The AI Bubble Is 17 Times the Size of the Dot-Com Frenzy, Analyst Says
Posted3 months agoActive3 months ago
marketwatch.comTechstory
calmmixed
Debate
80/100
AI BubbleDot-Com BubbleTech Investment
Key topics
AI Bubble
Dot-Com Bubble
Tech Investment
An analyst claims the AI bubble is 17 times larger than the dot-com frenzy, sparking debate among HN users about the validity of this claim and the potential consequences of an AI bubble bursting.
Snapshot generated from the HN discussion
Discussion Activity
Very active discussionFirst comment
38m
Peak period
71
0-6h
Avg / period
13.9
Comment distribution111 data points
Loading chart...
Based on 111 loaded comments
Key moments
- 01Story posted
Oct 3, 2025 at 2:13 PM EDT
3 months ago
Step 01 - 02First comment
Oct 3, 2025 at 2:51 PM EDT
38m after posting
Step 02 - 03Peak activity
71 comments in 0-6h
Hottest window of the conversation
Step 03 - 04Latest activity
Oct 7, 2025 at 3:01 AM EDT
3 months ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
ID: 45465969Type: storyLast synced: 11/20/2025, 4:26:23 PM
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
In other words, this is a statement on current US interest rates, not AI. The analyst just chose to single out the AI industry because they believe it has no ability to return on investment. But the actual numbers are not AI-specific.
Yet it doesn't show on the chart of misallocated capital. There was a huge amount of misallocated capital destroyed then:
https://images.squarespace-cdn.com/content/v1/5f3bf522f05eaf...
Maybe it would make more sense with the full context of the original report.
There are many many examples of companies that make little money or are actually losing money, yet their valuations are sky high. In other words, valuations are not in touch with reality. I see a massive bubble in the public markets as well.
As of the time of this comment, it's a whopping 247. There is no possible justification for that kind of valuation. Even if Robotaxi is a huge success, they sell more rides than Uber, Lyft, and traditional taxis services combined, it wouldn't justify it.
When the dominant voices begin to claim, "This time it's different!!" that's usually a sign we're nearing the end.
Look this up. During the dot-com bubble, the popular notion emerged that the business cycle might have ended. "The end of the business cycle!!" Many media outlets pushed the narrative that humanity would experience endless growth thanks to technological advancements.
During the housing bubble, towards the end, multiple media outlets suddenly started repeating the idea that housing prices, in real terms, have never fallen.
When I heard that farcical argument gain steam, I got out of housing immediately—before the bubble burst.
So, for me, that AI-bubble skepticism is still the dominant opinion suggests that this bubble is far from over.
I think we're closer to the Netscape IPO than we are to the .com crash.
Maybe we're at the halfway point. Just one person's opinion.
It has now progressed to what you might call a "mainstream contrarian" position. The relevant Overton window has shifted to where it is just barely in the acceptable discourse window.
It's been a while, but if I remember my dotcom bubble, that never happened from the zero-skepticism position. It was all RAH RAH RAH and then blammo it blew apart. Skepticism was always around... again, I know because I was a skeptic and I arranged my career, with modest but far from total success, around the bubble not being sustainable, but everyone around me had no idea what I was talking about (although since my alternative was "go to grad school" I didn't get much flak since they figured the market would still be as hot years later). The skeptical position was never in the Overton window, so I don't think the comparison is useful on this particular metric.
Personally I'm watching, intrigued, to see how far the general awareness that this is a bubble can go while the bubble refuses to pop. At some point that awareness needs to translate into ceasing to feed the bubble the feedstock capital it is so rapaciously consuming, but "I can always find a bigger idiot to unload on" can run quite a ways, it seems. (I don't find that an appealing investment thesis personally but I seem to be in the minority.)
There were some very prominent skeptics. Warren Buffett was regularly quoted as saying he didn't get what was going on and speculated that maybe it was time to retire.
Where I worked, people had seen tech boom and bust cycles before but what was different was the irrational exuberance wasn't just confined to a handful of stocks. Many people really did believe it was a new era and old rules didn't apply, and they were only too happy to check in on their rising Schwab and E-trade balances.
To be fair, I think it's definitely a bubble, but it's hard to compare something like this.
And you have a boom in hardware manufacturing, in Asia mostly, but even in europe you see new companies popping up (before Covid we had like 2 comapnies that could print custom pcbs, both german, now i found like 5 in France (2 of them are the german who expended but still))
Fortunately I'm employed but I have limited ideas on how to gain an edge for the next jobsearch, whenever that is.
1. Living in a tech hub like the Bay Area or Seattle. There may be tech hubs in RTP, Austin, Boston, etc but these tend to be inshoring offices and are oftentimes the first on the chopping block when offshoring is considered, because they never built the internal gravity needed to own P/L and roadmap, and those offices that did are few and far between.
2. Concentrate on doing a reputable online or part-time MSCS (eg. GATech, UIUC, UT Austin) and concentrating on fundamental courses. Coding is commodified, but programming isn't. Just understanding how to glue together Python, C/C++, or whatever language and associated libraries is not enough. Technical complexity is rising, and skills like understanding OS internals, understanding the fundamentals of SGD, or truly understanding how to derive a path tracing algorithm matters.
3. Hyper-specialize in a specific industry. "AI" is broad, just like "Mobile" was broad. What matters is how these platforms are applied to a specific industry subdomain. If I'm a cybersecurity company working on building an AI SOC, I'd rather hire Engineers who understand both core fundamentals of AI and Cybersecurity. If you are an engineer who can understand how to communicate both the business incentives as well as the engineering incentives you are worth your weight in gold.
The thing is, this isn't 2000 anymore. Eastern Europe, China, India, ASEAN, LATAM, and other regions of the world have fairly large and robust dev and tech scenes, and async+remote work has been proven out during COVID, thus removing one of the biggest barriers to offshoring. Additionally, there has been a reverse brain drain since the mid-2010s from the US to those countries that is making it easier to find talent to open an office following American norms.
I think as a mid-career engineer, you have the tools to survive this kind of a change. Any American SWE who graduated in the last 10 years is in a worse position because their universities failed them by watering down curricula to compete with bootcamps, reducing the business justification for building a domestic new grad pipeline aside from a couple top target programs.
As a former SWE-turned-PM, you won't survive as a PM or Designer if this is the reason you became one. This is how bad PMs become PMs.
The PM and Designer market is very competitive, and PMs are expected to be domain experts of their entire market category and have engineering chops to not get BSed by engineers and have showmanship to successfully manage sales.
> I simply lack the time (and brain, maybe)
I'd try to find the time if possible. Maybe take a single course a semester to ramp up. Despite the negativity on HN, there is a lot of interesting innovation in the AI/ML world - both on the math/algos side as well as the infra/platform side.
Alternatively, as mentioned by others, really leverage your network and friendships to get jobs. It's a superpower you have as an old-timer in the industry - leverage it.
Assuming you are located in a major tech hub, I don't see a reason that you should be too worried about chronic underemployment - especially in an AI-driven world which has made domain experience even more critical.
Honestly, just working on your side projects like you mentioned, and finding companies that are aligned with what your side projects aim to do puts you ahead of 95-97% of candidates.
The job hunt has been hard for people, but most candidates (especially the noisiest ones) don't really have your profile or background, but demand principal or staff level salaries and roles.
> thinking of something that combines left/right brain
Staff/Sr Staff/Principal Engineer or Architect type roles might fit the bill for you. And with your YoE, I think if you were to job hunt you could leverage your network to land that kind of role with ease.
Honestly, you are not the type of persona that has difficulty landing jobs (holding for geography ofc).
An AI/Vibe Coding market only makes domain experience more critical because hallucinations can cause significant monetary damage, so people who can act as the voice of reason for both the technical and business needs are extremely valuable.
If you can bridge the gap between nerd and corporate, you should be able to sail fine on to the next. You won't get any of that sweet prompt engineering money but pushing your skill set with a hint of AI should settle something.
I hope anyway.
Definitely interested to see where my thinking is off here, if anyone can help. This just doesn't seem much like a bubble to me.
Are there tons of uses for 2023 crypto farm hardware?
A trillion dollars of GPUs in various stages of obsolescence and/or shagged-outness in data centres might be useful to the right people but it's a depreciating asset.
If you're talking about purpose-built ASICs, probably still yes? It's not like it's performing some incredibly unique action when farming for crypto.
The secondhand electronics market is fairly strong.
Still, I think it's always something of a gamble. Maybe look for reputable dealers on there.
The property will still be worth something but nowhere near the expected value pre-crash. For those still liquid when the dust settles, they'll be able to get great bargains on compute or even whole data centers.
Maybe the land will be dotted with half-finished data centers that sit for years, like unfinished condo high-rises post 2008 crash
Jonathan Blow took this notion to task when said that the means of production are a $200 laptop you can buy at Walmart.
What's really going on is industrial grade existential angst. "If my competitor trains a robot to do this job, they can run their factory 24x7, dramatically lowering their costs. If they do it and I don't, I can't compete. Therefore, I've got to do it or die." Rinse and repeat for call centers, web shops with "assistants," or software feature mills.
Those CEOs and CTOs are convinced that the other guy is doing it, and their business will fail if they don't. There are a few that know better.
This isn't a class struggle. It's cold-sweats-at-night fear of the future.
That is bunk, in the current age. The means of production is a $200 laptop and a $1B nvidia cluster.
> What's really going on is industrial grade existential angst. "If my competitor trains a robot to do this job, they can run their factory 24x7, dramatically lowering their costs
This is not exclusive to what I said. All this "lowering of costs" is archieved by a reduced dependence on the working class.
No struggle but the class struggle.
Well, to the extent you're right, the capitalists are behaving rationally. What would Marx say about that?
But you seem to be looking for some sort of political argument on top of corrupt and twisted logic, so I don't know exactly what you're looking for here.
No, see, there are other struggles. The reason Marxism is fundamentally totalitarian and always devolves into a single party dictatorship is EXPLICITLY because it attempts to explain all suffering as caused by class conflict. Thus: if class is destroyed, utopia is achieved.
This is of course farcical.
"no struggle but x struggle" is the very the definition of a totalitarian ideology. You really think there's no struggle but class struggle? When you are suffering because a loved one has died of old age, is that struggle the class struggle?
Fuck off with that totalitarian nonsense.
But let's assume for the sake of argument that you are wrong. What are the logical implications? Can your worldview allow for the possibility that you might be wrong?
This kind of a statement assumes that all those CEOs did exhaustive, detailed, independent research to arrive at the conclusion that AI is the next big thing.
IME more often than not, they're just following the herd in a way that ends up being a circle.
The real issue is: what are the consequences if they are right - and what should be done to mitigate them.
This isn't a serious comparison, but Jonathan Blow is one to oversimplify concepts outside of his domain.
Could Jonathan Blow host his streams off his $200 laptop?
You may be right on the other parts, but, theoretically, yes? We've had the capabilities since BitTorrent to build p2p streaming tools: our main limitation for such situations is not capital, but the ability to coordinate making such tools simple and resilient enough for everyone to use.
Someone else downthread said "No struggle but class struggle", which it really does require a Marxist outlook. My take is "No problems but coordination problems" -- we have enough capital, labor, and raw resources now to solve most of our remaining problem set, we just need ways to coordinate better. Some of that is about aligning incentives. Marxists try to do this too! I just haven't seen them succeed very well without having to fall back on other, more general methods.
Digital technology, the domain of people here, is the main engine of better coordination in the modern era. We need to keep working on it to make it work better for everyone trying to solve problems.
You saying "he could theoretically stream off his laptop" is no different than saying a luddite could "theoretically sew clothes with needle and thread". It really comes across like a lack of respect for the argument Marxists make. I'm not asking you to be a Marxist but if you aren't willing to engage the argument seriously, I'm not sure what I can say.
Even your "solution", is exactly what the early Marxists thought. They needed to coordindate to overcome the sole power of factory owners and they did that through striking. What is striking, if not coordinated market power? The rest of your post comes off as Marxism, but in a different font. It's as wishy-washy as the solutions proposed by the Marxists on Bluesky. "if only we would come together and coordinate, the revolution will be here soon!".
The issue with "No problems but coordination problems", is that today, in our society, capital solves coordination problems. And before that was monarchy, and before that was religion. The entirety of human civilization can be distilled with "how do we get n-million people to do x?"
I don't consider myself as a Marxist but the actual problems he highlights are very salient even if his followers have failed to solve them. Just because there's no found solution for P=NP, doesn't mean the entire field of computational complexity is a wash.
The point though, is that you really can stream off a laptop: we had BitTorrent; we have WebTorrent. We've built peer-to-peer infrastructure that works. In the last 25 years we have built great collective works with minimal capital outlay -- so minimal that we saw an explosion of capital investments because the returns to low capital were so high. Heck, you're using one of those low-capital, capital-adjacent artifacts right now.
In the class of problems, scaling up peer-to-peer collective action is (I would suggest) difficult, but not as difficult as it would have been to coordinate peer production in the 19th century. We've just barely begun to explore their potential.
The guy is smart, but many of his opinions are those of a man who doesn't get much pushback on what he says. It's a common curse of success that you get surrounded by yay men.
I mean, you come on here and rave about how everybody should unionize and grab their pitchforks and struggle against the machine and what-not, and then you wonder why everybody is rushing to replace you with a robot.
When they take too much, that's how they make revolutions. They only have power and wealth because we let them.
> and then you wonder why everybody is rushing to replace you with a robot.
Why not rush to replace them with robots, instead? What social function do the rich serve that can't be replaced with a badly-tuned LLM?
That has nothing to do with it. Really.
Anyone in the powerful position you describe who "earned" it, that is got themselves there somehow, is fully aware of the basis of their power and of the risk represented by it resting on workers. They will try to replace them, regardless of whether or not they are currently at that moment agitating for unionization.
Not old enough for the dotCom (36) but my educated guess is that because the internet was new and exciting, companies were prepared to train and take risks.
Training for AI now requires specialist skills which can be syphoned from within an existing organisation or from those with existing skill sets.
Unlike the latter where anyone can pick up a programming language. Not everyone can do quantum algebra mechanics or whatever AI/ML uses.
Plus the cost of running AI is expensive. SME's don't have the resources to hire those for new jobs; AI kit and training. So it's easier to hire from within then outside.
AI is a rich boys toy.
Sounds editable. But from what I understood is that they're correction coordinates to something thats not sitting straight on a curve; I'll pass.
I never got given maths at school.
If we were still in ZIRP things would be insane.
E: getting some downvotes, feel free to elaborate where you disagree.
By that I mean that the learning we should take from the dot com bubble is that the bubble should not have burst, not that we should not have let it go that high in the first place.
The vast majority of the constituents of the DotCom bubble went entirely extinct and their stock prices went to zero. You are looking through a lens of survivorship bias.
I'm guessing you are referencing the price of market cap weighted index funds when you say "prices went back up", which is just one aspect of the DotCom bubble. The primary event was the prices of most of the hyped companies going to zero and the investors losing everything.
The number prints from market cap weighted indexes are somewhat arbitrary calculations that doesn't necessarily reflect the underlying market behavior.
>The dot com bubble was aberration.
It was a fairly typical bull mania, seen many times in history, and seen since as well (DotCom 2.0 was the 2020-era cycle with ten billion dollar Peloton, ten billion dollar space tourism companies, companies like Nikola with literally no product hitting 30 billion, etc, but don't take my word for it, look at the chart of the two events: https://static.seekingalpha.com/uploads/2024/10/9/48422113-1...)
>the bubble should not have burst
Bubbles are often based on a reality of future growth, which is what draws people in. The fact that Amazon was eventually worth obscene amounts of money doesn't negate the fact that hundreds of listed companies went to zero. That is the DotCom bust. The bust of a huge swathe of real companies in the real economy, and their listed stock prices going to zero.
Passive ETF domination is more of a recent phenomenon anyway. The world was much more actively managed and stock picking/mutual fund oriented until the last decade or two. Funds like Janus getting decimated with 330 billion AUM in year 2000 money was the catastrophe in the fund space. Index tracker funds like QQQ were secondary to that when it came to DotCom.
So I don't think there will be as big of a fallout. My bet is there will/should be a healthy winnowing, but not a crash (at least not one due solely to AI hype).
OpenAI is losing 4x its income from operating expenses.
Facebook is pouring tens of billions into it
Google, fuck knows.
These are all loss leaders, looking to make the others fold, rather than say a geared company that is profitable but geared to expand.
better AI costs much much more to make and run, and there isn't much brand loyalty.
For AI now, the question is (a) how well one can monetize a LLM (size of the market); and (b) how much better one LLM can be than another (enduring competitive advantage). Both of these are generation-long questions.
The other factor is who's investing. A lot of money now is from (mostly foreign) non-players who are paying to play, and who have a vested interest in managing their online and business ecosystems. They're not really subject to uncertainty about (a) or (b). Any good LLM will likely be good enough for them, and they already know what they're using it for. So when turbulence hits, those in it for profit will drop out, ceding control to those in it to control society.
The bubble bursting is not the important risk.
I don't believe this to be the case.
Worst case it's mostly a bolt-on for most companies not a "core" money maker. It will be mostly be the hardware vendors like Nvidia and the AI as platform that will get hit.
Given that a half (or 80% or whatever) trained model is worthless, there is a threshold for each new model cycle to continue. As soon as the anticipated ROI fails to justify that cost, new model production stops. Dead.
Old models will still be used, but will become more and more out of date, and less and less useful. This will be the end of AI.
(sorry, paraphrasing/developing my comment on an earlier thread)
Where I use it outside of work:
- Product review summarizations (Amazon.com, HomeDepot.com, etc)
- Search summarization (top of DuckDuckGo, Google, etc)
- Chat (eg ChatGPT. Better Search, Q&A)
It's not on my TV, barely on my iPhone, not in my smart home, just generally not much of a presence for me outside of work. Why isn't it changing my life yet? Or perhaps the average persons?