Every day brings new predictions that the so-called “AI buildout” is unsustainable, a bubble, or a crisis waiting to happen. The argument is that the US economy is being held hostage by AI spending, invoking the usual metaphors of a house of cards or a Ponzi scheme that will come crashing down.
A few examples from 2026:
- The biggest gamble in the U.S. economy is starting to look riskier — The Washington Post, July 31.
- AI’s Wobbly House of Cards Puts Markets and US Economy at Risk — Bloomberg, September 20.
- AI slowdown calls justified but collapse of bubble may be more immediate threat — The Guardian, September 20.
- Financing of historic AI buildout raises systemic risks in US, researcher says — Reuters, September 24.
- The Coming AI Meltdown — The American Prospect, September 25.
- The debt-fueled AI boom is cracking — Axios, October 16.
However, these predictions of unsustainability are hardly new. They go back to at least 2023:
- AI Is the Bubble to Burst Them All — WIRED, October 27, 2025.
- The AI Bubble and the U.S. Economy: How Long Do ‘Hallucinations’ Last? — Institute for New Economic Thinking, September 2025.
- Big Tech is spending billions on AI. Some on Wall Street see a bubble. — The Washington Post, July 24, 2024.
- Gen AI: too much spend, too little benefit? — Goldman Sachs, June 27, 2024.
- From boom to burst, the AI bubble is only heading in one direction — The Guardian, April 13, 2024.
- What kind of bubble is AI? — Cory Doctorow, December 19, 2023.
These people have been sounding an alarm for a long time, but where is the fire? Maybe they need to give it a rest or, at the very least, admit they are wrong or reassess their premises. Being wrong all the time is no reason to dig in. It means it’s time to reevaluate your arguments.
I am making a bold call: the AI buildout is sustainable. It will not collapse as so many are predicting.
The reasoning is two-fold. First, the people who keep making this prediction have terrible track records, as mentioned above. A good heuristic is that these people tend to be wrong more often than not, so doing the opposite of them can be a profitable investing strategy.
Second, scenarios like 2008 or 2000, where asset prices collapse, are outnumbered by instances where valuations keep going up and up, and the awaited crisis never comes.
So without even working out the math of debt or AI spending, this heuristic alone is useful. Recall the similar predictions about Meta/Facebook, Tesla, Uber, Amazon, and many others being bubbles or headed for collapse:
- Analyst Predicts Facebook Will Disappear Soon — CBS News, June 6, 2012.
- Facebook will lose 80% of users by 2017, say Princeton researchers — The Guardian, January 22, 2014.
- Tesla is just months from a total collapse, says hedge-fund manager — MarketWatch, March 27, 2018.
- Bob Lutz Says Tesla ‘Headed For The Graveyard’ — GM Authority, September 20, 2018.
- End of the line for Uber — Cory Doctorow, August 10, 2021.
- Analysts Send Amazon Down River — WIRED, June 23, 2000.
For every one they get right, like Theranos, there are others that defy predictions of their collapse. Theranos is often used as a case study, but its collapse was such a national sensation that it led to a television series, a book, and multiple documentaries.
People make documentaries about things that are uncommon or special, not routine. No one makes documentaries about how successful Uber was, which successfully pivoted to delivery during Covid. Or how Facebook orchestrated a remarkable turnaround in 2012–2014, pivoting to mobile and beginning to monetize Instagram, which was once considered a reckless purchase for $1 billion by those very “experts” (now estimated to be worth hundreds of billions of dollars on its own).
This suggests that VCs and other “smart money” are actually good at knowing where to put their capital. They are smarter than these pundits who keep insisting it’s a bubble or a crisis in the making.
There is also the goliath of coding. Recall, when generative AI entered the public consciousness around 2023, the main applications were writing, short-video generation, and images, as those were the most obvious. Remember AI making artists obsolete? This was a huge concern in 2023…not so much anymore.
Then in 2025, Anthropic and OpenAI launched coding agents, and everything changed. This opened up a potential trillion-dollar industry as major companies began adopting AI coding to save money, vastly expanding the market beyond “the arts.” This led to the so-called “SaaS Apocalypse” over fears of companies switching to AI. By February 5, 2026, roughly $1 trillion in market capitalization had been erased from the S&P 500 software and services index since January 28.
Grand View Research projects that the global AI code tools market will reach $26 billion in annual revenue by 2030. Mordor Intelligence puts it at nearly $39 billion by 2032, while Precedence Research projects $91 billion by 2035. Individual companies are already targeting billions: April reporting put Cursor’s projected annualized revenue above $6 billion by the end of 2026, while September reporting projected $4–5 billion for Cognition on the same basis. (These are projections for the tools and related services alone, separate from the economic value of the software being produced.)
According to a16z (yes, not an unbiased source):
Consider this: there are approximately 30 million software developers worldwide, with estimates ranging from 27 million by Evans Data to 47 million by SlashData. If we assume each developer generates $100,000 per year in economic value—perhaps conservative for the U.S. but slightly high globally—then the total economic contribution of AI software development is a total of $3 trillion per year.
Based on dozens of conversations over the past 12 months with enterprises and software companies we estimate that a simple AI coding assistant today increases productivity of a developer by about 20%.
When so many companies are suddenly using AI, suddenly these valuations do not seem as insane. In 2022-2024 much of the focus was on art. Now the discussion extends to AI making mathematicians or programmers obsolete.
A obvious counterexample are those same VC firms promoting cryptocurrency/blockchain companies in 2021-2022, which fizzled out. But there are key differences between the short-lived early 2020s crypto boom and the post-2022 “AI boom”. Crypto never generated anything remotely close to the sales figures or user adoption of today’s leading AI companies. Crypto was never taken that seriously even at its peak, as the adoption never really came to fruition, whereas with AI, we’re seeing many companies integrate AI, such as code. By comparison, the “reorganization of the financial system around blockchain” amounted to little more than wishful thinking.
Overall, the “market consensus” has decided that the buildout is sustainable, or at least so far. I agree. What seems like irrational exuberance today may in hindsight seem perfectly rational, much like how the World Wide Web became mainstream and immensely profitable (e.g. Alphabet and Meta) despite valuations being overextended during the 90s and early-to-mid 2000s. Of course, I could be wrong. Maybe this is like Enron all over again, to bring up another favorite poster child of corporate hubris. But if history is any guide, in 2-3 years from now we’ll still be waiting for the crisis that won’t come.