与 Jacob Helberg 谈：为超级智能世纪重塑美国经济
Jacob Helberg 的核心宏观判断是，AI、放松监管与充足能源，可能推动美国从消费驱动的服务型经济转向高投资的工业型经济。 资本开支目前已超过GDP的2%，明年可能翻倍；Helberg援引的分析称，过去1年AI为GDP贡献了整整1个百分点，而经济增速为3%。
供应链安全要求保护西方生产商，避免中国先压低价格、再将价格拉回的能力造成冲击。 Helberg指出，90%的关键矿产依赖在中国完成精炼，半导体依赖台湾制造；他还提到美国国防部与MP Materials达成的7.5亿美元合作。锚定买方、承购协议和价格底线，是他应对“经典垄断行为”的方案，也能避免美国制造商再低声下气地向北京求取磁体许可。
Helberg认为，AI赋予美国工人“超能力”，将抹平发展中经济体的大部分劳动力成本优势。 企业本可以雇用更少的人，但他预计，竞争对手以及他所说的无限人类需求，会推动企业将产出提升10倍。杰文斯悖论意味着，更便宜、更高效的生产会创造更多需求。乐观情景不是大规模技术性失业，而是Sarah Guo提出的框架：“如果经济规模达到45万亿美元呢？”
“超级智能世纪”可能造成早期采用AI的国家与落后者之间的第二次大分化。 Helberg将欧洲在20世纪初占全球GDP的65%、如今降至约15%，与技术领先的中东作对比；他表示，阿联酋和以色列的人均GDP已超过法国，也高于韩国。他的判断是，欧洲一次次“错过了船”，而欧盟《人工智能法案》可能确保欧洲不会成为先行者。
下一场平台竞争，在于全球南方进口由NVIDIA主导的美国技术栈，还是围绕Huawei Ascend与DeepSeek打造的中国打包技术栈。 Elad Gil强调了开源风险；Helberg认为Meta的努力很重要，但称DeepSeek“并不是真正的开源”，并指称其在算力规模上撒了谎，拥有10亿美元级集群，还蒸馏了ChatGPT的模型权重。战略目标不是某一种授权模式，而是让最好的美国模型得到广泛使用。
核电是Helberg眼中美国在2030年代可能将发电量翻倍的首选路径，但融资时间才是决定性变量。 大型电厂成本高达数百亿美元，建设周期可能达到7年；Gil指出，抗议和监管延误能把5年或6年的项目拖成12年，利息和法律成本也会不断累积。Helberg支持加快审批，并为可信赖的外国资本进入关键能源基础设施提供明确的CFIUS路径，同时主张在“全方位并举”的方案中搭配天然气和清洁煤。
更广泛的投资体系，旨在奖励能源、矿产、零部件、芯片、数据中心、模型、应用和物流各个层面的建设者。 Helberg将政府过去6个月的政策描述为通过放松监管、减税和引入外资实施的“休克疗法”；自动驾驶交通则提供了跳过旧基础设施的路径。国防也是主要支出领域：全球军费达到创纪录的2.7万亿美元，但政府如何配置资金，将决定其军队是真正具备作战能力，还是“纸老虎”。
1. 中国的工厂车间实力，始于对供应链的控制
Helberg首先指出美国创新生态下方暴露的基础：90%的关键矿产依赖在中国完成精炼，半导体依赖台湾制造，整个体系容易受到地缘政治冲击。因此，供应链回流和与其他国家建立合作，是让美国建设者继续建设的前提。
他的因果链条超越双边贸易。中国进口非洲原材料，在国内加工制造，再向全球转口；“一带一路”以及中国在非洲和拉丁美洲的影响力，都源于其作为“世界工厂”的地位。他认为，纠正贸易失衡，也会削弱中国在这些第三方市场的杠杆。
Gil有一个重要澄清：稀土并不算特别稀有，美国、加拿大和印度都有矿藏。Helberg认为，瓶颈在精炼环节：中国2015年宣布《中国制造2025》后，大举扩充产能、向市场倾销产品，并挤压包括田纳西州、亚利桑那州和佐治亚州设施在内的竞争性精炼厂。
应对方案是商业结构，而不是反复强调稀缺性。美国国防部与MP Materials达成的7.5亿美元合作，将锚定客户与价格底线结合起来，防止中国先压低价格直到西方竞争者倒闭，再将价格抬高：“我们可以用承购协议解决这一问题”（“We can fix that with offtake agreements”）。
2. AI可能推动经济从消费转向生产
Helberg认为，两股力量正在同时到来：审批、税收和能源改革，以及快速进步的AI。美国有时长期保持70–80%的消费驱动结构，超过三分之二的经济活动来自服务业，制造业约占10%；制造业占GDP比重仍然持平，但他称这只是滞后指标。
领先指标是资本开支已超过GDP的2%，Helberg称明年“可能”翻倍。他援引的分析称，过去1年AI为GDP贡献了整整1个百分点；在经济增长3%的背景下，他认为这一贡献相当可观。全国能源需求也在2008年以来首次上升，此前总电力供应长期横盘；能源基础设施、基础工业品和国防支出正成为主要活动领域。
Gil质疑服务型经济走向成熟是否不可避免，并以德国及其他西方工业基地为例。Helberg同意，美国的前提最终变成了自我实现的预言：全球化带来了“横向”增长，而过去7年则通过创新带来了重新加速的“纵向”增长。
他用英国第一次工业革命作为历史样本：尽管人口更少，英国人均工业产出却超过中国50倍以上；中国占全球GDP的比重则从1800年的约三分之一降至1913年的7.5%。在他看来，真正的变量是技术，而不是人口或工资水平；AI可能重新激活这一变量。
3. Helberg的AI基准情景是生产率提升，而非裁员
Helberg将智能体AI归结为两种结果：如果完成一项任务所需的工人数量降至原来的十分之一，企业要么削减劳动力，要么将产出提高10倍。他预计会出现产出扩张，因为任何拒绝这一机会的企业，都会面对一个没有拒绝的竞争对手；同时，他认为人类需求没有上限。
杰文斯悖论提供了其中的机制：当技术让资源的使用效率大幅提高时，其相对成本下降，总需求反而可能上升，而不是收缩。他的乐观结论是，AI不会彻底取代人类，而会赋予劳动者“超能力”（“superpowers”），拓宽每个人能够生产的范围。
这正是Guo那个刻意挑衅式问题背后的乐观情景：“如果经济规模达到45万亿美元呢？”（“What if the economy was $45 trillion?”）
4. 超级智能重排国家与技术栈
按Helberg的框架，这个世纪的决定性事件不是“东方崛起”，而是“超级智能崛起”（“the rise of superintelligence”）。早期采用者可能制造第二次大分化，跳过行动迟缓的国家，同时压垮支撑发展中经济体50年的廉价劳动力优势。
欧洲是他提出的警示案例：欧洲占全球GDP的比重从20世纪初的65%，降至1980年代和1990年代的约三分之一，再降至如今的15%。他指出，欧洲人将问题归咎于1970年代的石油冲击，但他认为，欧洲错过了互联网、数字化和消费者应用浪潮；《人工智能法案》和数字服务税则让欧洲持续“自断后路”（“shooting themselves in the foot”）。
中东是“完全出人意料的剧情反转”（“total plot twist”）。Helberg称，阿联酋和以色列的人均GDP高于法国，也高于韩国。技术领先的领导力、资本和廉价能源，可能创造一种新的美式伙伴关系。只要制度框架能阻止中国接触这些集群，中东的算力项目就有助于缓解美国的能源约束。
Gil继续追问中国的开源模型、主权AI和国家支持。Helberg表示，Meta的生态很重要，但更广泛的竞争在于分发：各国可能不需要“极其昂贵的Blackwell芯片”，但它们最终获得以NVIDIA为中心的美国技术栈，还是Huawei Ascend与DeepSeek捆绑的中国平台，将决定全球市场份额与影响力。Helberg还认为，DeepSeek“并不是真正的开源”，因为据他所述，DeepSeek蒸馏了ChatGPT这一闭源模型的权重；他并指称DeepSeek在算力规模上撒了谎，拥有10亿美元级集群。
5. 核电融资是连接AI需求与能源充裕的关键
Helberg“毫不怀疑”核电是获得充足电力的最佳路径。他援引过去2个世纪中廉价能源与经济增长之间的相关性，指出美国电价只有欧洲的一半，并认为已有承诺的中东资本可能成为美国建设提升生产率的国内核电基础设施的资金来源。
Gil指出，尽管美国自1970年代以来几乎没有新增核电产能，核电目前仍供应美国约17–18%的电力。他回忆称，原本5年或6年的项目最终可能变成12年的建设周期，收入延迟，而利息、法律费用和超支成本会不断累积，并非线性增加。
Helberg的答案是政策确定性：缩短建设周期，减少监管障碍，并明确释放信号，表明CFIUS将允许可信赖的外国投资者进入关键能源基础设施。法国约75%的总能源供应来自核电，证明即使监管负担沉重，也可以实现有意义的规模。
Helberg转述Elon Musk的观点称，有些统计显示，数据中心可能要求美国在2030年代将整体发电量翻倍，而再工业化可能进一步推高需求。Guo提到，市场已承诺在2028年和2029年建设大型数据中心项目，并可能让单座大型电厂与一个大型数据中心匹配。Helberg仍主张核电与天然气、清洁煤并用，但坚持认为，规模化扩张路径“必然要经过核电”（“definitely runs through nuclear”）。
6. 建设者经济需要每一层，包括物流与国防
Helberg的行动地图是一座分层金字塔：能源、矿产、零部件制造、半导体、数据中心、模型和应用。他表示，美国在大多数层面都处于有利位置，但矿产、零部件和芯片仍是最大的暴露点。
交通运输是另一个战略领域。中国的“一带一路”连接起非洲资源开采、中国精炼和全球出口；Helberg希望美国重新思考过去通过巴拿马运河等项目进行的大型交通与物流投资，并利用自动驾驶系统“跳过旧基础设施”。
全球国防支出已达到创纪录的2.7万亿美元，其中60%来自美国、中国、俄罗斯、印度和德国。这个“万亿美元问题”在于这些国家买了什么：乌克兰战场显示，AI与自主系统正在改变战场结果；资金配置不当，则可能让军队沦为“纸老虎”。
Helberg将过去6个月的政府政策描述为推动国内建设的“休克疗法”：加快审批、降低税负、放松监管并引入外资。其目标是打造一个对建设者友好的国家，让美国的政策环境成为“资本的最佳目的地”（“the best destination for capital”）。
Today, Elad and I are here with Jacob Helberg, the Under Secretary of State designate for Economic Growth, Energy and the Environment; co-founder of the Hill & Valley Forum, which connects Silicon Valley to policymakers in D.C.; and author of the book The Wires of War: Technology and the Global Struggle for Power. We talk about what America needs to change about its global supply chain, why nuclear is the key to energy abundance, the return of American manufacturing, and superintelligence as a means for productivity and economic growth. Jacob, thanks so much for being here.
Thanks for having me.
Very exciting, in terms of your new role as Under Secretary for Economic Growth, Energy and the Environment. Can you start by telling us a little bit about what your initial agenda is or what you're excited about?
1. Securing Fragile Supply Chains
A few of the topics that I discussed in my opening statement at my Senate confirmation are focused on securing our supply chains. Our economy is extremely over-reliant on a supply chain system that's very brittle. We have 90 percent reliance on critical minerals that are refined in China and on semiconductors that are manufactured in Taiwan.
We have one of the best innovation ecosystems in the world, but that innovation ecosystem is sitting on top of a supply chain system that is very exposed to potential geopolitical disruptions. Helping move the needle to forge new partnerships with other countries to secure that is indispensable, as well as supporting ongoing efforts by the administration to reshore as much as we can right here in the U.S. That would really go a long way toward giving our builders the tools they need to do what they do best, which is build products people love that are disruptive and help grow the American economy.
What are some of those things that you think are most important to reshore? I think you also have a broader purview of the anatomy of the U.S. economy changing and mutating. Could you give us the big picture and then the specifics in terms of how you think things are shifting, and what you think is most important to bring back?
2. The CapEx Economy
One of the fascinating things about this current era and decade that we're in is that we're really seeing the juxtaposition between the macroeconomic effects of the policies implemented by the Trump administration and incredibly powerful technological shifts, especially in artificial intelligence.
President Trump came in and instituted a blitzkrieg of policy reforms, fast-tracking data center permits. On day 1, he rolled out an executive order to unleash American energy and facilitate and support a surge in production capacity for energy sources like oil, gas, and nuclear, as well as clean, beautiful coal, as he says.
The net effect of all these different policies, combined with incredibly fast-paced progress in artificial intelligence, is changing the makeup of our economy. We're starting to see that in the data. For a long time, the American economy was primarily a consumption-driven economy. At different points in time over the last few decades, we've been between 70 and 80 percent consumption-driven. More than 2/3 of our economy has been entirely driven by services, and about 10 percent has been manufacturing.
We're starting to see those numbers move. The manufacturing makeup of our economy as a share of GDP has remained roughly the same as of now, but that's a lagging indicator. The more interesting one is that we're seeing massive CapEx investment, as you guys know, that has really picked up in a statistically significant way.
It's over 2 percent of GDP right now, and it's probably going to double by next year. Part of that is the result of tax incentives. Part of it is just making it easier to get permitting, because, as you guys know, a lot of this stuff boils down to how you compress the window when you want to make a CapEx investment.
Businesspeople make a decision about whether it's going to take 7 years or 5 years to actually get something up and running. Compressing that window as much as possible really moves the needle.
Are there common areas of CapEx that you're seeing in particular? Is it defense, space, industrial, other types of manufacturing, or biotech? I'm curious if there's a clear breakdown of—
I've seen analysis that basically says that it's added a full point of GDP just for AI in the last year, which is a lot because the economy grew 3 percent. If a third of that came from AI infrastructure, that's a lot.
Energy infrastructure is another really big one. For the first time since 2008, we're actually seeing an uptick in national demand for energy. Our total electricity supply has flatlined since 2008, which is an interesting statistic that I've recently come across, and now we're seeing that trend change.
The other one is raw industrials, so things like mineral production. The DoD has forged a strategic partnership with a firm called MP Materials for $750 million to reboot domestic rare-earth magnet production.
The last one is one that you touched upon a lot: we're seeing a global trend across the world where governments are spending a lot more on defense. A record high was reached this year, with $2.7 trillion in global spending on defense. Sixty percent of that came from very large countries like the U.S., China, Russia, India, and Germany.
The big trillion-dollar question is: What will they spend that money on? The answer might actually define the shape of hard power in the 21st century. As you guys know, a government is a little bit like a company. If you don't allocate capital efficiently and actually end up wasting it, you could have a situation where some of these governments have militaries that are paper tigers.
You're seeing a little bit of a window into this in Ukraine, where AI and autonomy are really changing outcomes on the battlefield. Those different trends are incredibly fascinating to watch.
What made you decide that you wanted to make supply chain security one of the primary things you addressed early? How does the vulnerability to China and others for minerals and components end up mattering?
3. China’s Manufacturing Leverage
That's such an important question. The answer is that China is obviously a systemic rival. Some people say it's an adversary. Regardless of the flavor that one wants to characterize it with, they're definitely a rival. They have a fundamentally different view of how the world should be run than we do.
A lot of what they have been doing internationally to compete with us really flows from the fact that they are the world's factory floor. Their presence in Africa—I'm sure you guys have read articles lamenting how China has taken over Africa—they're all over Latin America, and they have the Belt and Road Initiative.
All of that is a by-product of the fact that they produce the lion's share of the world's manufacturing output. They import raw materials from Africa, manufacture them in China, and re-export them everywhere else. If you solve the trade imbalance issue with China, you actually address all of the peripheral issues with their influence in these third markets.
It's a national security issue to do that because their footprint in some of these places has proven problematic. But it's also good for our companies because, as we've seen with their export controls on rare-earth magnets, the last thing we need as a country is for our best companies to beg Beijing for permission to get licenses for rare-earth magnets in order to manufacture cutting-edge technological products.
What's the solution to that? If you look at rare-earth minerals, for example, magnets are sort of a subset, and they actually aren't that rare, right? There are huge deposits in Canada, big deposits in the U.S., and deposits in India. Fundamentally, they're not actually rare, but they're called rare earth. They're fundamentally mined in a small subset of countries that have access to them.
Should the U.S. be changing its mining policy around this? Should we be—should Canada? I'm curious how you think about addressing that, because there are a few different ways to approach it. One is just to mine more in certain places.
China only emerged as a rare-earth mining superpower about 10 years ago.
Oh.
It came out in 2015 with its Made in China 2025 plan. From that date onward, we actually saw China's refining activity of rare-earth materials skyrocket.
They pursued a very aggressive industrial policy to build refinery capacity in China, and they started to flood the market, which sank the price and started to really squeeze refineries located in the West, Australia, Canada, and the United States.
We actually have refineries. Historically, we have had refineries. There’s a huge refinery in Tennessee, and there are refineries in Arizona and Georgia. So the solution to help put the genie back in the bottle is, I think, actually the DoD’s deal with MP Materials, which offers a good template: you need an anchor buyer and an end customer, and you need a price floor.
So you need to agree with the supplier, in this case MP Materials, on a floor for a price, because what happens with these big offtake agreements, when a Western refinery tries to compete with China, is that China will artificially sink the global price, depress it in order to put Western alternatives out of business, and then raise the price again, which is classic monopolistic behavior. We can fix that with offtake agreements and a price floor, and I think the MP Materials–DoD deal offers a good blueprint for that.
When you project forward, as you’re talking about some of the leading indicators on CapEx and what’s possible in terms of reengineering trade flows, if it’s not just consumption, what do you imagine the makeup of the American economy can be in terms of manufacturing and other elements over the next 10 years? I think a lot of people took it as a given: “Oh, American labor costs are just too high; it’s a service economy now.”
4. AI Rebuilds American Industry
In school, they used to teach us that it was almost part of a natural evolutionary process: when you reach a mature stage in economic development, your economy evolves into a service economy, and it’s just the natural order of things. And I actually think AI offers advanced economies, so to speak, a massive opportunity to violate that narrative.
Isn’t that narrative traditionally violated by Europe as well? So if you look, for example, at the German industrial base or—
Totally.
There are lots of examples in the Western world where that didn’t happen—
Yeah.
The underlying premise was something that became self-fulfilling in the US, but didn’t necessarily translate into a number of other Western economies at all.
Completely. I think the fascinating thing is, as you guys know, in Peter Thiel’s *Zero to One*, he talks about how you can either compete vertically or horizontally. Horizontally is globalization; vertical competition is innovation. And I think the basic paradigm is that, for a lot of the 2000s, we were really growing our economy horizontally through globalization, and we weren’t really growing a whole lot vertically.
The interesting thing is that over the last 7 years or so, I would argue, vertical growth has actually picked up a lot. And to really appreciate the potential impact that AI can have on productivity, if we increase productivity, it will totally erode the competitive advantages in labor costs that developing countries have. We have an opportunity to reindustrialize.
To appreciate the extent to which AI can give us that opportunity, I think you can look at history and the First Industrial Revolution, when industrial output in Britain rose because Britain industrialized and China at the time did not. It’s an interesting comparison because Britain was obviously a tiny country from a population standpoint and a much more advanced country than China. But because of technology, Britain had an industrial output per capita that was over 50 times the industrial output per capita of China.
Britain’s GDP far surpassed China’s. China’s GDP in the 1800s totally collapsed. It went from being about a third of the world economy in 1800 down to about 7.5% in 1913. So it just shows the power of technology.
Today, you can see differences between Israel and Nigeria. Nigeria is a huge country from a population standpoint, and Israel, a tiny country that’s smaller than New Jersey, has a bigger GDP and a stronger military, and it’s all because of technology. So the people who say that we can’t reindustrialize because China has a bigger population or our people are more expensive, I think are totally missing the plot.
I believe that AI will, far from replacing humans altogether, actually give workers superpowers and massively increase productivity. And I’ve become somewhat fascinated with this macroeconomic theory called Jevons Paradox, which is the basic economic principle that when you have a technology that massively increases efficiency, demand for a resource actually increases. It doesn’t decrease because the relative cost of that resource goes down. So that’s my basic take—my optimistic take—on manufacturing in America.
So I feel like you are perhaps the first policymaker I’ve talked to whose first instinct on AI is that it’s about productivity versus addressing some very real risks. But you said to me, “What if the economy was $45 trillion?” Right? There are historical analogies for that kind of increase in productivity. But as you also recognize, it’s not the dominant narrative today around AI. What do you think will help more people see that opportunity, or what do you think they should understand about that potential productivity gain?
Yeah. I guess the way that I would frame it differently is, if you believe that agentic AI is going to make each individual worker able to do a lot more stuff, if you’re a company or if you’re a country, you’re basically looking at 2 outcomes. If you’re a country with a GDP of $10 trillion, and all of a sudden you only need a tenth of the workers to perform the task that 100 workers previously were able to perform, you either need a lot fewer workers, or the totality of your workers will produce 10 times more.
And I actually think companies will choose to increase output because, if they don’t, their competitors will. If you believe, from a first-principles standpoint, that human wants are unlimited—which I would argue they are, just look at everything we consume today compared to our grandparents—I think we’re looking at a world that’s just going to produce a lot more stuff.
Workers will do a much bigger range of things because of AI. So I’m actually quite optimistic about the future of work.
You’ve talked a little bit about this being the superintelligent century. Could you explain what that means? Has it started? Is it about to start? How do you think about that concept?
5. The Superintelligence Century
We’re starting to see the contours of a totally new world. If you think about the narrative 10 years ago, it was that the 21st century was going to be the Chinese century, or the century where the East rises, so to speak. What it’s proving today is that the defining feature of this century is not the rising of the East or the rise of China. It’s really the rise of superintelligence.
And the way we’re seeing this change the global landscape is, first, we’re likely to see a second great divergence. For the first time since the First Industrial Revolution, I think we’re going to see the economies that are first movers in integrating AI into their economy reap massive productivity and growth benefits and start to leapfrog the rest of the world that’s lagging in AI adoption.
Number 2, a byproduct of this is a collapse of the cheap-labor advantage that a lot of developing economies have benefited from for the last 50 years. The third big feature, which really hasn’t been discussed a lot in the press, is that Europe’s economy has been collapsing.
The narrative today isn’t that China is rising. The plot twist that no one saw coming is that it’s actually Europe that’s completely collapsing. Europe’s economy went from being 65% of global GDP in the early 20th century to roughly a third in the 1980s and 1990s, and now it’s down to 15% of global GDP.
What are the drivers of that? Are there specific policy things that happened, specific decisions?
The Europeans blame it on the oil shock of the 1970s, but that was 50 years ago. Reasonable people have different takes. I would argue that they missed the boat on a lot of really big technological revolutions. They were very late to adopt the internet. They were very late to embrace the digital and internet revolution and the consumer app revolution.
And now, with the AI Act that the EU passed and the digital services tax, they just keep shooting themselves in the foot. I think the AI Act is basically single-handedly ensuring that Europe will not be a first mover in AI because it’s now subject to this incredibly punitive set of rules.
That’s actually a great segue to another interesting feature, which I find to be a total plot twist: the part of the world we’re really seeing surge is the Middle East, which is just really interesting because GDP per capita in the UAE and Israel is higher than in France today, which is wild.
It's higher than in South Korea. And so you're seeing parts of the Middle East actually emerge in completely unexpected ways. I think in the West, we've long talked about the Middle East as a war-torn region, a region that struggles with all kinds of geopolitical instability issues and regional conflicts, and you're seeing a totally new Middle East emerge.
You're seeing leaders in the Middle East that are super tech-forward. A silver lining of the recent conflict is that Iran's influence in the region being greatly diminished actually paves the way for a much, much more peaceful region that's not being held hostage every day by terrorist groups. So I find the Middle East fascinating.
The last 2 features are that I think the US and China are going to be locked in a very aggressive race to control the scaffolding of the AI architecture for the world. Obviously, the rest of the world at one point or another will need to import intelligence. A lot of them will not need the super-fancy Blackwell chips; they'll need the normal stuff. But who sells them that, whether it'll be NVIDIA or Huawei, will really make a huge difference.
Both companies and the Chinese will definitely bundle the stack, so they'll have AI out of the box with the Ascend platform, DeepSeek, and all of these Chinese tools. The Chinese are very good at aggressively competing for market share. So, obviously, having a strategy to compete in the Global South and third markets will be important. And the last is the one that we talked about earlier, which is the global rearmament across the world.
How do you think about open source in that context? Because really, a lot of the Chinese companies are pushing open-source models, and those are ones that can be optimized in all sorts of ways by enterprises and others, by governments, et cetera. There's a lot of sovereign AI rising.
In the US, obviously, we have Meta as a champion for open source. In Europe, there's Mistral. But my sense is that there's a lot of Chinese government involvement in some of these open-source models in terms of funding them, promoting them, or accelerating them. What role, if any, do you think the US government should play in our own sort of open-source AI ecosystem?
Well, I think we need to have a strategy to figure out how we promote the American stack overseas, and whether that's through open source or through other models. I would argue, again, reasonable people have different takes on what happened with DeepSeek. I think the basic takeaway of how DeepSeek achieved its performance was incremental efficiency gains. They lied about their compute capacity because they have a billion-dollar cluster, and they distilled ChatGPT's model weights.
So while DeepSeek is open source, I would argue it's not really open source if they stole the model weights from a model that is closed source. With that being said, it will definitely be an integral part of China's strategy to try to get market share by using the open-source ecosystem. In that sense, I think Meta's efforts are very important.
But I think having a holistic approach to making sure that we have the very best models and they're as widely used is super important.
What do you think is the relevance of the Middle East, given their level of investment and this set of leadership that's very forward-leaning on AI? I want to go broadly into energy, but what about the availability of energy for gigawatt data centers there? And, plus, in this fight, right? Is it a swing vote? Is it the capital that matters? Do you believe in these compute partnerships? Should they be a closer ally?
Yeah. Well, I think the Middle East actually has the potential to be a completely new kind of partner for the US. First of all, they actively say they want to move in a much more pro-American direction, which is obviously a good start. Second of all, as a country, we're energy constrained, so we can expand our energy supply, and obviously, the administration, along with the private sector, is actively working to do that. But that's going to take time.
And so if we want to move really fast, working with partners that have abundant, cheap energy offers our companies an opportunity to actually compete on raw energy power, combined with compute and speed, against China's approach. The trick there is really going to be finding the right framework that satisfies the security concerns that national security professionals have in Washington.
Some of those concerns include making sure that China doesn't get access to those clusters. But I think it's eminently doable, and ultimately, I know that this is something that the administration is looking at closely.
Maybe we can move to that then, given that we're at, at best, low-single-digit actual energy production growth in the US right now. People have said numbers like we need to double energy production in the United States, maybe beyond that if you believe we're going to be a manufacturing hub again. What's a feasible way to get there and close that gap?
6. Nuclear Enables Energy Abundance
I think we need nuclear energy. There's no doubt in my mind that nuclear energy offers the best path. And this is where a partnership with the Middle East could also be very interesting, because the president has done a superb job securing very, very large foreign investment commitments in the United States.
There's a lot of room for those commitments to be channeled toward productivity-enhancing areas, and I would argue that energy—nuclear energy infrastructure—is a productivity-enhancing area because it makes our electricity and energy supply more abundant and cheaper. We know from 200 years of history that there's a direct correlation between the cost of energy and economic growth.
The cost of electricity in the US is half of the cost of electricity in Europe, and we see that difference play out in GDP growth. The challenge is that large nuclear plants that actually produce a lot of energy take a lot of CapEx. They're dozens of billions of dollars. They also sometimes take 7 years to build.
Now, the administration is doing a superb job at taking a very hard look at regulations and figuring out ways of actually compressing that window. But it's still very capital intensive, and so working with partners, including in the Middle East, to make sure that we actually get those projects capitalized could really move the needle.
I think it's kind of interesting because, in the US, we're still at 17% or 18% nuclear power—
Yeah, that's right.
—from the perspective of the overall base, and we haven't really added any capacity, roughly, since the '70s.
Yeah.
Fifty years later, we're at, like, 17% or 18% of all of our output, which is kind of amazing, with minimal accidents and high safety profiles—
Yeah.
—clean energy. It's dramatic that that didn't really take off as an energy source. What do you think is the path to actually deploying more nuclear? Because I know that there are some initiatives from the DoD, and there are some initiatives more broadly from the government. Is there an initial entry point or starting point to actually start to rebuild our nuclear industry, or do you think it's still TBD in terms of the right policy approach?
I think it really starts with policy uncertainty. Compressing the window has a huge impact on the cost analysis that—
Yeah.
—a lot of investors make when they decide whether or not to invest. And then, as far as pockets of money go, there are a lot of—
Yeah. Sorry to interrupt, but—
Yeah.
—but to pause really quick on that first point: I looked into this years ago, and I remember seeing that a lot of the cost of nuclear is actually financing costs and overruns. You start building a plant, there are protests and other things organized against the plant, and other regulatory red tape suddenly crops up. A 5- or 6-year project suddenly takes 12 years, and you have huge CapEx loans that are put out against that.
And so, to your point on the timeframe, A, you're losing time on ramping up the actual plant, right? You start producing energy later and making money later. But also, those delays are incredibly costly from a financing perspective.
Yeah. And as you know, when you have delays like that, the costs compound because you're paying interest on loans and legal fees. It's not even just a perfectly linear extension of the cost. Your costs actually go up on a compounding curve.
The time value of money with nuclear energy investments is actually super valuable. One of the ways in which the US government has an opportunity—and, if confirmed, I hope to help play a role—is through a government body called CFIUS, which has historically scrutinized foreign investment in critical infrastructure, including nuclear facilities.
My hope is that there's an opportunity to actually create partnerships with strategic foreign investors to absorb foreign investment and use that capital to boost our domestic energy supply. Because there is so much capital that's been committed, it would be beneficial for the country to use some of it to expand our overall electricity supply.
Would you imagine the administration being more directive in this area? Because if you look at some of the analogies, it doesn't have to take a decade to build a nuclear plant, including in first-world countries like South Korea. They chose reference designs, and they made industrial policy about it. How do you think about the feasibility of that sort of directed investment in the US?
Yeah.
I think it's very promising to me that even the scale and shape of demand really matches nuclear in the US. I look at data center demand all the time, but people are very committed to large-scale data center projects in 2028 and 2029. That's not quite long enough, but you have the desire to build data centers that actually take all of the energy from a single large nuclear plant. The matching problem should give us a huge advantage here.
I mean, I think it's eminently doable, and I actually think it's doable with just the right incentives. I think there isn't even that much state-led direction that's necessary. If you really reduce the regulatory barriers and the costs, you can create the right environment to direct a lot of that capital.
Then the government has to signal to the market that the Committee on Foreign Investment in the United States, or CFIUS, will not block foreign investment from trusted partners into this sector. Energy has historically been considered, rightfully so, critical infrastructure, and therefore foreign investment is subject to all kinds of scrutiny. But we can channel investment from trusted partners into this sector to grow our energy supply.
The one point that I'd add is that the one thing France got right in the last 45 years is that it actually gets 75% of its total energy supply from nuclear. They don't have natural gas like we do, and they don't have oil rigs like we do. We're very blessed as a country because we have a lot of resources.
But they prove that even in a country that has an insane regulatory burden, you can get really statistically meaningful amounts of electricity from nuclear. Elon rightfully pointed out that some statistics say our data center capacity will require us to double our overall electricity production in the 2030s. I think it's possible to do that, and, Sarah, you pointed out that if we want to reindustrialize, those numbers might even be higher, and that's totally true.
The way we get there definitely runs through nuclear. It also runs through natural gas and clean coal. I think we really need an all-of-the-above approach, but nuclear provides a massive amount of very low-cost energy. Whatever we can do to turbocharge that would be very meaningful.
One last question for you on what other parts of the economy you focus on. There's energy, there's obviously intelligence, and there are inputs like rare-earth magnets and minerals. What other domains do you think are essential for competitiveness from a security or strategic perspective?
7. The Strategic Supply Chain Pyramid
I tend to think of my work as being very supply-chain-focused because it gives me a mental framework for thinking about these issues holistically, by looking at the supply chain as a layered pyramid that includes energy, minerals, component manufacturing, semiconductor manufacturing, data centers, models, and apps.
As a country, we need a strategy that's holistic across the different layers of the supply chain. We're actually in a really good position at most of them. We have abundant energy, although we need to increase our supply. Our biggest exposure points are component manufacturing, semiconductor manufacturing, and minerals, and there's a lot that we can do to move the needle there.
Transportation logistics is another really interesting area where policy can actually play a role. The Chinese have been masters, through their Belt and Road Initiative, at having a supply chain plan that includes a global transportation and logistics network to get minerals from Africa back to China, refine them in China, and export them back everywhere else.
I think we need to do what we used to do with the Panama Canal: make these big investments in transportation and logistics infrastructure. The president's appetite for having a very robust economic policy agenda is exciting because it gives us an opportunity to take a hard look at things that, as a country, we haven't done in a while, including reimagining how we move goods in a supply chain system that looks different from the one that we have today. We can use technology to leapfrog old infrastructure. We can use autonomous technology to leapfrog old infrastructure, so I think there's a lot of opportunity there.
One last question for you, Jacob. We have a predominantly tech-focused audience. What should they understand about the way they should interact with the administration and the administration's stance on the technology industry's role in economic growth over the next few years?
Sure. Part of what we've seen over the last 6 months is that this is fundamentally a builder-friendly administration. We have a builder in the White House, and that's really been reflected in the policies rolled out.
Fundamentally, the policies of the administration have amounted to shock therapy to help facilitate building in America as much as possible: removing roadblocks through deregulation, lowering taxes and the tax burden, and promoting foreign investment in the US. Ultimately, the job of the White House is to empower builders as much as possible and make America the best destination for capital. I believe that we've fundamentally already started achieving that.
Amazing. I think your view that America can be a country, potentially, of builders rather than just services is also really compelling in terms of broader opportunity.
David Sacks and I hosted an AI summit with the president not too long ago, and it was incredibly inspiring to see the president declare that America would win the AI race. In that statement, he acknowledged that we were in a race, and he declared that America started the race and that we're going to win it.
It's inspiring because, in a way, it was reminiscent of John F. Kennedy's moon speech. That's the kind of optimism, bullishness, and boldness that we need from the White House, and I think it's eminently reflected in policy. This is a pro-builder administration.
Great. Thank you, Jacob.
Thanks so much for having us.