How to accelerate catch-up growth after transformative AI

Published on October 16, 2026 7:12 PM GMT This post was crossposted from ChatGDP by the Forum team, with the author's permission. Subtitle: Do everything you can to help your firms accumulate AI-enhanced capital In my last post, I argued that, if AI accelerates growth in rich countries, it becomes crucial to accelerate catch-up growth in poor countries. If we manage to do it, we create a lot of welfare. If we don’t, hundreds of millions will lag behind. This matters not only because people would miss out on consumption goods, but because they would suffer from hunger and deprivation and die earlier than necessary.[1] So I think accelerating catch-up growth after transformative AI (TAI) really matters! The multi-trillion-dollar question is: how? I think the answer is this: do everything you can to accumulate productive AI-enhanced capital. In the TAI world, frontier growth is driven by robots doing activities of economic value. Whoever owns the robots gets rich, so poor countries should try to own more robots. Throughout, I take TAI as given and ask what follows. I’m not predicting that it will happen, or when. This sounds obvious, but it’s often overlooked. I’ve seen smart people saying that countries should create a sovereign wealth fund to buy stocks in frontier AI companies, or resist the urge to do anything and let markets operate, or bet on sending their workers abroad so they will send remittances back.[2] So how can poor countries accumulate the AI-enhanced machines that drive growth? At a high level, the problem looks close to hopeless. Exporting gets harder, capital flows uphill, and poor countries get a relatively small growth boost from the first wave of cognitive AI. The export-led labor-intensive growth model that has reigned since the Second World War probably stops working. To get unstuck, we go down to the firm level. The core of this post is therefore a deep dive into the drivers of firms’ technology investment decisions. The exercise provides a map for countries to trace their own growth paths after TAI. The key task is to diagnose where firms get stuck and use policy to unblock them. Then countries can get into a positive feedback loop, where investment drives technological upgrading, which raises incomes, which funds more investment. The map doesn’t give one answer, but the answers rhyme. Run it on almost any poor country after TAI and the same workarounds keep showing up. I think they add up to a new growth model, the AI-intensive bootstrapping growth model. I am, of course, uncertain about this and would love to get a debate going. I hope this will help start an economic policy and philanthropic grantmaking agenda to help accelerate catch-up growth post TAI. How rich countries grow after TAI To answer how poor countries can grow in the aftermath of TAI, let’s recall how rich countries grow. As AI capabilities grow, machines become more capable of performing economic activities, so more tasks are allocated to machines, and the growth rate increases.[3] Higher automation leads to higher growth because, as the economy approaches full automation, its growth rate gets closer and closer to the growth rate of capital. Producing anything takes a bundle of tasks. Today, people do most of those tasks and machines do some. If you add machines but not workers, you run into the tasks only people can do. A factory with twice the robots and the same staff doesn’t make twice as much. Economists call this diminishing returns to capital, and it’s why capital accumulation has never been enough to make growth explode. Under TAI, automation erodes that bottleneck. The more tasks machines can do, the less each new machine is held back by the scarce human ones. At the limit, as Moll and Imas put it in a great recent post, “machines produce output and output is used to build more machines, so the economy keeps growing without labor being a bottleneck.” In economists’ terms, the economy becomes an “AK” model, and the growth rate can become very high.[4] Moll and Imas then argue this won’t happen soon, because diffusion is slow, most work is physical, and spending shifts toward whatever stays scarce. Their bottom line is that double-digit growth in the next 10 to 15 years is “extremely unlikely,” and that “a much more reasonable baseline is, say, 4-5%” a year. I am more AGI-pilled than that, but even this smaller number leaves catch-up enormously important.[5] How poor countries get stuck What I will paint here is close to the worst-case scenario. Things might turn out to be easier than this, and in later sections I discuss how prescriptions change if we relax the scenario. But if we want to create a robust solution to the challenge of post-TAI growth, we should start from the hard case. I expect frontier takeoff to create two big problems for poor countries. First, it gets much harder to export. Historically, poor countries used cheap labor to become more competitive in exports, but automation erodes that ad