science & techOur World in Data
Most widely used AI models come from American and Chinese companies
Dozens of companies around the world build large AI models, but a small number of them account for most of the systems people actually use. Those flagship models come mostly from US labs and Chinese developers. Training a frontier model takes huge amounts of computing power, specialist talent and data, which keeps the field concentrated. That concentration shapes which languages, values and default answers reach billions of users.
Large AI models are trained on enormous datasets using vast clusters of specialised chips, a process that costs huge sums. Only a few organisations can afford that scale, so the most capable systems tend to come from a handful of well-funded labs.
Whoever builds the most-used models sets the defaults for search, writing and software tools. Knowing where these models come from helps explain the geopolitics of AI and why countries worry about depending on foreign systems.
