. By contrast, no other economy in the world has more than two of these companies manufacturing on its shores…
The United States is projected to produce at least 20 percent of the world’s leading-edge logic chips by 2030 (up from zero percent in 2022) and ~10% of its leading-edge DRAM chips by 2035 (also up from zero percent)—both technologies that are essential to the future of artificial intelligence (AI), high-performance compute, and advanced military systems. TSMC’s Arizona facility has already begun volume production of leading-edge chips, marking the first time in roughly a decade that a new fab is making these technologies domestically.
And The Economist, certainly no friend of industrial policy in general, has grudgingly admitted that U.S. reshoring of the semiconductor industry is succeeding:
Early returns are impressive: the [CHIPS Act] programme has catalysed about $450bn of private investments. And this money is spread across much of the industry, from high-tech packaging to memory chips. One marker of success is the production of the most advanced chips, measuring less than 10 nanometres in size. In 2022 America made few such chips. By 2032 it is on track to have a share of 28% of global capacity.
As in the case of U.S. auto manufacturing a generation earlier, foreign direct investment — especially from Taiwan’s TSMC — has played a crucial role.
In early 2024, some poorly informed pundits were writing stories declaring that “DEI killed the CHIPS Act”, while others were wondering whether Americans had a culture capable of making chips. Those articles were spectacularly ill-timed — obstacles were quickly overcome, and the factory is now pumping out 4nm chips. Those are, by at least some measures, the most advanced semidconductors ever made on American soil.
And what’s more, those chips are being made with yields (i.e., quality) that are comparable to, or even higher than, what Taiwanese factories get. The idea that American workers are unable to produce high-quality products turned out to be wrong. The chips made in the U.S. are a bit more expensive (around 30% more right now), but that cost differential will probably come down as volume ramps up and as chipmaking experience spreads throughout the country.
In fact, the reshoring effort is going so well that TSMC is now planning to build even more cutting-edge chips at its U.S. plants:
So far, the effort to reshore semiconductors is turning out to be a big success.
Batteries look like a third reshoring success. Currently, most batteries are produced in China, but the Inflation Reduction Act may be starting to turn things around:
It’s not just factories being announced, either; production in the U.S. is way up:
The reshoring of the solar, chip, and battery industries is a direct rebuttal to the naysayers, and proof that American manufacturing can succeed.
Although these are only three industries, they will inevitably facilitate the reshoring of related industries that either supply these manufacturers or consume their output. American reindustrialization isn’t just about a few key tentpole industries — it’s about a whole web of suppliers, customers, related industries, and talent. Fortunately, we can already see this web starting to form in the U.S. SEIA reports that America’s solar manufacturing boom isn’t just limited to the panels themselves, but related industries like solar tracker, solar inverters, and upstream materials production like wafers and ingots.
Meanwhile, the CHIPS Program Office reports that the semiconductor boom also includes downstream activities like packaging and testing. The Economist points out that this ecosystem, as well as the talent that gets developed for the CHIPS Act’s projects, will reduce costs and help sustain future expansion of chip manufacturing in America:
The subsidies have helped to shrink a gap of roughly 30% in the cost of building and operating fabs in America compared with in Asian countries. In part costs are lower in Asia because Asian governments lavish handouts on companies. But Asian producers have also reaped the benefits of dense manufacturing clusters, with well-trained workforces and plenty of suppliers nearby. The hope is that CHIPS has started this process in America. “It’s enough to get the flywheel going,” says [outgoing Commerce Secretary Gina] Raimondo.
At this point, whether reshoring continues is largely a matter of political will and smarts. If Donald Trump keeps attacking the solar industry, or follows through on his earlier threats to cancel the CHIPS Act, it could shift production decisively back to China. It would be ironic if a President who rose to power promising to restore American industry ends up being the person responsible for strangling our industrial revival in its crib.
2025-01-23 16:16:43
There’s a popular perception that the U.S. presidency has become an “imperial presidency”. Part of this idea is that Congress is so hopelessly gridlocked that the President has begun governing by executive order — basically, an elected king. I don’t think this is true — at least, not compared to the past. Executive orders were more common in the early 20th century than they have been in recent years — Calvin Coolidge issued 215 a year and Herbert Hoover 242, compared to 55 by Donald Trump (so far), and 38 by Joe Biden. Even in terms of word count, we don’t see much change over the years.
As for whether executive orders have overreached in terms of their scope, that is definitely a concern (though it was also a concern a century ago), and I’ll talk about that later. But many modern executive orders are just one President revoking or canceling some executive order of the previous President — often some symbolic culture-war thing. Yo-yoing back and forth between conservative and progressive cultural stances isn’t necessarily the most stability-inducing thing, but it doesn’t make the President a king.
Donald Trump’s flurry of executive orders should be seen in this context. A bunch of what Trump did on his first day in office was just to cancel out stuff Biden tried to do. Ultimately, the effect on the nation won’t be noticeable. But in addition, Trump issued some orders that could have far-reaching consequences, and in a few cases they represent a broadened assertion of presidential power. Time has a good summary. Of course I’ll write more about these orders in the days ahead, but I thought I’d give some rapid reactions. The five main areas I’m going to focus on here are:
Immigration
Environmental review and permitting
Affirmative action and DEI
Energy policy
DOGE and the federal workforce
Trump’s executive order on birthright citizenship is his most important, because it’s the biggest test of executive power by far. The 14th Amendment of the U.S. Constitution guarantees birthright citizenship:
All persons born or naturalized in the United States, and subject to the jurisdiction thereof, are citizens of the United States and of the State wherein they reside.
This is a common practice in the Western hemisphere, though it’s rare in the Old World. Immigration restrictionists have been wanting to go after this one for a while, and they finally decided to just go for it and see what happens.
Trump’s executive order says that visa holders — people here on O-1, H-1b, and so on — are not subject to U.S. jurisdiction, and therefore their kids are not citizens even if born in the U.S. (It also says the same about the kids of illegal immigrants.) It doesn’t revoke anyone’s citizenship; it applies only to future kids.
Trump’s EO is not declaring the Constitution null and void, but it’s asserting an interpretation of the text that has already been contradicted by the Supreme Court, way back in 1898. SCOTUS ruled that “subject to the jurisdiction” of the U.S. means that you’re under U.S. law, and visa holders are definitely under U.S. law, so there’s a good chance this executive order will be struck down. Already, states are launching a flurry of lawsuits against the order. (Update: A federal judge has already blocked this executive order, so we’ll see what SCOTUS has to say.) I’m no legal expert, but it seems pretty nonsensical to argue that green card holders are subject to U.S. jurisdiction, but O-1 holders aren’t.
Even if the order does get struck down, however, it’ll probably exert a chilling effect on skilled immigration. The chance to have a kid who’s an American citizen is one reason that the world’s most talented people come to America to work — often on an O-1 visa. Getting those people to work here is a national security priority, as well as a boon to the U.S. economy":
Now those people know that the citizenship of their future children is a precarious thing — a political football. This will probably exert a chilling effect on America’s ability to recruit top talent. The EO also seems badly written, in a way that could potentially be interpreted turn the kids of visa holders into illegal immigrants the moment they’re born (though I bet they’d find some fix for this).
What will the voters think of this move? Birthright citizenship is broadly popular, though not overwhelmingly so. A bunch of recent polls find support for birthright citizenship in the 60% range. But Americans are about evenly split on granting it to the kids of illegal immigrants. As usual, Americans like immigration in general, but really hate the idea that people are violating the country’s rules and getting away with it. It seems possible to me that Trump’s people added visa holders to the EO in order to give a conservative SCOTUS a face-saving way to split the difference — ending birthright citizenship for the kids of unauthorized immigrants, but keeping it for the kids of visa holders.
Anyway, I also think this order confirms a trend that I’ve been noticing since Trump’s first term: The MAGA movement is opposed not just to illegal immigration, but also to high-skilled legal immigration, especially immigration from India. The massive blowup on right-wing social media around Christmastime was all about Indians, and there continues to be a huge amount of hate directed specifically at Indians in right-wing circles. Elon Musk stood up for H-1b workers on X, but if there was any internal battle over it within the administration, his faction does not appear to have come out on top.
All in all, this is Trump’s worst executive order, for any number of reasons, and has the potential to do lasting harm to the fabric of the nation.
It’s also part of a suite of executive orders attacking immigration in general. A lot of this stuff — suspending asylum claims at the border, suspending refugee resettlement, reimplementing the “remain in Mexico” policy, and so on — is stuff that the American people wanted (or probably would want if they understood it). Many of these orders will be challenged in court, as they were in Trump’s first term; permanent resolution of these problems will require Congress to change the asylum law.
But Trump’s legal justification for these orders relies on his declaration that the wave of illegal immigration constitutes an “invasion”. Obviously, this is nonsense — it’s just a bunch of poor people looking for work, the same as it ever was. Declaring that this is an “invasion” seems like a clear recipe for executive overreach.
If the birthright citizenship EO is Trump’s worst, the two orders on environmental review — basically, limiting NEPA — are easily his best.
2025-01-21 16:40:56
Financial disclosure: I hold a modest amount of BTC and ETH.
Don’t worry, I’m going to write a post about Trump’s flurry of executive orders very soon. But in the meantime, I thought I’d write about crypto.
Just before his inauguration, Donald Trump and his wife Melania launched a pair of cryptocurrencies, called TRUMP and MELANIA. (The image at the top of this post is from the TRUMP launch.) The coins have both declined in value from their initial peaks, but their market capitalizations, as of this writing, are still around $7 billion and $800 million, respectively:
A lot of crypto folks weren’t too happy about TRUMP and MELANIA, believing that a stunt like this could undermine the respectability of crypto in general. Their interest in the matter is very simple to understand. Most crypto people — like me — hold Bitcoin, and probably Ether too. If crypto gains respectability and wide adoption by financial institutions, pension funds, etc., demand for Bitcoin and Ether will increase by a lot. This will cause BTC and ETH to go up enormously in price, enriching pretty much everyone in the entire crypto ecosystem.
On the other hand, the launch of new “memecoins” doesn’t enrich the average crypto holder. Yes, you can buy these coins and hope to make a profit. One bold and lucky investor turned $1 million into $120 million investing in TRUMP (though that number could go down if he holds onto it and it crashes). But in general, new memecoins like TRUMP and MELANIA don’t enrich crypto people nearly as much as increased adoption for the existing coins that they already own.
And the crypto people are right that TRUMP and MELANIA just make the whole space look bad. The original idea of cryptocurrency was that it was going to replace fiat currency as the standard means of payment. So far that hasn’t happened, or come anywhere close to happening, but the dream remains that one day people could buy their pizza and beer with BTC or ETH, stablecoins like Tether and USDC, or maybe even a BTC imitator like DOGE. But absolutely no one thinks that one day we’ll be paying our grocery bills with TRUMP or MELANIA. The purpose of these coins is not to be currencies, but speculative assets. Their reason for existence is to transfer money from investors to other investors, not from buyers of physical goods to sellers.
This does not mean TRUMP or MELANIA will ever go to zero. As we’ve already seen with a huge number of other cryptocurrencies, it’s possible for assets with no fundamental value whatsoever to retain market value over long periods of time. Some people hold on to the speculative assets — and are willing to pay some price to keep holding on to them — on the off chance that someday other people will notice those assets and speculate on them in turn, allowing the current holders to sell at a profit. This is exactly what happened to Dogecoin.
Speculation is a game of higher-order beliefs. I might believe TRUMP is fundamentally worthless, and you might believe the same, but if you believe that someone out there thinks it’s not worthless, you might think you can sell it to that person. And then I might buy some TRUMP so I can sell it to you, and so on.
There are a number of economic models that explain how this can lead to financial bubbles. The most famous is probably DeLong et al. (1990), which argues that if there are some people called “noise traders”1 who mistakenly believe an asset is more valuable than it really is, and if smart traders who know the true value aren’t able to get enough money to trade against the noise traders, it can force the smart people to temporarily go along with the herd and cause a bubble.
Another paper in this vein is Abreu and Brunnermeier (2003). They argue that even if the smart traders would together have enough financial firepower to pop a bubble, they can’t always do it, because they would have to short the bubble at the same time. Short-sellers don’t coordinate with each other (in fact, it’s illegal to do so). So each short-seller doesn’t know when the others are going to make their move, and so instead they ride the bubble for a while.
For models like these to explain the value of TRUMP and MELANIA, there would have to be someone out there who thinks these memecoins have long-term value. But in fact, the models will work just as well if there are people out there who love to gamble, even when they know it’s likely to lose them money. We know from the existence of craps and roulette that there are some people who like to take financial risks that have a negative expected payoff, as long as there’s a chance of winning big. Observationally, people who just love to play roulette aren’t really distinct from people who mistakenly believe that the odds of roulette are in their favor. So TRUMP and MELANIA might have value because some people just like to gamble on exciting memecoins, even knowing they’ll probably lose money.
But on the other hand, it is possible that TRUMP and MELANIA do have fundamental value. One simple way this could be true is if President Trump directs the federal government to buy his own memecoins with taxpayer money. He has already discussed the idea of creating a “strategic Bitcoin reserve” (i.e., using taxpayer money to buy Bitcoin). What if his reserve also bought a little bit of TRUMP and MELANIA? That would of course be incredibly corrupt, and in any well-functioning society it would be illegal. But in the age of Trump, who knows what will fly?
In fact, though, there’s another, much more plausible reason why TRUMP and MELANIA could actually have real, fundamental value. It’s because these coins allow people to give Donald Trump money without actually transferring him funds.
Suppose you wanted to buy a favor from Donald Trump, and he wanted to let you buy a favor from him. How could you do it? You can’t just pay him a giant bribe — that’s illegal. Maybe you could pledge him a bunch of cash for his presidential campaign. But there are campaign finance laws that will get in your way, and even if you succeed, he can only use the money for his campaign, not to buy yachts or whatever else he might like to use the money for.
Instead, what you can do is to buy a bunch of TRUMP or MELANIA. When you buy one of those memecoins, you increase the demand for the memecoin. Its price then goes up. This makes Donald Trump richer, without any money actually having to change hands.
Let’s think of a little example to see how this works. Suppose the price of TRUMP is $50. Then you buy $1M of it, driving the price up to $55 — a 10% increase. Through the magic of mark-to-market accounting, this means that Trump’s entire stake in TRUMP goes up by 10%. And since Trump himself owns almost all of TRUMP, this means he suddenly gets hundreds of millions or even billions of dollars richer on paper.
In other words, by spending $1M, you were able to make billions of dollars appear in Donald Trump’s crypto wallet. And it’s all perfectly legal, too — you didn’t actually transfer a single dime to Trump’s bank account or his crypto wallet or anything. All you did was buy something that he owns and drive up the price. And Trump will know you bought it, too, so maybe he’s inclined to do you a favor in return.
Note that TRUMP is perfectly set up to do this. The “float” — i.e., the amount of TRUMP that actually gets traded on the open market — is very small compared to the amount that Trump owns. The smaller the float, the cheaper it is for one rich guy to come in and pump up the price.
Now, can Trump actually get all of those new billions of paper dollars out, and turn them into actual dollars he can spend? No — at least, not quickly. For that he needs what we call exit liquidity — he needs to find someone willing to trade him USD for TRUMP. And as soon as he starts selling huge amounts of TRUMP, the price will crash hard, eliminating much of his paper wealth. That’s the downside of the tiny float. Live by the mark-to-market accounting, die by the mark-to-market accounting.
But in the long term, he can get a substantial portion of it out. Remember that every time someone wants to pump up Trump’s wealth, they have to trade USD for TRUMP. Trump himself could sell some of his TRUMP to the guy who’s trying to do him a favor, but this might violate some pesky anticorruption law. Instead, there will be middlemen — people who pay Trump for his TRUMP whenever he wants to sell off a bit of it, and then hold onto it, knowing that eventually they’ll be able to sell it at a higher price to someone who wants to pump up Trump’s wealth. These middlemen look like speculators, but they’re really just market makers.
The exit liquidity doesn’t have to be an idiot or a speculator or a compulsive gambler — it can just be the next guy who wants to scratch Trump’s back. And so on and so forth, into infinity. Trump is old, but he has kids, and he will presumably leave them his crypto when he dies. This has the potential to sustain the value of TRUMP indefinitely, without it being a speculative bubble.
Could this run afoul of antibribery laws? I’m no legal expert, but it seems to me that if I buy a lot of TRUMP, and the cops and prosecutors demand to know why I did that, I can just say “I thought it would go up!”. And who could prove otherwise? And if Donald Trump then does something that helps my business, well, he just appreciates a smart crypto investor, right?
And now here’s the thing — this doesn’t just work for Trump. It works for anyone. All you have to do is to create your own memecoin, and hold onto most of it, and float a small amount of it on the market. Now anyone who wants to give you money without paying gift taxes or running afoul of anticorruption laws can do so simply by pumping your memecoin. Now, most people don’t have hordes of favor-seekers trying to give them money, so in practice, only famous and/or powerful people will be able to pull this off.
In other words, memecoins are a form of payment technology — not the normal kind of payment, but an approximate, sloppy one. The uncertainty about how much money you’re actually giving someone by pumping their memecoin is the price you pay for the immaculate, plausibly deniable nature of the gift. It’s a little bit like a financial hawala — a way of sending money through informal relationships without ever formally transferring it.
This is, of course, an invitation to vast corruption. But with Trump in power, it seems doubtful that the Department of Justice will crack down on this sort of thing anytime soon. Meanwhile, the possibility of memecoin pseudobribery means that these coins might actually have fundamental value after all. Unfortunately, as with so many other crypto use cases, this one would be all about evading the spirit of the law.
Correction: An earlier version of this post claimed that Trump’s stake in TRUMP peaked at a value of over $50 billion on the morning of 1/19, as reported by many news outlets. But the CoinGecko data that the reporting was based on appears to have since been revised, and the true amount of Trump’s stake appears to have only been about $10 billion. The remaining amount was based on the coin’s “fully diluted” value, which includes all the TRUMP that Trump might create in the future. Creating those coins wouldn’t just flood the market and lower the price; it would automatically dilute holders of TRUMP, lowering the value of their stakes. By the rules of mark-to-market accounting, stock shares that haven’t been created yet have no financial value, so we’ll stick with the same definition for crypto. But do note that the ability to create shares has an option value that’s not reflected in typical mark-to-market accounting.
The authors were too polite to call them “idiots”, but only just barely.
