How countries can end the capability overhang

OpenAI dropped a report this week with a term that's likely to stick around: the capability overhang. The idea is simple enough. Frontier AI models can already do a lot for productivity, healthcare, education, you name it. But most countries aren't using anywhere close to that capacity. The gap between what AI can do and what nations are actually doing with it, OpenAI argues, is wide and getting wider.

The report leans on adoption data across dozens of countries and finds the usual suspects, the US, parts of Western Europe, some Gulf states, out ahead, while much of the rest of the world lags well behind. That's not exactly shocking. What is worth pausing on is the framing: OpenAI isn't just describing a digital divide, it's describing a divide in access to a technology that compounds. A country slow to adopt AI in 2025 isn't just behind, it risks staying behind as models get more capable and more embedded in everything from logistics to diagnostics.

To its credit, OpenAI didn't just publish the report and walk away. It's rolling out new initiatives aimed at governments, apparently meant to help them build the infrastructure, skills, and policy frameworks needed to actually capture productivity gains rather than just watch other countries do it. Specifics are thin in what's public so far, but the direction is clear: OpenAI wants to be the company that helps nations climb the adoption curve, not just the one selling the models at the top of it.

There's an obvious self-interest angle here too. Every government OpenAI helps onboard is a government more dependent on OpenAI's stack, more likely to write policy favorable to OpenAI's business, and less likely to end up building sovereign alternatives. None of that makes the underlying observation wrong. Capability overhangs are real and they do have consequences. It just means the company diagnosing the problem also has a lot to gain from being the one who fixes it.