What happens to economic growth when Moore’s Law speeds up?

For sixty years, computers have gotten cheaper and more powerful on a set schedule. It’s called Moore’s Law: a doubling of computing power roughly every two years. That steady pace powered most of modern economic growth across three technology eras: the personal computer, the Internet, and smartphones. All of it was built on computing that kept getting cheaper on schedule.

Earlier this year, my colleague Adam Egelberg asked a question I haven’t stopped thinking about: What happens to economic growth when Moore’s Law speeds up? Because it just did.

Intel CPUs drove the expansion of computing capacity for decades. Nvidia GPUs have accelerated this trend.

 

In the AI era, computing capability is doubling roughly every 10 months, not every 24 months. So does the economy that it powers grow faster? Or does computing just get cheap faster, while the economic gains stay the same size?

We think growth accelerates. And we think the evidence is already here.

Our economy ran this experiment before, and Adam also found the historical test case. In 1985, Intel exited the memory chip business and bet the company on microprocessors. This was a pure bet on Moore’s Law. 

The table below shows what happened to Intel’s revenue over the next sixteen years:

Period	Intel revenue growth	What was happening
1986–1992	~29% per year	The 386 and 486 era
1992–1997	~34% per year	Pentium and “Intel Inside”
1997–2001	~1.4% per year	The franchise matured

 

Intel’s chips got cheaper per unit of computing power every single year of that run. Revenue grew about 30% a year for 11 years anyway. The computing power Intel shipped was accelerating, and revenue accelerated right alongside it. Falling prices and rising revenue, at the same time, for over a decade.

How is that possible? It’s called Jevons’ Paradox, and it’s one of the most useful ideas in economics. When something useful gets cheaper, we don’t spend less on it. We find so many new uses for it that total spending rises. Cheaper computing didn’t shrink the computer business. It puts computers in every office and every home, and then in every pocket. We wrote about this last year when DeepSeek claimed to have made AI dramatically cheaper and the market briefly treated that as bad news for computing demand. It’s the opposite. Cheaper means more.

Now the same movie is playing again, but faster. Through the Moore’s Law era, the world’s total computing capacity grew about 58% per year. The installed base of AI computing is growing at 2 to 3 times per year. The doubling that used to take 24 months now takes about 10.

And the semiconductor industry’s finances are doing exactly what Intel’s did, at ten times the size. Global chip revenue hit $796 billion in 2025, up 26%. The industry’s own forecast for 2026 is $1.5 trillion, up 90% in a single year. This is a mature, fifty-year-old industry suddenly growing like a startup. Falling unit costs and accelerating revenue, again. Same experiment, same result, but with bigger numbers.

Semiconductor industry revenues are seeing accelerated growth, and at much larger dollar amounts.

 

As unit costs fall and capabilities rise, you can point these AI models at more and more economic activities. Tasks that were too expensive to automate last year make sense this year. All that new demand soaks up the available hardware, so utilization rates climb. That’s why three-year-old AI chips now rent for more than they did last fall, not less, even as newer chips ship in record volume. The hardware fills up and captures value. The users of the models capture value. When demand grows faster than supply, everyone in the chain does well at the same time.

Computers have been getting cheaper for our whole lives. What’s new is a more direct link to getting things done. 

Computation now literally equals intelligence. You can spend money and get intelligence, directly, at scale, and point it at all kinds of human endeavors that were never available for this before. That link did not exist five years ago. A cheaper spreadsheet is nice. Cheaper intelligence is a different thing entirely.

And we already know what intelligence costs, because businesses pay for it every day. It’s called a salary. Human cognitive labor is the high-priced version of intelligence, and it gets a little more expensive every year. Machine intelligence is the version whose price falls roughly 90% per year. 

When the unit cost of intelligence keeps falling like that, we’ll simply start applying intelligence to all kinds of things we previously wouldn’t have tried, because it was too expensive to try. That’s Jevons’ Paradox again, pointed at the largest expense category in the world economy.

Is there a limit? We don’t think there’s a meaningful one, because there is no upper bound to the value of intelligence. Nobody wants less intelligence. And people are competitive. Applying more intelligence to whatever you’re doing than your competitor is something humans are always going to pay for. Cars and planes are useful but ultimately have fairly limited applications. Intelligence has nearly boundless ones. 

This boom may not show up cleanly in GDP. Big deflations never do. GDP counts what is paid, and this whole story is about paying less and less for more and more. The iPhone is the classic example. As it swallowed the camera, the GPS unit, the music player, and the map, measured output in those categories shrank. An explosion of real capability showed in the statistics as falling prices. AI runs the same math at a far larger scale. When true efficiency arrives, GDP mostly can’t see it.

So when the productivity statistics look unimpressive next year, and skeptics declare the AI boom was hype, remember where efficiency actually shows up: in profits, and in accelerating growth at the companies that adopt it. Across the 900 largest US public companies, cash profit per employee is now above $120,000 per year. That’s nearly doubled from a decade ago, and the line is steepening. Profit per employee is productivity.

Employees increasingly have access to better tools, making each employee more profitable.

 

The chart above shows a productivity boom in action. Maybe we’ll see it in the GDP report. Maybe we won’t. But we’re very likely to continue seeing rising corporate profits per employee as each person can do more with useful, intelligent tools. 

The cost of intelligence keeps falling. The value of intelligence keeps rising. Accelerate Moore’s Law, and you accelerate the economy built on top of it.

 

Best regards,

Evan McGoff

 

Disclosure: Dock Street Asset Management, Inc. and/or our clients may own Nvidia (NVDA) and Intel (INTC). This article is not intended to be used as investment advice.