The Geography of the AI Boom
The AI investment boom does not spread evenly just because a model can be accessed from many places. This chapter compares the complementary conditions that let firms and countries absorb AI—compute, energy, data, skills, finance and institutions—and traces how those differences become market concentration, trade effects and divergent development paths. The practical judgment is conditional: preparedness creates an advantage, but the way policy manages contestability and diffusion determines whether that advantage becomes broad capacity or a more concentrated source of dependence.
Access is the first threshold, not the final outcome. Two firms may be able to call the same model while receiving very different economic value from it. One has reliable power, data, skilled workers and processes that can be redesigned; the other faces expensive connectivity, weak managerial capacity or uncertain rules. The difference is not an abstract “digital divide.” It is a difference in how quickly a unit of AI capability can be turned into dependable output.
What preparedness actually buys
Preparedness is best understood as a stack of complements rather than as a single technology score. Compute and data provide the immediate technical inputs. Energy, grids and data centers make those inputs available at scale. Skills and organizational capacity determine whether workers can use and supervise them. Finance allows firms to make the required investments. Institutions determine whether contracts, standards, taxation and public trust make adoption durable.
Each layer can be a binding constraint. A firm may have skilled workers but no affordable compute; a country may have data centers but unreliable power; a region may have energy but lack the networks and workers needed to use it. A government may fund infrastructure while lacking the administrative capacity to deliver support or monitor concentration. Preparedness is therefore not a medal awarded to a country. It is a diagnosis of which missing complement prevents the next unit of capability from becoming output.
The IMF's analysis places advanced economies in a stronger starting position because they tend to have more digital infrastructure, AI-ready labor pools, access to finance and institutional capacity. That does not mean every advanced economy will gain equally, nor that emerging and developing countries cannot catch up. It means that the same frontier improvement encounters different absorption conditions. A country with weak complementary infrastructure may spend scarce resources importing AI equipment while capturing less of the income generated by its use.
The distinction between access and absorption matters for firms as well. A large company can spread fixed costs across many users, hire specialized staff, negotiate infrastructure and carry the expense of experimentation. A smaller firm may face the same model price but a much higher total cost of integration because it lacks data systems, technical oversight or the slack to redesign work. If regulation and compliance are also fixed costs, they can reinforce the advantage of scale even when the underlying service is technically available to everyone.
This creates a difference between using AI as a purchased input and building an AI-enabled production system. A business can subscribe to an external service without owning the model or compute, but it still needs to adapt its workflow and verify the result. A country can import AI services without hosting frontier firms, but it may then capture less of the corporate rent and have less influence over standards and supply. Importing capability can raise productivity; it does not guarantee that the institutions and incomes around that productivity will be built locally.
Scale economies are central to frontier AI.