I’ve been reading with some interest the debate about the AI bubble, mostly written by folks who didn’t participate in previous technology bubbles. I have the (good?) fortune to have participated in three of these during my career. The most extreme that I have personally experienced was the dot-com bubble. At that time, it felt like the entire world had changed. It was now possible for a software company to compete directly with “real” companies like Sears or General Motors. This was INSANE at the time. Yes, software was a big market. No, we didn’t go after Fortune 500 companies on a regular basis. Suddenly, we could.
So, let’s talk about what a bubble actually is. According to Wikipedia:
An economic bubble (also called a speculative bubble or a financial bubble) is a period when current asset prices greatly exceed their intrinsic valuation, being the valuation that the underlying long-term fundamentals justify. Bubbles can be caused by overly optimistic projections about the scale and sustainability of growth (e.g. dot-com bubble), and/or by the belief that intrinsic valuation is no longer relevant when making an investment (e.g. Tulip mania).
https://en.wikipedia.org/wiki/Economic_bubble
As an aside, I was quite pleased to discover Tulip Mania when I read this article on Wikipedia.
Getting back to our point here. The definition of a bubble is a financial one. In the business we tend to talk about things like markets or wild product claims, but those are just external symptoms. The actual definition is all about valuations vs. reality. During the dot-com bubble for example, Commerce One had a valuation around 210x. They had basically no revenue but their valuation was sky high. They eventually folded, declaring bankruptcy in 2001. I had a few friends who worked there and they told me I was dumb for not moving over. I was at Microsoft at the time. It’s crazy when someone tells you to leave Microsoft for financial reasons.
So, are we in an AI bubble?
Well, at 42x and 27x respectively, OpenAI and Nvidia certainly seem a bit high, but not as high as CommerceOne. Of course, they are WAY WAY WAY higher than “regular” companies like Adobe at 6x. So, it does feel a tad high.
OpenAI and Nvidia are often cited as examples, but they are nowhere near the top of the list. Perplexity is at 180x and Anthropic is at 61x. Of course, OpenAI, Perplexity and Anthropic are not public companies. They are startups so by definition they are going to be in a mode where they invest and focus on growth and customer acquisition rather than revenue.
Naturally, VC funding reflects this also. In the USA, AI companies captured about 35-40% of the total VC funding in Q3 2024 which is quite a jump from 14% in 2020. This tells us that VCs think that this market is hot and the valuations for pre-IPO companies reflect this.
I’m not an economist, but the data seems to be a bit hit or miss. Yes, there is significant growth in valuations and yes, they look a bit high, but not quite as high as the height of the dot-com bubble.
However, I can certainly tell you from experience that it FEELS like a bubble.
What does a bubble feel like?
Companies get funded for basically nothing. No real business model, no customers, no revenue.
Money is spent like water. Hiring goes insane. Salaries explode.
Everyone is suddenly in that business even if they’re not really.
“This changes everything” mentality. All previous knowledge is suddenly obsolete.
Are these things happening? Let’s look at the evidence.
Funding
During the dot-com boom, you could get VC money for almost anything. It was literally insane. A friend of mine started an online sock business. Yes, a website where you could buy socks. He got millions. Of course, they failed. But, Amazon is now a thing. To put it another way, real e-commerce companies came out of the bubble and if you had invested in Amazon you would have done quite well.
Today? Take the example of Safe Superintelligence (SSI) which has a $32 billion valuation and $0 revenue. Founded by ex-OpenAI chief scientist Ilya Sutskever in June 2024, SSI has raised $3 billion total and reached a $32 billion valuation with approximately 20 employees, zero products, and zero revenue. Sutskever has explicitly stated the company won't release anything until they achieve "safe superintelligence" at some unspecified future date. The valuation jumped sixfold in less than a year, from $5 billion to $32 billion. It's literally betting billions on a technology that might not exist for decades.
Spend
During the dot-com boom, companies did all kinds of crazy things. They would send the entire company to some tropical island. They would offer free services to employees like massages or a concierge. Most offered free food. Hilariously, this is a Google thing now but at the time it was a crazy excessive perk.
Today, OpenAI spent about $9 billion in order to lose $5 billion this year. Their spend on infrastructure is truly insane. They have also signed San Francisco’s largest office lease two years in a row and occupy nearly one million square feet of office space in San Francisco. Meta recently paid $100m to a single AI scientist.
The Cool Kids
During the dot-com bubble, it suddenly became popular to have a .com address. The web went from an interesting nerd hangout to the center of the business world. Literally everyone needed a website even if they had no idea what that was. Things like Myspace were huge for a short time, but quickly faded. Of course, we also got Facebook which is still here.
Today, it’s hard to find a software company that ISN’T claiming to be an AI company. Cisco, Oracle and Intel are all claiming to be AI companies now:
Oracle: What is Oracle AI?
Cisco: Cisco AI Solutions
Intel: AI Ready Datacenters
The New Paradigm
When dot-com happened, the general feeling was that this was the most amazing thing ever:
In the Web's first generation, Tim Berners-Lee launched the Uniform Resource Locator (URL), Hypertext Transfer Protocol (HTTP), and HTML standards with prototype Unix-based servers and browsers. A few people noticed that the Web might be better than Gopher. In the second generation, Marc Andreessen and Eric Bina developed NCSA Mosaic at the University of Illinois. Several million then suddenly noticed that the Web might be better than sex.
— Bob Metcalfe, InfoWorld, August 21, 1995, Vol. 17, Issue 34.[11]
As it turns out, that wasn’t completely wrong. Ten years after the bubble burst, Marc Andreesen wrote the famous Why Software Is Eating the World essay. That was 2011. Today in 2025, we still have brick-and-mortar businesses but we also have things like AirBnB, Uber and Spotify which are software-only companies disrupting very traditional brick-and-mortar businesses effectively (hotels, taxicabs and music stores, respectively).
Today, everyone is pretty convinced that AI will change everything:
“AI is going to reshape every industry and every job.”
– Reid Hoffman, Co-founder of LinkedIn
Of course, this isn’t a new thing either. The economist John Maynard Keynes famously predicted a fifteen-hour work week by the early 2000s in a 1930 essay.
So, yes. I can say that all of this feels eerily familiar. Some of the more insane claims are almost word-for-word things I’ve heard before including the idea that we won’t need to work. Of course, that claim was made in 1930 about the year 2000. Twenty five years ago.
I’m still waiting.
No comments:
Post a Comment