NYTimes shows how fast AI data center chip counts are growing

In short: A New York Times interactive highlights the huge and fast-growing physical infrastructure behind AI, from chips to new data centers.

What's going on

The New York Times published an interactive report on July 29, 2026, and later updated it on October 12. It tries to make the math behind AI computing easier to picture, since the scale can be hard to grasp.

The report says data centers around the world contain about 20 million AI chips, based on estimates from Epoch AI. AI chips are specialized computer parts that are good at doing the kinds of calculations used in modern AI. If recent growth continues, that number could roughly double every nine months and reach about 200 million by the end of 2028.

Training a top AI model can require tens of thousands of these chips working at the same time, like a large crowd all solving different parts of the same puzzle. The report says the cost of that training can reach hundreds of millions of dollars. After a model is trained, data centers also handle “inference,” which means generating answers when people type in prompts, and that ongoing work is increasingly driving demand for even more facilities.

What to watch

The report’s main message is that AI progress depends on more than software ideas. It also depends on a growing physical system of chips, buildings, electricity, and fast network connections. The New York Times describes a construction boom from the American Midwest to the Persian Gulf, so it is worth watching where new data centers are built and how quickly chip counts keep rising.

Source: NYTimes