Dan Denning: So Much Depends on a Green Wheelbarrow

“The world has yet to turn any of today’s AI hype and spending into a meaningful lift in the actual economy.”

— Satya Nadella

 

December 4, 2025 — It turns out that you CAN determine the total factor productivity (TFP) of an egg-laying chicken. But it depends on which type of chicken (a commercial Leghorn hen is ‘high efficiency’ while a backyard breeder hen is ‘low efficiency.’)

A green wheelbarrow, on the other hand, has ‘zero’ TFP because technically, it produces nothing. As a capital good, however, it is a ‘productivity multiplier.’ And that may make all the difference.

Or not!

At stake is trillions of dollars in AI investment and tens of trillions of stock market value. Whether it pays off or not–in real economic terms–is the central topic of today’s letter.

History shows that the great debt and money crisis of the past precede radical political revolutions and violence. The only ‘Hail Mary’ that can save us is a radical increase in productivity from a new technology, something on par with the railroads, electricity, or the personal computer. Which brings me back to the Green Wheelbarrow.

Wherever I’m trying to figure something out that is full of technical jargon, I go back to what Professor Zapatka said in my intro to Russian Literature class: the first step in characterization is appellation. What you call things–their name and definition–matters (remember Adam spends a lot of Chapter Two in genesis naming things).

What I’m trying to figure out today is if AI is a General Purpose Technology (GPT) on the same scale as electricity and the Internet. And IF that’s the case, what the Total Factor Productivity of AI would have to be to justify the trillions invested in chips, data centers, and the power both require to function. What would make it all worth it, in economic terms?

Those are big questions. It is beyond the scope of a weekly research note to answer them all today. But we can make a start. The first step is defining what a General Purpose Technology (GPT) is. After that, a brief note about chickens and wheelbarrows.

The first part is actually not that complicated, once you strip away the big words. The term General Purpose Technology was first coined in a paper published in 1995 in the Journal of Econometrics (General Purpose Technologies: Engines of Growth?). Some technologies, the paper argues, are more ‘transformative’ than others, and create new economic growth. Why? Three reasons.

1. They’re pervasive. They are used everywhere in a wide range of applications and activities. Think of the steam engine, or later, electricity. Or semiconductors and the internet.
2. They continuously improve. They get better, faster, and cheaper–resulting in big efficiency and productivity gains over time. Think of Moore’s Law.
3. Innovational Complentarities. A fancy way of saying that you see increased productivity from research and development of new products and services downstream of the GPT. The technology itself creates and disrupts but mostly in a wealth-creating way (think of AI’s designing drugs and therapies based on your DNA).

There’s a lot going on there. But the basic question is whether AI can reasonably be compared to electricity and the Internet. And if it can, how much investment will it take to ‘take off’ and how long will it take before we get there (wherever ‘there’ is)?

Again, all big questions. I don’t propose to answer them all here. But one big headwind for AI: any benefits from its adoption as a GPT are offset by its deflationary impact on the value of human labor. It had better make the world safer, richer, and more abundant. Or there are going to be a lot of out work people with lots of debt who are unhappy, desperate, and primed to be politically radicalized (even more than they already are).

What do you think?

Are AIs really GPTs? Are they about to become pervasive and continuously improving and lead to breakthroughs in energy, materials design, and medicine? Do all those changes justify the trillions in capital investment in ‘more compute’ based on the future growth and prosperity that will inevitably result?

I use the word ‘inevitably’ with qualification. We just don’t know. But in making our assessment, we do get some help from chickens and wheelbarrows.

Electricity automated some physical work. Computers automated some routine cognitive work (calculations, record keeping etc). AI, so we’re told, will automate a lot of NON-routine cognitive work, including invention and scientific discovery. As a result, the Total Factor Productivity of AI will put it on par with other transformative technologies (so we’re told).

The idea has its roots in a paper published in 1957 by Robert Solow called Technical Change and the Aggregate Production Function. Solow was trying to explain why output per person (productivity) doubled in the United States between 1909 and 1949. This increase in productivity wasn’t just because of better tools and better educated workers.

There was some ‘X’ factor working, according to Solow. He concluded it wasn’t just more machines and more workers that increased productivity. It was ideas and innovation that accounted for more than 90% of the increase in productivity. In his own words, “Economic growth is mostly about figuring out better ways to do things, not just piling up more machines and more workers.”

Solow invented the term Total Factor Productivity to describe this dynamic. Ideas, technology, processes, people, knowledge…they all combined, ideally…to use resources efficiently, increase productivity…and produce more growth/wealth.

That’s the theory anyway. Paul Romer, an economist at the New York University Stern School of Business, would add to the theory and win a Nobel Prize for it in 2018. His argument was that growth doesn’t come from a mysterious process but a deliberate and measurable investment in technology and people (what he called ‘endogenous growth theory) that leads to new discovery.

But just because you invest heavily in something doesn’t mean you get what you want. Trillions in capex in AI may not result in a big increase in Total Factor Productivity. And AI may turn out to not even be a GPT.

Yann LeCun, who used to run AI at Meta and has been referred to as ‘The Godfather of AI’, said what we’ve been saying for awhile now, that AI investment may be a large misallocation of capital and human resources. LeCun said that Large Language Models (LLMs):

‘Are not a path to human-level intelligence. They’re just not. Right now, they are sucking the air out of the room anywhere they go — and so there’s basically no resources [left] for anything else. And so for the next revolution, we need to take a step back and figure out what’s missing from the current approaches.’

Chickens convert inputs to outputs. Some do it better than others. The genetic ‘technology’ of a commercial Leghorn hen allows her to produce 300-320 eggs a year. That’s what we can now call ‘high-TFP.’

Wheelbarrows are not chickens. A chicken is a biological production unit. A wheelbarrow is a capital good. A wheelbarrow doesn’t produce work. But it CAN be a productivity multiplier.

And that’s how we have to think of all those GPUs the hyperscalers are spending money on. If their thesis is right, trillion in AI and data center spending now, will translate into a massive burst in productivity and new technologies in the next two decades. That is the only justification for the current valuations/multiples at which these stocks trade now.

The American poet William Carlos Williams wrote, “So much depends, upon a red wheelbarrow, glazed with rainwater, beside the white chickens.”

Today the wheelbarrow is Nvidia Green. And so much of the stock market depends on that wheelbarrow being a big enough productivity multiplier to offset $340 trillion in debt.

Dan Denning
Bonner Private Research & Grey Swan Investment Fraternity

P.S. from Addison: Dan Denning of Bonner Private Research — editor, investor, and as godfather to my middle son, contractually obligated to keep me honest – joined us on today’s Grey Swan Live!

Dan and I covered multiple topics, from the Fed’s pivot from tightening to easing, the rise of Dollar 2.0, and what it all means for your personal balance sheet before the next Enron or Lehman Bros. signals the historical start of the next crisis, spawn of Fed’s perpetual bubble machine.  The replay will be ready for members soon… stay tuned!

If you have requests for new guests you’d like to see join us for Grey Swan Live!, or have any questions for our guests, send them here.