Ask HN: What Makes AI a Bubble?
A lot of people seem to take it as a given that the AI bubble will "pop", leading to a mass devaluation of AI companies from their current peaks.
What I'm confused about though is what makes current AI evaluations a bubble.
Bubbles usually exists when future speculation outpaces productivity: eventually some realization leads the market to no longer believe in that future speculation, causing devaluation which triggers a mass sell-off.
However, AI companies currently have very high revenues and are growing extremely fast. Their valuation is backed by actual commerce. I can't imagine that there is any room for a bubble, as it is very clear where the market is at, and why demand for AI is so high.
Now, certain specific companies I can imagine losing a lot of valuation, but only contingent on the fact that they serve a middle-man role in the market that improvements in the underlying AI models will solve, which would likely only mean more revenue for the frontier labs, and thus less reason for a bubble. Bubbles happen when a) someone buys something where they don't understand the value to, b) sell it off to another sucker who also doesn't understand the value. If a is not true, the prices will fall to their value. If b is not true, overpriced goods don't move, and the price doesn't blow up as fast as it does. Yes, they increase productivity, but how much? Is OpenAI really contributing a trillion dollars of value to the world? A thousand billion, a million million. 50 billion months of a $20 subscription. 80 million lifetime subscriptions of $20/month. That much value? The valuation of a company != an approximation of how much it is contributing to the world, it's more of an estimated total future potential value, a market aggregate of demand for a piece of it. I would say that the valuation of OpenAI et al exceeds a normal multiplier based on how much productivity it is contributing, but much of its value is based on the premise that the improvements to its models that have occurred in lockstep over the past 5 years won't just suddenly stop, and will continue for at least as long as the underlying compute for training scales. All AI companies are selling are tokens, which aren't really a speculative asset in that they are consumed at the moment of inference. The question is whether this token of intelligence has a multiplicative effect on the applications towards which it is directed, which currently I believe many companies are seeing positive signals towards, which is why their revenues are accelerating so much so fast. Since it's not actual A.I., I'm reminded of nuclear fusion,
which has long been only 25 years away.
It's not an actual invention yet. Yet, thanks to our times, at least one major company appears
to be thought-bubbling. It appears to hope (if it's not
just window-dressing) that fusion will suddenly appear
in the next 2 years ... to avoid driving regional
electric rates sky high. > Their valuation is backed by actual commerce. Is it? I think their annualized revenue is 25 billion with 3.4x yearly growth, with 1 billion weekly active users Now do costs. You're also ignoring the fact that these companies have been shifting things around to make their books look better than they actually are. Here's a good example explaining how META has been keeping debt and lease obligations off its books to fuel growth (and who's at risk if META doesn't pay up): https://www.reddit.com/r/economy/comments/1soent7/if_the_ai_... Many tech companies operate at a loss initially, that is the point of venture markets in firms that invest heavily in R&D, the initial investment will pay off once the technology matures. As for Meta’s shady accounting, I also inside most tech companies leverage whatever they can to remain competitive in a high growth market. They certainly have the money to get away with it though for now. > the initial investment will pay off once the technology matures That's not guaranteed. Just look at the Metaverse. It did not pay off. Hasn't it cost them $100s of billions to earn that money? Don't they need $100s of billions more to keep the ball rolling? You were talking about actual commerce, though. Is that revenue actually tied to something in the market, or is it just all of these companies and investors blowing air into the bubble? I am operating as an advisor for Indian firms targeting the international market. What I’ve noticed is that many of the companies have moved away from Intercom, Zendesk, or other such tools and built in-house simpler versions. The same thing applies to Chargbee, Chargify, and other such tools. These ready-made solutions have many features, and they are complex. Most companies only need a subset of those features but the ability to customise them. Making a general-purpose tool is very difficult. The same thing I’ve noticed for Uptimerobot, PagerDuty, and others. As a result, I suspect SaaS revenue will drop further and further. Companies are questioning why we pay $x,xxx per month for a SaaS solution when we can roll it out for some token expense, highly customizable? And it's nothing new, Google has its software managing its internal stuff. It's not about the companies having high revenue, but rather investors being really interested in AI because it's the new flashy object everyone must have nowadays. Yes, I think it will still be a thing a few years from now and later, even. But, at the moment, the AI trend is staying afloat due to how much people are investing on it. Except that companies are just losing ridiculous amounts of money due to compute costs. Which is why OpenAI had to close Sora AI and cancel their contract with Disney to allow Sora AI generated media in Disney+. I think cancelling SORA doesn’t mean their endeavors in AI are ultimately futile. It seems that with how scarce a resource a compute is, and how much of a legal minefield video models are, the GPUs would serve better working on something else. I’ve been thinking about this: Has current AI truly led to a significant improvement in productivity? For many jobs, AI seems to be getting more and more powerful, but has it fallen into the problem of a mismatch between show cases and real-world cases? If there is a bubble in AI, I think it will burst because, in the current environment of a sluggish global economy, people’s excessive expectations of AI have not been realized. >I’ve been thinking about this: Has current AI truly led to a significant improvement in productivity? In most sectors, definitely not. My day job is in commercial real estate, and in this business I keep hearing from the C-suite that it's transformative, but they just have to figure out what to do with it. They're literally salivating over the hype, but cannot seem to apply it at all. Despite this going on for years now, they remain absolutely certain that it will allow them to take the business to the next level. This is the very nature of the hype around "AI" - living in the fantasy and completely ignoring the reality. That's why I think this will go down as the largest mass delusion in the history of humanity, created by crossing social internet virality with the ancient human tendency to anthropomorphize anything and everything. All of that said, I want to be clear that I think LLMs are extremely useful for certain tasks. They just aren't nearly as useful as the hype would have us all believe. They're also about to get exponentially more expensive when investors get tired of deferring their returns, which is another problem alltogether. Because too many new companies are popping up with the business model of "We're going to use AI" and they don't actually have any explanation for how or why they're going to use AI. This is just like the Dot Com Bubble, where a lot of companies popped up saying they were going to "use the internet" without actually having a plan. P/E ratios are much lower among top AI and AI adjacent firms now compared to the dotcom bubble. There is hype, certainly, but it isn’t an entirely speculation driven market like the dotcom bubble. Even so, the internet only became more pervasive than ever after the early 2000s It’s important to keep in mind that railroads, airplanes, and the internet also caused bubbles. Just because of an invention is useful and world changing doesn’t mean it won’t cause a bubble. Exactly that. There was a very nice talk by Warren Buffett explaining that to business leaders at the height of the dot-com bubble - if I remember correctly it's full text is the introduction to his biography 'The Snowball: Warren Buffett and the Business of Life'. Airlines are a great example - they are everywhere, nobody can imagine life without them, and yet they are yet to make any money ! Maybe they will figure it out before oil runs out on planet Earth. As for Buffett speach - ther is a specific quote about airlines in it: "If a capitalist had been present at Kitty Hawk back in the early 1900s he should’ve shot Orville Wright" They make no money because competition has essentially turned them into commodities. There is no differentiation thus no producer or consumer surplus. Yes about commodities, but what you said about surplus is actually wrong. Consumer surplus is at a maximum when there is no differentiation. That is because when producers struggle to differentiate amongst competition, they must compete fiercely on price, so customers often pay far less than their maximum willingness to pay. And arguably your own statement here doesn't help the view that the industry is not a bubble, as one could argue most LLM models are not substantially different from each other, and most products integrating them are offering similar features. The industry may need to do better at differentiating if it wants to avoid being commodified. Conversely, just because an invention causes a bubble doesn't mean it is useful and world changing. What are some examples of this? Blockchain is a strong candidate Tulips? NFTs? South Sea Bubble? Mortgage Backed Securities and derivative products (CDOs etc)? Beanie Babies? LLMs, for instance Revenues start looking lot less impressive if the margins are very low at same time... Software was special place where producing more units costed only tiny bit more. But AI seems to be something where producing more will well cost almost the same. It is widely agreed users are not paying enough to cover the costs of inference. This is what "subscription" plans are. So, many users are losing the companies money. This is not discussed publicly and is covered up for by raises, because there is growth and the hope that at some point the economics could work out. Which remains to be seen. It's a variant on a Ponzi scheme. Investor hope is that at some point someone invents a way to stop losing money. If at any point investors start to lose faith that this is going to be the case, the bubble pops. If companies start to raise the token prices, at some point it won't be affordable to people. I think that no matter what they do they will just keep losing money. If they raise prices, less people will be buying the paid plans and if they don't, they are still losing money like now What percentage of Anthropic’s and OpenAI’s revenue is subscriptions? I haven’t seen a single ai-based product that’s relevant and making money. In all the companies i have worked for, ai hasn’t been a productivity multiplier It feels like the 2000 dot-com bubble all over again—excessive investments in every internet company out there. 1.) Exceptions from a society of people, half of which don’t know what a computer is, 2.) the speculative debt - Oracle is (was?) the most buried running an ai product on top of anthropic's api. from this side: customer demand is real, recurring revenue is real. but margin lives or dies on whether anthropic keeps cutting token prices. real revenue, fragile pricing. that's the bubble shape most people miss. When the tide recedes, those who are running naked will understand. Just let the bullets fly for a while. Is it a bubble? I'm not sure. Most bubbles are characterised by people overpaying for investments because the prices are going up - whether dot com shares, tulip bulbs, houses in 2006 etc. With AI there's not so much of that. A lot of investment is out of cash from the likes of Google and Microsoft who probably have an idea what they are doing. what makes AI a bubble is the return over investment. By Scam Altman's spreadsheet, openai should be spending some $100B/year with computing from partners that are building the datacenters for that. They should also be buying 40% of all available RAM. Those things are not happening.