✨ How the AI Boom is like the Internet Boom (and how it isn't)
Stock prices are one thing, big economic impacts another
My fellow pro-growth/progress/abundance Up Wingers,
First things first. There’s going to be a lot happening for the rest of 2025 into 2026 — the march toward superintelligence, a nuclear energy renaissance, and a New Space Age. And Washington is going to play a big role in all of it.
I look forward to taking this journey with all my readers as I analyze the economic and societal impacts of this transformation.
So before we get started, I wanted to let you know about a 50% off sale on new subscriptions. I don’t do many sales, so I’m not sure when I will make this offer again.
AND THIS ONE ENDS TODAY!
Please click on the blue button below and get 50% off for the next year! And let’s go!
History is rhyming again in Silicon Valley — or at least sounding a clear echo. Just as Cisco Systems and Intel rode the internet wave to stratospheric heights in the go-go 1990s, Nvidia and Microsoft are surfing AI euphoria to trillion-dollar market caps.
But the dot-com crash teaches a harsh lesson worth considering: When revolutionary technology fails to deliver the economic outcomes built into investor expectations, even the most promising boom can turn to bust.
AI for automation, only
Granted, current AI models seem evermore clever at automating existing tasks, which is a valuable function, no doubt. But they still need to prove they have enough silicon smarts to deliver the kind of fresh breakthroughs that can truly justify soaring equity valuations. Microsoft Chief Executive Satya Nadella's vision of autonomous AI agents sounds transformative: software that plans, executes, and learns across every system. But this sort of use case may fail to provide anything like the productivity boost that the most enthusiastic AI advocates dream about.
Oxford University economist Carl Benedikt Frey cites the Nadella example in a new FT commentary, noting that despite decades of faster computers, labor productivity growth in advanced economies has slowed from two percent annually in the 1990s to just 0.8 percent in the past decade. (China's once-soaring output per worker has stalled, too, by the way.)
From the piece:
A survey of over 7,000 knowledge workers found heavy users of generative AI reduced weekly email tasks by 3.6 hours (31 per cent), while collaborative work remained unchanged. But once everyone delegates email responses to ChatGPT, inbox volume may expand, nullifying initial efficiency gains. America’s brief productivity resurgence of the 1990s teaches us that gains from new tools, be they spreadsheets or AI agents, fade unless accompanied by breakthrough innovations.
Delayed impact
Georgetown University economist Dan Cao's new research paper, “Tech Booms: From Dot-Com To Dot-AI,” offers a sobering parallel in how the expectations game played out decades ago. His model of the 1990s dot-com bubble shows that market valuations didn’t peak when, say, investors suddenly realized that the internet was no more valuable than a fax machine.
Rather: when the broad economic benefits from the emerging Internet Revolution were slow to materialize. (People had the idea for a ride-sharing service a decade before Uber was founded in the 2000s.) The key factor was "spillover speed" — how quickly productivity gains from high-tech sectors spread to the rest of the economy. When that diffusion disappointed, markets corrected regardless of the underlying technology's merit.
Cao: “In the current Dot-AI boom, amid the real and rapid progress in AI, AI related stocks have been experiencing significant appreciation and have remained at elevated levels. If the model is also applicable to the current Dot-AI boom, it suggests that the rapid advancements in AI are expected to continue for the foreseeable future.”
Indeed, it does seem like the AI boom is following a similar pattern: explosive growth at companies like Nvidia, massive capital investment, and soaring valuations betting on economy-wide transformation. One difference is that today's tech giants have stronger balance sheets. Another: The expectation for accelerated productivity and economic growth is even higher for the AI Revolution than the Internet Revolution.
AI for ideation
What’s more, as Frey notes, those expectations hinge on AI helping find new things to do, not just automating existing tasks. Yet AI today optimizes for efficiency within existing paradigms rather than discovering new ones. While markets price in transformative productivity gains, current AI primarily delivers incremental improvements.
Again, from the Frey piece:
Had the 19th century focused solely on better looms and ploughs, we would enjoy cheap cloth and abundant grain — but there would be no antibiotics, jet engines or rockets. Economic miracles stem from discovery, not repeating tasks at greater speed. … AI could still ignite a productivity renaissance — but only if we use it to dig deeper for new and previously inconceivable endeavours rather than merely drilling more holes.
As it happens, a new McKinsey report on AI seems to provide evidence for the point about incrementalism that Frey makes, according to Ethan Mollick:
Now, if human-level AI is a sooner rather than later thing, maybe the problem Frey identifies is one of fleeting importance. But it might not be. Google DeepMind's Demis Hassabis, creator of the celebrated protein-folding model AlphaFold, admits that artificial general intelligence may require "several more innovations." (Again, this reminds of the need for higher bandwidth and smartphones for many 1990s business ideas to be realized.) And without institutional shifts — in funding, incentives, and culture — AI risks becoming just a turbocharged email assistant rather than the catalyst for the next medical wonder, energy breakthrough, or transportation advance.
Even if Frey is correct (and I would guess that there is so much money flowing into the sector that plenty of riskier ideas can be funded, at least in the US, and execs will become bolder as their familiarity with generative AI increases ), a so-so productivity performance from the AI Revolution might still deliver a half-point increase in productivity growth. That’s estimate I get from the more cautious economists who follow AI closely.
The most bullish case for growth, then, combines the need for diffusion and discovery. Faster, please!
Micro Reads
Please check out the website or Substack app for the latest Up Wing economic, business, and tech news contained in this new edition of the newsletter. Lots of great stuff!