I think that your fundamental assumption is wrong. While I agree that replacing white collar workers is a "benefit" of AI (more of a detriment IMO). I think the biggest gain will be the increased productivity that it gives workers. What I am noticing in my work and those in IT (and from my own reading on AI) is that with current AI capabilities the trend is to assist the worker. I liken it to the early 90s when PCs became ubiquitous and we received a huge productivity boost as a result.

In addition, you can see a lot of new companies utilizing AI to assist. Grammarly for example. It's main goal is to augment the worker, not replace them. Agentic AI can both replace and augment the worker. Anyway, I think you had half the ROI, but forgot the existing worker productivity increase.

My own view on whether we are in a bubble is that new technology has always seen bubbles. However, that's normal and isn't always something to worry about in the long term. When something is new it's hard to tell the winners from the losers so capital is going to flow to both of them. When companies start to fail then we see the size of the bubble. However, the overall sector may be net positive even with those failures. We saw it in the early 1900s with car companies, we saw it again in the late 1990s/early 2000s with Internet companies, and again with electric vehicle companies. This really isn't anything new and I wish the media would stop pushing so many bubble items. Negativity really is a self fulfilling prophecy. It seems to me a little historical context would allow people to understand what's going on. Maybe I'm the one that's wrong here though - confidence isn't certainty.
 

The dot com bubble led to an overinvestment in lay fiber optic cables, which were then underutilized for a while, but the usage eventually caught up to the installed capacity. That is the good case. In the railway bubble some rail lines were built that were then barely used, and which were eventually dismantled. That is the bad case. The worry with AI data centers is that a good portion of the investment is in those Nvidia GPU chips, and if AI demand lags behind it is possible that the chips are outdated by the time somebody needs them.
 

The fundamental problem with the AI industry is that there is currently no realistic path to generating ROI. Like you pointed out trying to do it via layoffs isnt feasible even if you ignore the fact AI can't fully replace most of these works and maintain the same level of quality and service.

Productivity isnt it either because even if you magically increased everyone's productivity by say 30% that doesnt directly translate to more revenue.

AI as in industry is being propped up by the prices of chips and data center contracts and that cannot last forever. When those valuations eventually go down the industry bubble will pop and I think we will be left with only a few big players with deep pockets remaining.
 

I'm at a loss of how to discuss this without spending too much time writing a dissertation. In summary, my view is that AI is going to enable individuals or small teams create things which were prohibitively expensive for individuals or small teams in the past. I see three things happening both of which directly affect profitability: 1. current team size in all types of industries (from movie and game making, to drug developing, and architectural design) being able to produce currently levels with much smaller teams which allows the companies to produce more with the same teams or reduce team size while producing at current levels, 2. the creation of things that are currently not possible, new industries that we cannot forecast, kind of like no one forecasted Meta in 1995, 3. the impacting on supporting technologies - specfically I'm thinking power and then all of those offshoots (construction, etc). What I envision is AI having a fundamental impact on every single industry and not just "AI companies".

That's where the growth will come from in my opinion, and a large part of that is productivity gains. Trying to view AI through the lens of just AI companies to me is like thinking that PCs only affect IBM, HP, and Gateway in the 90s. The computer impacted every industry and I believe AI will as well. I'm not talking about artificial super intelligence, just AI being another tool to do more with less and enabling new opportunities. I think the current DC build out is no different than the DC build out for cloud computing. I didn't see people equating that to a bubble (or maybe I missed it).

Anyway, that's my thought and I hope I'm right because if I am we'll be in another age of prosperity like we were in the late 90s and early 2000s (post '91 recession in the US).
 

Usually your analogies are well thought out Tobold, but this one is way too simplistic. AI doesn't just simply replace people. It's an enabler. It will increase productivity, just like industrialization and robotization did over the last few decades. Those innovations did WAY more than cut the manual labor by 50% and we didn't get a Great Depression from that. People and markets simply shift to a new reality. New jobs and roles that we currently can't even imagine will arise. With higher overall productivity, scarcety will decrease and prosperity will increase. Where we were heading for a generation that for the first time would be less prosperous than their parents, AI has the ability to turn this around. In short: it's not a matter of dividing the pie differently, it's about increasing the size of it for everyone. Now - this doesn't take away from the fact that there might very well be a bubble. The dotcom bubble was very real, but it didn't crash and burn. The internet is still there and thriving - it weeded out the under-average implementations and brought investments down to more realistic levels. That I can see happening.
 

I am sceptical of the idea of “productivity gains”, as that would require work that was previously supply limited. But generative AI is useful mostly for administrative work. For example if a company has a team that compiles monthly sales figure into a report / presentation for upper management, doubling productivity would logically mean firing half the existing team, not putting out twice as many reports / presentations. That is different from the industrial revolution, where the technology actually increased the amount of physical output.
 

Here's a perfect example of increased productivity utilizing AI. It enabled a small team to do something that would have been extremely expensive for much less.That's a productivity gain in my opinion. Of course this isn't going to generate any profits, I think that we are viewing this very differently and am trying to show an example of a possibility. I've already heard of many other things like this (e.g. accurate AI diagnoses much quicker than with traditional methods).

https://www.sciencedaily.com/releases/2025/10/251027023748.htm
 

This shows all the signs of being a bubble. AI is likely a great long term investment if you can pick the winners, and a terrible short term one. The current numbers are very badly off, and I don't see how this could end well. Good explenation here:

https://www.youtube.com/watch?v=Q0TpWitfxPk

The biggest issue I see is one that has has already been brought up. By the time peaple figure out how to make money from AI, all the hardware in the data centers that they are spending trillions on will need to be replaced. Building out huge data centers before you know how much money you can make and how is very much putting the cart in front of the horse.
 

I think it is important to remember that the idea of a financial bubble isn't about whether the technology will eventually be useful and make money. The question rather is whether it will make money A) fast and B) to the people who are currently heavily investing in it.
 

The thing with this line of thinking is that it assumes the small team being able to create something they couldn't before will result in more money made. Productivity increasing with intangible goods like software is not a direct line to more profits.

In video games and movies we are already in an era where technology has advanced enough very small or even one man projects can put out work rivaling large billion dollar companies.

But for every Stardew Valley or Expedition 33 there are hundreds or even thousands of well made and fun games indie games that don't top sales charts.
 

This is only slightly related but I felt like commenting about what is essentially the cancelation of New World by Amazon. New World as of the recent update was getting around 30k concurrent players on Steam daily. It's recent update was praised as by many as the best the game has gotten since launch.

By all accounts a game with 30k players on Steam and also console players should be a successful game.

Yet Amazon just fired tons of people from their games division to "refocus on their core businesses" and supposedly put more resources into AI.

I dont know New Worlds Financials but if Amazon couldn't make enough profit to sustain that game with those player counts there is nothing Generative AI could have added to that studio to help.