S9 Ep1: Trump, trade, and AI growth

Today on Vox Talks Economics, Trump, Trade and AI Growth. Welcome to another of our special episodes recorded at the CEPR Annual Symposium in Paris. My name is Tim Phillips. The Trump Administration says it wants America to lead in AI. But what does that mean in practice for trade and productivity? Gary Gensler of MIT and CEPR and also, of course, a former chair of the Securities and Exchange Commission joined me to unpack the Administration's AI Action Plan, work out what's happening to export controls and untangle the deal-making geopolitics of AI hardware. We also discussed whether AI will make growth great again or just inflate a short-term capital spending boom. So Gary, welcome. It's good to be with you, Tim. We spend a lot of time talking about AI. The impact is going to have on productivity on the economy. But in policy terms, has the Trump Administration made big moves on regulating AI? Tim, let me just start by saying, I do believe that artificial intelligence is a transformative general-purpose technology. But it's not new. Some of the first seven of work was around when I was born, and I'm not the youngest dude around any longer. But particularly the last 15 years, it's widely adopted in finance and in parts of the economy, the streaming apps like Netflix and all the big tech companies. And my gosh, even the postal services using it to a postal service. Yes. You would think, no, they can't be, but just to figure out our handwriting. But of course, all the buzz is about generative AI in these last two or three years and will it take all our jobs away or lead to some version of dystopian futures? But back to the question you asked. It's interesting. US policy prior to Donald Trump taking his second term was pretty forward leaning towards the continued development. And I stress continued development of artificial intelligence. We did not have any national law like in the European context where they passed the EU AI Act. Have some smattering of domestic state laws around privacy about fake intimate pictures of any of us. I think that's a good thing that there's laws about that. But the Trump administration came in and I would say generally speaking is a bit more deregulatory, but it's not a dramatic shift for artificial intelligence. And they've continued leaning in to being restrictive on exports, which actually started in Trump 1.0 against Huawei, the large Chinese technology company, President Biden leaned even further into export controls. And I would say that President Trump is there as well. So you'd say what's different? There's a lot of big narratives, big talk, supporting artificial intelligence from this administration. And mind you, they're using it as a geopolitical negotiating chip with mid East nations and even with China. But by and large, it's a bit deregulatory, less concerned about response for AI. And yet still pretty tough on export controls to China. He came out with an AI action plan within the first six months. What was in the AI action plan? What was it for? Monswell questions. What is it for? It's ultimately like many doctrines of this by Democrats, Republicans, liberals, conservatives and elsewhere. It's a bit of a message document. It's a political document. We are for something and we're against something and revs up news stories and podcasts like this. When you ask what it's for, I am and have lived in the political world. But the substance of it was a pro-American AI policy fitting in with the overall president's view of America first or so-called making America great again. And so in that context, going lighter touch on regulation, lighter touch on the ethics bias and fairness and privacy issues around AI. Lean heavily into saying that we should be supportive of data center and infrastructure development and make less sand in the gear to development. And it did speak to some of these issues around export control and having an American first policy. Interesting for those European listeners. It also as Vice President Vance had in an earlier speech in Munich said, to Europe, we think you've been too restrictive in these areas, too regulatory. So here is the US and the AI action plan speaking to Europe. And then again, just in December of 2025, they spoke to Europe again in the US publishes this thing called the National Strategic Strategy. Once a year, but the first year of president does it, it really speaks loudly to the world. And boy, did President Trump speak loudly towards the end of 2025? On a lot of things, but in part was about artificial intelligence that the US needs to be first amongst this push and this technology. The impact of AI is something that Gary's colleagues at MIT have had a lot to say about on Vox talks over the years. We spoke to David O'Tour on automation and the value of expertise in July 2025 and in June 2023. We also spoke to Doron Asim Moglu about power and progress. The book he wrote on the influence of technology on society with his colleague, Simon Johnson. Find these episodes and much more on the influence of AI when you subscribe to Vox talks wherever you get your podcasts. So let's look at how it's intending to be first. Now those exports that you talked about, the export strategy, it's been pretty tough on what can be exported to other countries, particularly China. What are the restrictions here? How strong are they at the moment? President Trump and his first term had already put some restrictions on technological exports to some Chinese companies, Huawei being the lead. President Biden brought that out and had a three tier process where some nations like China would get very little if anything. Some nations were in the middle and then the top tier were allies. I would say President Trump's plan has looked at all of this in a more transactional way. A part of a tool set for this administration to use to achieve other geopolitical goals. As it relates to China, it's still pretty tough restrictions. Some of the leading chip manufacturers like Nvidia have pressed hard that they want to be able to sell their middle tier chips to China at first early in 2025. The president's team said no and Nvidia took a big $5 billion right off on that because they couldn't sell the chips to China. Then the president said, well yes, maybe these lower tier H20 chips can go. And then interestingly the Chinese government said to their major state enterprises and even private enterprises, not so fast. Maybe we don't want you to be buying these Nvidia chips and I know people like the game of soccer. The ball keeps bouncing back and forth. And of course China has a dominant 90 plus market share. Dominance on a really important piece of the supply chain called rear earths. Yeah. More importantly the refining and processing of rear earths. Even more importantly the magnets and so China also exerted that economic leverage. So we are now in a world of competing interest using economic leverage to extract that from other countries. And the US is doing that to Europe in a more broad way. But interestingly on AI when the president went in 2025 to the Middle East, he also decided in the spring, well maybe I'll relieve some of the stress here. I'll allow more chips to go to key Middle East during countries as part of an overall negotiation. But was it a negotiation to promote AI presumably yes or was it a negotiation that maybe ended up with a multiple hundred million dollar airplane being gifted to the US. It's all part of a broader tactical transactional presidency. You've been inside government Gary when you're looking at some of these deals that are being made around AI technology and of course famously now Nvidia's paying 15% to the government of its sales of eight 20 chips in China and the Chinese government is said not so first We don't want to buy us money. We don't want to buy it. Yeah Easy to all about doing the next deal the next thing along making a small advantage. I think it's accurate to say that every leader in probably history has a mixture of strategy and tactics or one might say strategy and transactions. This presidency really does pride itself when that's transactional approach. Yeah And so I think that a lot of what's going on in artificial intelligence does fit strategy there is a broad strategy of this administration and I would say even the prior administration the Biden administration to support ship development in the United States. President Biden had a big bill that went through the US Congress to do just that and support it. It was bipartisan a big strategy to support AI development in the US but I think there's a transactional overlay that is even more significant in this administration about these deals whether it's in the Middle East or even with China as we saw when the two presidents met last year and President Xi and President Trump came away with a one-year truce on the restriction on rare earths and the US is very elevated tariffs. Two great nations are holding the proverbial guns to each other's head and they each say well all right we'll back away from that but the US still left very large tariffs on China and China is still trying to build their resiliency China has a remarkable capability of their computer scientists are families and they have more of them than Europe or the US just because the sheer size of China but also because of their educational system their culture this is not China of the 1990s or even of the 2010s they are really the world's hegemon and I will repeat the world's hegemon when it comes to rare earths electronic vehicles solar panels magnets the US is definitely the world's hegemon in terms of a deep and liquid capital markets hegemon in terms of the US dollar who are really are the most innovative economy and culturally the support of finance for that innovation but no one no one should count the Chinese out let's have a look at how AI is being used domestically your colleagues at MIT are leading the debate we've spoken to David Oatall we've spoken to Darron a subo-look what is very clear is we don't quite know yet exactly what effect it might have on domestic productivity where do you stand on this thank you for quoting two remarkable colleagues Simon Johnson who won the Nobel Prize for as well with Darron and Simon my co-editor of books that we put out in 2025 about the economic consequences of the Trump administration with CEPR and there's a range of views one of my former colleagues at MIT Eric Brennielsen Eric is in the end of the spectrum where he forcees about 2.5 percent productivity point gain over 10 years so this is a 10 year window not along one year about 2.5 points Darron is around 0.3 percent over 10 years you know so it's a wide range and David Alters a little bit more positive than Darron Simon is closer maybe to them I find myself in the middle universe of Pennsylvania has an estimate out late in 2025 about 1.5 percent productivity points in the next 10 years here's the question we already have in San Francisco and in Austin Texas and elsewhere significant portion of ride hailing you're hailing a waymo autonomous vehicle to take you around town how many years is it before the entire ecosystem of taxis are moved over what about the two to two and a half million Americans that work in trucking so there's going to be big transformations over the next two to three decades maybe not the next two or three years of labor and what we found in the past whether it was the canal era 200 years ago the railroad era 150 years ago electricity and the automobile 100 years ago or the internet 30 years ago in each of these eras there's big disruptions big waves of transformation and economies change I think we're in one of those eras so I'm more optimistic than Darron and David but I looked Darron one the Nobel I didn't so I still time no that's not my path but I do appreciate that to the extent that it's just about automation it doesn't boost productivity it just displaces folks here's the two things what happens to the people that no longer are in those roles the truck drivers or the office workers what else are they doing productively in society producing goods and services that we humans actually want to consume and two and this is where the big question mark is what does it do beyond automation because every general purpose technology in the past electricity the automobile the internet so forth it's real boost of productivity came not from automation alone but from what it does to transform an economy and is the Trump administration in its policies doing anything to deal with these effects this administration is deeply into what they would call individual sovereignty and national sovereignty and stepping away from supporting either countries or individuals that's a unified theme in this I would say they're leaning in to be a ultimate booster in chief for the technology but far less in terms of what it means for the working public it's a one thing that I really wanted to ask you about because of your experience is the AI bubble that everyone's worrying about at the moment there is a lot of market concentration in the sorts of technologies that are producing AI is the administration worried about the level of concentration I can't speak for the administration or what they think about I had the great honor to be in government when I was not supposed to speak about the markets but here I find myself I'm not in government and you can and I guess I can unambiguously artificial intelligence in 2025 boosted the US GDP growth just in the first half of 2025 JP Morgan estimated that it was well more than half of economic growth but here's the catch it's not about AI productivity it's about capital expenditures on all those data centers and buying all this Nvidia and AMD and TSMC chips we definitely have a boom going on and even if these accelerated higher spending on capital expenditure infrastructure just flat lines which is approximately four hundred billion dollars a year which is well over one and a quarter percent of US GDP even if it flat lines you're taking away an accelerant of growth if it reverses itself it dampens the economy here's the second thing about it the revenues don't currently cover those expenditures yeah look at open AI just for one example for all of 2025 and while it's not a public company but for all of 2025 it's estimated it might be somewhere in the vicinity of 12 or 13 billion of revenues and it's making commitments a company that's probably had a negative cash flow in 2025 of 20 ish billion it's making commitments not just in the hundreds of billions but add it all up and say closer to a trillion dollars yeah no don't got me wrong it's ambitious it's bold but we could definitely say that but will it play out so now back to the stock market unambiguously there's a disequilibrium the spending on AI is boosting the economy and it could be about four hundred billion dollars a year direct revenues matching that might be 50 billion somewhere that disequilibrium has to write itself there's an awful lot of concentration that you said that the largest players in the market have the largest market share not just in the stock market but in this ecosystem of artificial intelligence and thirdly in the stock market these companies are driving the US stock market rounded out in early December was about 70 trillion dollars the European total stock market 15 trillion this is not about American exceptionalism alone I mean I do think that US economy is quite strong and unique but it's also about valuations and you talk about concentration the top four of the NVIDIA Microsoft and Alphabet and Amazon in aggregate those four companies are worth all of the entire European stock market. You kind of go, what's a buy and what's a sell and I'm not giving investment advice. But I think the one big difference from prior booms around general purpose technology, railroads of the mid-19th century, automobile, radio and electricity in the 1920s, the internet of the 1990s. You can study each of these and they don't end well in terms of stock markets, whether it's the 1870s for 1990s. It's fascinating history. A debt-fed bubble like the 2008 financial crisis, housing bubble. Those end really hard. This boom or bubble, if you wish, is much more equity fed and the big concentrated tech platforms that have a lot of free cash flow. Where my worry is is part of the US economic growth will go away even if the capital expenditure is just flat-lined. The stock market valuations themselves are supporting a lot of the spending, particularly of the high net worth and the ultra net worth. The US economy right now is not being supported by the average median families consumption. The third thing I do worry about is there's a fair amount of debt building up in the AI ecosystem around the interconnectedness between Nvidia and others providing supply chain financing to the model makers. And remember, the model makers are just leasing the chips and the big data centers rush to build all these data centers into connected, isn't it? History tells us when you have problems and finance. It's usually around debt, interconnectedness, a lack of transparency and complexity. Those four together, if any one of them gets knocked over the chair falls in that four leg, it's still. So what you have here is a lot of this interconnectedness. You have debt into some of these neo-cloud companies like Core Weave and others. We don't have revenues yet to support this. Now maybe OpenAI's revenues will go up 10 fold in the next few years, but what if it only goes up threefold? And so you have this wobbly system, and I think that's a real risk. We're full for fuller on your measures of risk. I don't have a number. Thank you Tim for trying to get that out of me. I'm just saying that in the system, you have some very strong, relatively low debt companies like the Microsofts and the Googles and so forth, and in China, the very big tech companies. They're not debt heavy. If the values come down significantly, those companies will survive. The risk area is some of these second tier cloud companies, are generally called neo-cloud companies. Some of the model developers that have committed to lease chips down the future, I think it's the public itself, when the economy slows down if the valuations come off. Gary, thank you very much. Thank you. If you want to read Gary's chapter on AI and much more of course, then download the book, the economic consequences of the second Trump administration, a preliminary assessment from cepr.org. It's a free download and there will be a link in the show notes. Vox Talks Economics is a talk normal production. The assistant producer is Megan Viber and our editor is Andrei Zagarian. Next time on Vox Talks Economics, has AI Eton the economics major.