AI Computing Capacity Set to Multiply as Data Centers Expand Globally
The number of AI chips in data centers is projected to double every nine months, reaching around 200 million by 2028, as companies in the United States and China race to build new facilities and increase computing power.
The global stock of artificial-intelligence chips used in data centers is expected to double roughly every nine months, according to research firm Epoch AI. With about 20 million chips in operation today, the estimate is for roughly 200 million chips by the end of 2028, a ten-fold increase from current levels.
Industry leaders say the expansion will be driven by new data-center projects across the United States, Europe and the Middle East, as well as by rising chip production. Amazon’s head of foundational AI models, Peter DeSantis, noted that the company has doubled its computing capacity since 2022 and plans another doubling by next year. > “It’s hard to get your mind around the scale,” he said.
Rob Wachen, co-founder of microchip firm Etched, described the build-out as the largest infrastructure effort in human history. > “This is the largest scale infrastructure build-out in the history of humanity,” he said.
The surge reflects belief in the so-called “scaling laws,” which hold that larger data sets and more compute enable AI systems to tackle increasingly complex tasks. Executives from firms such as Anthropic and DeepMind have suggested that continued scaling could allow AI to perform substantial portions of white-collar work and drive industrial advances at unprecedented speed.
The expansion also raises concerns. Environmental groups have protested data-center construction over potential impacts on electricity use, water consumption and local ecosystems. Economists warn that rapid investment could outpace profitability, citing past technology booms that ended in market corrections. Nobel-winning economist Philippe Aghion warned that “each time you’ve had a technological revolution, this kind of bubble bursting happened.”
Geopolitically, the United States currently hosts about 5,500 data centers and controls roughly 80 percent of global AI-related computing power, according to Epoch AI. Google is estimated to have four times as many AI chips as all Chinese companies combined, though China is accelerating its own chip production and data-center projects. Chinese officials have framed AI development as a collaborative international effort, while U.S. analysts expect the American lead to persist for the next several years.
Investment in AI infrastructure is projected to exceed $1 trillion annually by 2029, up from $318 billion in the previous year, according to IDC. Jeff Dean, Google’s chief scientist, said larger scale enables capabilities that were not possible at smaller scales and that the goal is to make those capabilities available to billions of users.
The growing compute capacity is expected to increase the use of AI “agents” for tasks such as coding, research, and customer support across corporations like Uber, PepsiCo and Walmart. Economists caution that while AI could create new jobs, it may also displace many existing roles, leading to significant labor-market adjustments.
Overall, the rapid build-out of AI-focused data centers and chip production is reshaping the technology landscape, prompting both optimism about new capabilities and scrutiny over economic, environmental and geopolitical implications.