AI Chip Numbers Double Every Nine Months in Record Boom

The number of artificial intelligence chips in data centers worldwide has reached approximately 20 million H100-equivalent units and is now doubling every nine months, according to a new investigation by The New York Times. Projections indicate the total will reach 200 million by the end of 2028 — a tenfold increase in just over two years — signaling what technologists describe as the largest infrastructure build-out in human history.

The Scale of the Build-Out

This explosive growth in computational capacity, driven by the conviction that more computing power creates more capable AI, has triggered an unprecedented wave of data center construction across the United States and around the world. US companies including Amazon, Google, Microsoft and Meta now control approximately 80 percent of global AI computing power, according to Epoch AI, and are projected to spend roughly $750 billion this year on data centers, chips and related infrastructure, up from approximately $400 billion in 2025, according to Goldman Sachs.

The scale of the build-out has drawn comparisons to the railroad boom of the 1800s, Franklin D. Roosevelt’s New Deal, and the Manhattan Project. “This is the largest scale infrastructure build-out in the history of humanity,” said Rob Wachen, co-founder of the microchip firm Etched, which has raised more than $1 billion to meet demand for AI components.

Peter DeSantis, who leads foundational AI models at Amazon, told the Times that the company has doubled its computing capacity since 2022 and plans to double it again by next year. “It’s hard to get your mind around the scale,” he said.

The Driving Philosophy

Behind the surge is a core belief known as “the Scaling Laws” — the tenet that AI systems become dramatically more capable with more data and computing power. This philosophy has become the driving force of the current technological era, with tech leaders arguing that those who command the most computing power will create the most advanced systems.

Confidence in the Scaling Laws has led to increasingly bold predictions. Dario Amodei, CEO of Anthropic, has said that if the laws hold for another year or two, AI will be able to perform vast amounts of white-collar work. Demis Hassabis, head of Google DeepMind, wrote that AI could usher in “10x of the Industrial Revolution at 10x the speed.”

Jeff Dean, Google’s chief scientist who has worked in AI research for more than 30 years, explained that new capabilities emerge at larger scale that simply do not occur at smaller scale. “You’re also now trying to bring these capabilities to not just a few million users but to hundreds of millions or billions of users,” he said.

The Infrastructure Race

The physical infrastructure to support this vision is taking shape at remarkable speed. In March 2024, there were approximately 2.4 million H100-equivalent chips worldwide. Today there are 20 million. By the end of 2028, Epoch AI projects roughly 200 million.

In Wisconsin, Microsoft opened the Fairwater data center on June 23 — described as the world’s most powerful supercomputer — with nearly 550 full-time employees on site. A second facility is already under construction, scheduled for completion in 2028, which will be more than 100 times as powerful as the machines that trained the early version of ChatGPT. In Indiana, Amazon’s Rainier complex spreads seven interconnected data centers across what was once cornfields, serving Anthropic’s AI workloads.

Meanwhile, China is racing to close the gap. Chinese companies had approximately 1.16 million H100-equivalent chips at the end of 2025, up from roughly 244,000 at the beginning of 2024. The Chinese Communist Party’s five-year economic strategy, released in March 2026, mentions AI more than 50 times and calls for building “next generation supercomputing.” Xi Jinping, in a speech in Shanghai this month, declared that “AI development should not be a solo performance by a single country, but a symphony of international cooperation.”

However, analysts expect the US lead over China to grow for another four to five years before China’s domestic chip production reaches sufficient scale. Jordan Nanos of SemiAnalysis said the US data center lead would likely grow for four to five years before China begins closing the gap. “The advantage will run out,” he said.

The US-China rivalry threatens to leave much of the world behind. Europe currently holds just 5 percent of global AI computing power, constrained by electricity access, land availability, permitting and financing. Amandeep Gill, the UN under secretary general and special envoy on technology, warned: “If 75 percent of the compute today is in a few postal codes in the US, 12 to 15 percent in China, and 5 percent in the EU, where does it leave the rest of the world?”

Growing Pains: Energy, Backlash, and Bubble Concerns

The breakneck pace of expansion is raising alarms on multiple fronts. Global data centers consumed 64 gigawatts of electricity last year — roughly equivalent to Germany’s total usage, according to SemiAnalysis. By 2030, that figure is projected to quadruple, surpassing the combined electricity consumption of all of South America and Africa. Every gigawatt of power for advanced AI data centers translates to approximately $40 billion to $60 billion in total costs.

Community opposition is mounting. According to IDC, global AI infrastructure investment is forecast to top $1 trillion by 2029, comparable to the GDP of Switzerland. But approximately $130 billion worth of data center projects have been blocked by protests in 2026 alone. Data centers are shaping up as a major issue in the November 2026 US midterm elections, with a federal AI Data Center Moratorium Act introduced in March and state-level moratoriums proposed or passed in Maine and Oklahoma.

Economists are also warning of a potential bubble. Philippe Aghion, who won the Nobel Prize in Economic Science in 2025, drew parallels to past technology cycles: “Each time you’ve had a technological revolution, this kind of bubble bursting happened. AI is like the fourth industrial revolution and it has this aspect to it that generates a bubble.” Past infrastructure booms — railroads in the 1800s, electrification in the 1920s, the dot-com boom of the 1990s — were followed by recessions and crashes as overinvestment outpaced realizable benefits.

Labor and the Accelerating AI Loop

At the same time, AI’s real-world capabilities are advancing rapidly. According to the Remote Labor Index, developed by the Center for AI Safety, the share of freelance tasks that AI agents can complete at a professional level has risen from 2.5 percent in October 2025 to 15.8 percent by July 2026 — more than a sixfold increase in under a year. AI systems are now capable of building mobile video games, designing 3D models, producing animated advertisements, and generating architectural floor plans.

Carl Benedikt Frey, an economist at Oxford University, characterized the competitive dynamic as an arms race: “These companies are essentially in an AI arms race. If they don’t invest, they are acknowledging defeat.”

Erik Brynjolfsson, director of Stanford’s Digital Economy Lab, warned of profound labor market disruption: “There’s going to be millions of jobs destroyed, millions of jobs created. That’s going to be very difficult. Even if new jobs are created, they’re not the same jobs.”

Perhaps most strikingly, leading AI labs are pursuing a breakthrough known as recursive self-improvement, where AI systems help build their own successors. Google’s Jeff Dean described a future where thousands of “very tiny models” generate ideas autonomously. “With more computing power coming, you can fully automate the loop,” he said. “We are at the beginning stages.”

What to Watch For

As the number of AI chips continues its relentless climb — doubling every nine months, with no end in sight — the world faces a series of profound questions. Can energy grids keep pace with data center demand? Will the benefits of AI be broadly shared or concentrated in a handful of tech giants and nations? How will societies manage the labor disruption that economists increasingly warn is coming? And what happens when AI systems begin improving themselves?

The answers will shape not just the technology industry but the global economy, geopolitical balance, and the daily lives of billions. For now, the only certainty is that the deluge of computing power is accelerating — and the world is only beginning to grapple with its consequences.