AI will not cause 20% unemployment and 10% GDP growth in the next couple of years

Who does Anthropic's CEO expect to buy all the things he envisions AI being able to produce/provide with 20% unemployment?

On May 28th, Axios ran a piece with excerpts from an interview with Anthropic’s CEO. It was your pretty standard tech leadership mixture of actual information mixed with marketing disguised as wisdom. However, one thing that he said stuck with me simply because of how perfectly it highlighted the limited understanding some of these leaders have of things outside their perview.

The comment is this: "Cancer is cured, the economy grows at 10% a year, the budget is balanced — and 20% of people don't have jobs."

I’m not going to try to argue with the idea that AI could or should cause mass unemployment, there are plenty of other actual experts who can do that. What I do take issue with here though is the ignorance of how GDP growth works and why we will not have 10% GDP growth with 20% unemployment. Mathematically, that simply won’t happen. It’s possible you could see vast productivity gains with 20% unemployment, but that still would not result in 10% GDP growth concurrent with Depression-style joblessness. That’s because the math for GDP is surprisingly simple and problematic for that forecast.

GDP at the national level is a daunting task to complicate primarily because you have to accurately count all economic activity. However, once the counting and labeling is accomplished, the actual formula itself is very simple.

Gross Domestic Product (GDP) = Consumption + Investment + Government Spending + Net Trade (Exports - Imports)

Consumption is the vast majority of US GDP. In the most recent release from the BEA, GDP for the US for Q1 of 2025 is calculated below (dollars billion)

Consumption accounts for over two thirds of the calculation. Consumption, unsurprisingly, is driven by consumer spending and is, therefore, extremely sensitive to employment. If there are rapid and extreme shocks to employment levels, consumption will react accordingly.

In order for AI to generate 10% GDP growth with rising unemployment, you would need to see massive increases in spending in Investment or Government (or the reversal of our 40+ year old trade deficits). Let’s generously assume that Consumption manages to remain constant with huge numbers of people becoming unemployed. 10% growth would mean GDP growing to almost $33 trillion. If Consumption holds steady at $20,493.3 then I+G+ Net Trade somehow needs to see a net increase of 31.6%.

To give you a sense of how ridiculous that is, the average rate of growth for Investment spending, the category most likely to grow in an AI boom, over the past 7 years is 2%.

The numbers get much worse very quickly if you expect Consumption to decline if there is mass unemployment. In the Great Recession (2007-2009), consumption declined by 5% from it’s 2007 level and remained below 2007 totals for several years. That was in a world with just 10% unemployment, half of our supposed total.

Back to the math. If Consumption declines by 5%, then the other categories need to increase by more than 42%. It’s just unrealistic. Mass unemployment will lead to at best stagnant GDP and, more likely, a recession or worse.

If we face a future where AI renders white collar work unnecessary and the US experiences crushing unemployment, we will not experience a wave of GDP growth, we will experience a 19th-century style bust where GDP stagnates in the near-term with the possibility that rising efficiency and reshuffling of workforce roles results in growth at some point several years down the road.