About the AI Market Bubble

Last week, I ran a poll among my followers on the topic “When will the AI bubble burst?” First, I noticed that not everyone understands that a bubble is a financial term. It refers to how overvalued company stocks are — and how much they could fall. The underlying technology itself can still be extremely useful and transformative — like the Internet 30 years ago.

Now to the poll results. Around 600 people took part, so the sample is fairly large. The largest group believes there is no bubble (38%). Only 3% think the bubble will burst within six months, 10% — within 7–12 months, 24% — in 1–2 years, 19% — in 3–5 years, and 6% — in more than 5 years. Personally, I think we’ll see the turning point in 2026. To explain why, let me share my personal experience with the ICO boom.

In the summer of 2017, we had a closed meeting of venture investors, where some shared their experience investing in ICOs. One speaker’s phrase struck me: “I invest in pre-ICOs of companies I’d never fund as a venture capitalist, just to sell them during the ICO.” That’s when I felt how speculative the whole thing had become. Another participant said it takes about five months to really figure it out and invest properly. I realized I probably didn’t have those five months — so I decided not to participate.

About four months later, my colleague got a call from his grandmother in a remote village asking what Bitcoin was and how to buy it. That was my signal that the end was near — when the idea of making money on Bitcoin had reached the masses. Combined with my sense of intense speculation, I knew a correction was inevitable. In January 2018, the market crashed. Sure, I could have traded for another 3–5 months and made a nice profit, but that’s not consistent with my principles.

Now, about the AI market bubble. Why do I think the correction will come in 2026? I can’t fully justify it, but several factors make me uneasy:

• The idea of investing in AI startups has completely captured investors’ minds. Just look at Y Combinator’s Request for Startups — it’s all AI. That could mean returns are truly exceptional, but it could also be a warning sign.

• When I talk to people implementing AI solutions for clients, they often say customers have vastly inflated expectations about AI’s capabilities and how easy it is to develop such systems.

• Several prominent figures in the AI industry have started talking about a “useful bubble” in AI — likely as public disclaimers for the future.

• From time to time, reputable brands publish reports showing that many AI projects fail to deliver positive ROI for clients (see the articles below).

From all this, I have a strong sense that expectations are overblown — but most people are still ignoring it. I think that unless there’s a sudden technological breakthrough, it will take around 6–12 months for a critical mass of people to realize there’s a bubble. After that, a major correction will follow. It might even get to the point where investors won’t open an email if the subject line mentions “AI startup.” Don’t believe me? Ask those who once launched solar panel startups — the technology was great, but the market overheated and then corrected sharply.

So the main question is: what should you do if you’re building an AI startup? I recommend taking an honest look at your business and asking yourself: are you creating real value for customers, and could you still sell your product if you removed the word “AI” from your marketing materials? I’d also suggest developing a plan for what to do if venture funding suddenly dries up. Could your company survive? Your goal is to outlast the period when less farsighted competitors die off.

Let’s see if my prediction comes true. In any case, I wish you success in growing your business — regardless of whether there’s hype or not.

References mentioned above:

S&P, October 2025: The share of organizations that abandoned AI solutions before production grew from 17% to 42%. Only 19% of companies achieved a strong positive effect, while 46% didn’t find a single case with a strong positive outcome. Overall, the share of failed AI implementations is rising.

MIT, August 2025: 95% of corporate generative AI pilot projects fail to deliver business results (only about 5% achieve significant revenue growth).