The Promise and Peril of Trump’s Big AI Plan

0:00So there's a lot of noise about AI, but time's too tight for more promises. So let's talk about results. At IBM, we work with our employees to integrate technology right into the systems they need. Now, a global workforce of 300 ,000 can use AI to fill their HR questions, resolving 94 % of common questions. Not noise. Proof of how we can help companies get smarter by putting AI where it actually pays off, deep in the work that moves the business. Let's create smarter business. IBM. Healthcare doesn't always work great. If you've ever waited on a refill or couldn't schedule an appointment, you get it.

0:36That's the kind of stuff Optum is changing. They're using data and technology to integrate patient care, pharmacy, and everything else. So healthcare is connected, not complicated. What's that look like? Cheaper prescriptions that are easier to get and care that looks at the whole person. How you need it. Optum is helping make healthcare work as one for everyone. Learn more at business.optum.com. Wise is the smart way to manage the currencies you need around the globe. When you send money abroad using your bank, you could get hit with hidden fees and exchange rate markups. There's a better way.

1:09Try Wise. Wise uses the exchange rate you'd usually find on Google with no unwelcome surprises. Plus, most transfers happen in under 20 seconds, which means your money arrives in less time than you've been listening to me. It's simple and free to sign up when you download the Wise app. Be smart. Get Wise. T's and C's apply.

1:31Bloomberg Audio Studios. Podcasts, radio, news. America is the country that started the AI race. And as president of the United States, I'm here today to declare that America is going to win it.

1:57Stephanie Flanders:I'm Stephanie Flanders, Head of Government and Economics at Bloomberg, and this is Trumponomics, the podcast that looks at the economic world of Donald Trump, how he's already shaped the global economy, what on earth is going to happen next. One thing President Biden and President Trump and any other elected US politician can agree on is America needs to win the race for AI supremacy. You just heard Donald Trump say as much at the administration's recent AI summit, and he went on to vow to do everything it takes to achieve that. So we're asking a simple question. How's he doing? At that summit, he unveiled a 23-page AI action plan, which, to one expert, read like an industrial-scale mobilisation order.

2:44Stephanie Flanders:The emphasis was all on freeing up companies to move fast and innovate, getting rid of the red tape in the way of building data storage centres, junking Biden-era guardrails on safety and pretty much everything that might be standing in the way of big AI companies doing what they want to do. So will it work? And if so, at what potential cost? Well, naturally, we have two brilliant Bloomberg voices to tell us the answer and to consider whether other Trump policies that we often discuss on this podcast might be working against the administration's goals on AI. With me in the London studio is Parmi Olsen, the Bloomberg opinion columnist covering technology.

3:25Stephanie Flanders:She's a former reporter for the Wall Street Journal and Forbes and the author most recently of Supremacy, AI, ChatGPT and The Race That Will Change the World. That's the book that won the Financial Times Business Book of the Year award in 2024. Parmi, so great to have you on. Thank you for having me. I think the last time I saw you may have been at the award ceremony itself. Yes, that was nice. It's a very pleasant surprise for you to win. And in Washington, D.C., Michael Deng joins us. Michael's the geoeconomics technology analyst for Bloomberg Economics, and he previously served in the Biden administration as a policy advisor and presidential management fellow in the CHIPS program office at the Department of Commerce, among other things.

4:03Stephanie Flanders:Michael, welcome to Trumponomics, your first time too. Thank you. Great to be here.

4:12Stephanie Flanders:PAMI, since you have just written this book, and we have heard a lot about DeepSeek in the last six months or so, if the president wants the US to win the race in AI, is the US still the leader in this technology? Or did DeepSeek show us that China had already caught up? Well, it depends how you define leader. But I would say when it comes to AI capabilities, the models and how good they are at reasoning and math and coding, then yes, U.S. models are still ahead of the ones coming out of China. The thing is, they're not ahead by much. It's a very narrow lead. Now, when it comes to actual adoption of AI services, I would say that in China, it's much more sort of readily available.

5:00People use AI much more in their daily lives. And China's ability also to kind of create the energy infrastructure they need to build AI models. They're able to kind of spin up the gigawatts that they need. I think it's something like 400 gigawatts that they're already planning for AI that they have available versus, you know, dozens of gigawatts that the U.S. has. That certainly puts them in a very advantageous position. But as it stands, you know, the U.S. is still ahead. DeepSeek, of course, caused this kind of freak out moment for Silicon Valley back in January when that dropped, because here was a model that was not only as good as China GPT, but it was free.

5:39And the blueprints had been put on the internet for anybody to copy and develop on, which is absolutely what people have done. And that really spoke to why China's tech sector has moved so quickly, is there's been this real focus on open source or so-called open weight models. You know, when you keep your tech proprietary and secret, you kind of have to start from scratch as a company each time you build AI. But if it's open, companies are collaborating more with each other and they're building on top of one another's innovations. And then collectively, the whole industry can just move a lot more quickly.

6:12And that's exactly what's happened in China. But as I say, they haven't completely caught up, but they're close.

6:18Stephanie Flanders:Okay. So Michael, we heard at the beginning there at the summit that Donald Trump at least thought his action plan was all about maintaining US leadership and winning that race. What does this administration think is the answer on this? Yeah, so I think the AI action plan is interesting in that it does correctly identify some of the key areas in which the US really needs to develop further and is falling behind on China if it wants to keep up and win this AI race. One of the areas that it really clearly highlights is in power infrastructure. Parmi has already mentioned that China's being able to build on the hundreds of gigawatt scale and is bringing on much more power than the US is capable of bringing on to be able to match the demand that data centers will require.

7:00And this action plan specifically accelerates a lot of the permitting processes required to build these data centers. Though interestingly, it doesn't outright call for building new infrastructure. A lot of it is mainly about accelerating interconnection, optimizing grids, stabilizing grids, etc. But it's still headed in the right direction. The contradiction there, however, is that, you know, last year, 90 % of new power generation capacity added to the U.S. was all came from renewables. And the big, beautiful bill explicitly rolled back a lot of the clean energy tax credits that was going to support that sector of the economy.

7:32And another wrinkle in the AI action plan itself is that the aversion to solar and wind is still very clear in the text there. You'll notice that they highlight nuclear fission, geothermal, and fusion as the only new sources of energy they think should be developed further, while excluding solar, wind, battery storage, etc. So it doesn't really make sense, especially mentioning fusion, which is not close to being technically viable at this point. On workforce itself, it does target a longstanding gap in the U.S. ecosystem in that it's trying to build up that vocational workforce to operate and deploy AI.

8:06The issue there being, however, that a lot of President Trump's recent actions in the past few months have targeted that high-end research talent pipeline. That's really critical for advancing the frontier of AI research. Universities, government funding, and also immigration. Jensen Huang has said something of 50 % of all AI researchers are Chinese. So while you're building up the vocational workforce, you're undercutting or undermining the elite research workforce you need to really advance that frontier. And so again, a mixed bag on what it's actually doing. And the last thing I'll bring up is expert controls, which is a massive topic, lots of implications for foreign policy, how President Trump has acted so far in the past six months towards a partner as tariffs are involved as well.

8:49So very big topic. Right.

8:50Stephanie Flanders:Well, you have introduced the idea that there might be some tensions within the plan, but certainly between the plan announced at the summit and other aspects of Donald Trump's policies. And we are going to get into that. But, Parmi, you describe this as a gift to big tech. We know that the people involved in this seem to have played a central role in devising this plan. Your book is all about the people behind AI. I suspect that you have some insight into them as well. That's always the most interesting part. I think one of the most important people behind this plan is Jensen Huang, who is the CEO of NVIDIA.

9:28And Jensen, you see him on stage walking around with this leather jacket. He is not someone who likes doing lobbying in Washington, D.C., but he's had to do that in the last few months after the Trump administration came into power. And first of all, back in April, the White House enacted a ban on NVIDIA's chip sales to China. That's a huge problem for NVIDIA because it said it would cost$15 billion in revenue. And so there came now this effort from Huang to try and turn that view around. And he kind of found an ally in the Trump administration in David Sachs, who was the so-called crypto and AI czar.

10:05And both of them were really pushing this idea that actually we need to continue selling to China because if we don't, China's going to and Huawei are going to build their own chips and then they're going to build their own AI services on top of that. And guess what? They're going to sell all that AI technology to other countries around the world and they're going to be the world's leading AI provider. What we need to do is kind of clutter the market in China with our own tech, with NVIDIA chips, so that Chinese companies become reliant on our own infrastructure. And it's really interesting that this is actually fundamentally a very Silicon Valley worldview.

10:43This idea that if you kind of blitz scale your product or service, dominate the market, eventually consumers, businesses, your customers get locked in into your ecosystem. And we've all kind of experienced that with Apple iPhones and Microsoft, the hyperscalers, you know, cloud services. It's actually hard for their customers. They feel they're locked in. And so this was the idea that Huang and Sachs really pushed in the Trump administration. Eventually, Howard Lutnick, the Commerce Secretary, also picked up on that. And we saw that play out with this big U-turn that Trump made on the ban on NVIDIA chips.

11:21He lifted it. So NVIDIA is now able to continue selling specialized sort of export controlled chips called the H20s. They're kind of framed as being low power, not as good as the ones that American companies can use. But actually, they're still pretty powerful and Chinese companies can still do a lot with them. And so I think that's raised a lot of questions in Washington among kind of traditional China hawks in the administration about, is this really a good idea?

11:46Stephanie Flanders:And I think, Michael, you've described this as a strategic pivot from technology denial to export dominance. You know, this idea that instead of trying to keep China from getting the technology, you make everybody, the whole world, including China, absolutely dependent on it. I mean, that seems to me a major shift, not just from policy a few months ago, but actually from the Biden administration's kind of instincts on this of actually trying to put controls on technology. But I know that there's different levels of export control involved here. So just talk me through it a bit, Michael. Is this a complete change of strategy or is it a bit murkier than that?

12:22Absolutely. I would characterize it as very murky in the sense that if you notice in the AI action plan itself, there's a section specifically for export controls. And that section almost looks like it was written by the same Biden administration export control specialists and thinkers. That you used to work alongside. Exactly. In that one, it hints and signals at targeting component subsystems of semiconductor tools. And what that means is, if you think about the progression of export controls over the past few years, we essentially went from controlling the chips, then to controlling the tools.

12:55And now the administration is thinking about controlling the components that go into those tools. So again, with those components, there's a heavy reliance on the U.S. and Western suppliers. So it is an avenue of vulnerability for China in terms of building up their domestic semiconductor manufacturing ecosystem. But this has been a direction that export controls have been going into for a while, and those in the field have been seeing if this is going to happen. But it really does seem to be at odds with letting H20 chips go to China. And now I can see maybe like a targeted strategy of we'll deny them access to building up their own manufacturing ecosystem while still letting them have access to chips so a Huawei competitor doesn't build up.

13:35But on the flip side, what it now means is you've enabled the entire Chinese AI ecosystem to build off of that. And if you notice what Anthropic and OpenAI have said, they've always been fairly pro export control. Anthropic in particular has gotten into it publicly with NVIDIA, sometimes around export controls. Because they understand that if you give H20s and advanced compute to China, you're empowering their AI companies to become more competitive and to compete directly with DeepSeq, with Alibaba, with US AI companies. Quick one before you jump back in. You're listening for ways to help teams move faster, make sharper decisions, and turn scattered context into work they can use.

14:12ChatGPT for Business can help. Chat GPT for Business gives teams a shared workspace with admin controls, permissions, and access to work and codecs in Chat GPT. This means your business can move from question to answer and code to rollout quicker. Join over 10 million business and enterprise users worldwide already using Chat GPT for Work. Download the Chat GPT desktop app or contact sales to learn more. What if the vacation you thought you couldn't afford is already paid for? That's exactly what Point.me helps you discover. First, if you have credit card points, stop wasting them in your card's travel portal.

14:50Point.me searches 150-plus airlines and loyalty programs to help you book flights with far fewer points. It even shows you step-by-step how to book, making award travel simple whether you're a points expert or just getting started. With Point.me, you'll spend up to 90 % fewer points on your flights. That flight you thought would cost 100 ,000 points could be just 10 ,000 with PointMe. Visit PointMe, that's P-O-I-N-T dot M-E today and see where your points can really take you. PointMe, turn your points into the trip you thought you couldn't afford. Join Bloomberg for the Canadian Finance Conference proudly sponsored by National Bank of Canada Capital Markets on September 29th in New York.

15:34Hear from influential corporate and government leaders as they discuss the strategies shaping Canada's economic future. Connect with senior decision makers, gain actionable insights, and be part of the conversations driving business forward. Register at bloomberglive.com slash Canadian Finance. That's bloomberglive.com slash Canadian Finance.

15:55Stephanie Flanders:So, Parmi, I would imagine other parts of the sort of AI ecosystem are a bit concerned, I guess, on two fronts. that there's this quite abrupt change of strategy from the US and also that it is in itself a bit inconsistent and seemingly unstable. We often talk about in Trumponomics about the sort of uncertainty being costly for trade and for the real economy. But I imagine it's, if anything, more costly if you're trying to think about AI investments. Absolutely. I'm sorry, you said inconsistent and seemingly unstable. I feel like that should be the tagline for the Trump administration. And that's precisely like, as Michael was saying, when you read the action plan, the actual behavior of Trump doesn't correspond to what is being said in this plan.

16:38Tough talk on export controls. And yet he has...

16:42Stephanie Flanders:Given in on the most important one or the biggest, most obvious one. Arguably the most important one. I think this kind of uncertainty is perhaps not great for companies. But on the other hand, in Silicon Valley, there's this term accelerationist. So if you're an AI accelerationist, you want to move fast and break things. Essentially, you believe that AI development should speed up because the benefits to humanity will outweigh whatever risks and problems come up along the way. And what seems to be happening is the White House is adopting this accelerationist approach that has taken hold in Silicon Valley.

17:15You know, just even a year and a half ago, so-called AI doomers were lobbying very hard in Washington for more guardrails out of concern that AI could bring existential risk to human civilization. those voices are now completely quiet. They've been drowned out by the accelerationists. So yes, while the uncertainty isn't great, I think for tech companies, you know, AMD is another one in Silicon Valley, anyone making hardware, anyone selling to China, this is really good news.

17:43Stephanie Flanders:Michael, I do want to go back to you on how some of this sits within the broader economic policies, broader Trumponomics that we've seen over the last few months. And you referenced some of it at the beginning. Just talk me through how important the negatives have been so far on the issues that really matter to the development of AI and how that might sort of stack up against anything that was said at the summit. Yeah, absolutely. So I think the biggest issue with a lot of this is the AI action plan lays out very ambitious objectives and correctly diagnoses the areas on which the U.S. needs to work.

18:18But the environment and the ecosystem currently that the Trump administration has created with many of their policies is not conducive to running as fast as possible on AI. I've talked about the workforce issues already. I've talked about the energy issues already. But even if we're speaking about globally, for example, on export controls, you need close collaboration with a lot of these suppliers in other countries to be able to set up a regime that actually keeps China out of these key technology choke points. and the A action plan itself have some very coercive and muscular language in essentially saying that we welcome allies to collaborate with us.

18:54And if not, we're going to use secondary tariffs, the foreign direct product rule to induce allies to participate in this export control regime. And so that was something that really surprised me when I read it. It wasn't something that the Biden administration, when I was there, ever seriously considered. Though when negotiations got tough, it came close a few times, I will say. It essentially shows this Trump administration's view that they can force US tech supremacy on the rest of the world. They're going to do that through the export packages that they've contemplated, despite the fact that this is, at the end of the day, a very global industry, a very global ecosystem.

19:27And so it's very unclear to me how exactly they're going to make the US outrun China on this AI supremacy race if they're going to be so unilateral in their approach.

19:39Stephanie Flanders:Just a small footnote to that, and you've made the point in other analysis that you've done. If you're also making the US itself a more closed economy, it's going to be this sort of inside the fortress of America, you're going to build all this stuff and that's going to sustain the supremacy, but you're making it harder for those crucial inputs. I remember, Michael, you highlighted just that copper tariff. Just remind us that even that one piece is going to make things harder for a certain bit of the chips industry. Yeah, absolutely. Donald Trump in this AI action plan talks about building out AI infrastructure, data center, semiconductor fabs.

20:13All of those require a lot of inputs from abroad imported. Data centers use immense amounts of copper in power and cooling. Semiconductor fabs, even more so, you know, not copper just for power and cooling, but copper is found in chips themselves. And not just to talk about copper, we're contemplating Section 232 sectoral tariffs on semiconductors, and those might involve semiconductor tools themselves, all the different consumables and materials that will go into these fabs being built in the US that they need to actually produce the chips that go into AI. At the end of the day, chip manufacturing and AI data centers require a lot of global inputs that just doesn't square with this vision of a fortress America, this unilateral vision that Trump is all in for.

20:54Stephanie Flanders:I mean, you weren't at the summit, though I imagine that you had your spies there in one form or another. And was there any of this, these sort of tensions, parent contradictions voice there? I mean, even the sort of basic fact that all that he's talking about building this stuff and getting rid of red tape and all the bureaucracy at the same time as actually the imported components that you really need to make this stuff is becoming more and more expensive and harder to get. Was there anybody who was saying, hang on a minute? Not that I could see. I mean, I was checking out a lot of social media.