AI Bubble Debate: Both Sides, Steelmanned

Illustration: chili pepper mascot illustrating AI money pit

The Great AI Money Pit: Bubble or Boom?

Welcome to the AI money pit.

The air is thick with the scent of freshly minted billions and the faint whiff of panic. Everyone’s got an opinion on AI, and they’re usually shouting it from opposite ends of a very expensive conference room. Is this the dawn of a new era, or just a particularly shiny, silicon-infused bubble waiting to pop? Both sides are selling a story, polished to a blinding sheen. The real test, however, isn't who sounds smarter, but whether revenue decides to catch up with spending before patience packs its bags and leaves.

The bull case, steelmanned

Let's be clear: this isn't just a bunch of nerds playing with algorithms. The smart money, the really smart money, is pouring in like it's going out of style. Tech earnings per share are up a staggering 55.7 percent year to date, with tech prices gaining 22.1 percent. And get this: the tech price-earnings multiple actually dropped 21.6 percent, according to an a16z summary from September 2026 https://thebusinessperspective.com/ai-bubble-bursting/. This isn't just price speculation; it's growth. Real, honest-to-goodness growth.

Venture capitalists, those notoriously cautious creatures, are practically throwing money at anything with "AI" in its pitch deck. AI grabbed 86 percent of US venture dollars in one summary, and globally, it was 61 percent of all venture capital in 2025, totaling a cool 258.7 billion dollars, per the OECD https://thebusinessperspective.com/ai-bubble-bursting/ and https://intuitionlabs.ai/articles/ai-bubble-vs-dot-com-comparison. That's not a trend; that's a stampede.

Look at the titans. OpenAI, the poster child for AI ambition, secured 110 billion dollars in February 2026 at a 730 billion pre-money valuation, closing March 31 at 122 billion dollars and 852 billion post-money. Amazon chipped in 50 billion, Nvidia 30 billion, and SoftBank another 30 billion https://intuitionlabs.ai/articles/ai-bubble-vs-dot-com-comparison. These aren't small bets from small players. These are corporate behemoths betting the farm.

Anthropic, another major player, saw its valuation jump from 30 billion at 380 billion in February 2026 to 65 billion at 965 billion post-money by May 2026 https://intuitionlabs.ai/articles/ai-bubble-vs-dot-com-comparison. These numbers aren't just big; they're astronomical, reflecting a belief that AI isn't just a feature, but the entire operating system of the future economy.

The market agrees. The Nasdaq, a benchmark for tech enthusiasm, hit about 27366 on October 9, 2026, while the S&P 500 hovered around 7812. Nvidia, the chip king powering this revolution, was valued at about 5.7 trillion dollars in early October 2026 https://intuitionlabs.ai/articles/ai-bubble-vs-dot-com-comparison. This isn't a speculative frenzy; it's the market recognizing a fundamental shift in how everything gets done. The infrastructure is being built, the applications are emerging, and the world is changing. Fast.

The bear case, steelmanned

Alright, let's hose down the champagne and look at the cold, hard numbers. Yes, money is flying, but where's it going, and what's coming back? OpenAI's revenue run rate is described as about 40 billion annualized by 2026. Sounds great, right? Until you hear their 2026 GAAP losses are estimated near 62 billion, versus a more palatable 14 billion non-GAAP figure cited in the press https://tbsgraduates.net/writeup/ais-trillion-dollar-bubble/. When OpenAI's GAAP losses are estimated near $62 billion in 2026, while their non-GAAP figure is cited at $14 billion, it makes you wonder if their accounting department is powered by a particularly optimistic large language model.

Anthropic isn't much different, with a revenue run rate described as 47 billion, having raised about 132 billion dollars since its founding in 2021 https://tbsgraduates.net/writeup/ais-trillion-dollar-bubble/. That's a lot of capital for a "run rate" that still leaves a gaping hole.

The capital expenditure (capex) for AI is estimated at about 450 billion dollars in 2026, chasing a mere 80 to 120 billion in direct AI revenue. That's roughly 25 cents of revenue for every dollar spent https://www.linkedin.com/posts/goldenbridgeinvestments_investing-ai-marketvaluation-activity-7516022589836689408-U-1m. If that's not a red flag, it's at least a very aggressive shade of orange.

And let's talk about valuations. The Shiller CAPE, a long-term valuation measure, hit 41.1 in October 2026. Compare that to 44.2 in December 1999, right before the dot-com bust, and a long-run average of about 17 https://www.linkedin.com/posts/goldenbridgeinvestments_investing-ai-marketvaluation-activity-7516022589836689408-U-1m.

The verdict neither side wants

Here is the part both camps skip when they are winning an argument online. The bulls are right that real money is moving, real earnings are up, and real users are showing up every week. The bears are right that the bill is enormous, the revenue per dollar spent is thin, and the valuation cushion is almost gone. Put those two sentences together and you get the actual position: this is an expensive bet on a real thing, not a fraud and not a sure thing.

A press release would never print this: nobody in this debate knows whether 25 cents of revenue per dollar spent turns into a dollar next year or stays at 25 cents, and the people who pretend they know are usually selling something. The honest test is boring. Watch whether run rates keep climbing, whether the 5 percent who pay become 10 percent, and whether the giants can cut a deal that does not need another 100 billion dollar round to stay alive. If those three things move, the bulls win the argument without needing to win the shouting match. If they stall, the bears do not need a crash to be right, they just need patience.

Sources and further reading

AI bubble bursting, earnings and multiples AI bubble versus dot-com comparison AI bubble numbers, OpenAI versus Anthropic AI trillion-dollar bubble writeup * AI valuation rhymes with 2000