GenAI Bubble Update 10/25
An update on the following hope - GenAI is just hype and bubble and in a relatively short time is going to collapse back into irrelevancy ala 3D TV's, block chain and NFT's.
Let's review some of the postulates and confident proclamations of the bear position on GenAI and see how they are holding up over time.
GenAI training is theft
So far the complainants are losing EVERY court case. Some sources of data are managing to wrangle significant payments from the AI companies as expected. But those looking for GenAI to be knocked out on the basis of IP laws should recocgnize that ship has sailed.
To clarify - some folks win suits and get paid. Nobody has come remotely closed to "cease and desist". Even Anthropic lost only because they trained on priated copies of books, not that training on books itself was illegal.
GenAI is losing money and the money train is coming to an end
So far the investments continue to pour in. Huge profits for Nvidia, Microsoft, Amazon, and Google. OpenAI and Anthropic revenue soaring but still in the investment phase. No loss of appetite for either company.
While certainly there WILL come a reckoning if the profits don't materialize - we are already past relying on the big spending fronteir models. A plethora of open source models are already available that come close to the performance and lots of companies hosinting these models in a profitable way. We may not know who is going to be the Amazon or who will be the Lycos of GenAI - but we do know there will be web search and web shopping after any correction happens.
GenAI is going to get dumber
This theory is that GenAI trains on internet data and so much of that data is now AI slop and so the models will get dumber as they train on their own poor quality output. The old "garbage in equals garbage out" truism. And it is a truism, just being poorly applied. Turns out the billions being invested in building and training these models has enabled them to anticipate data quality issues and to work to solve for it. The result - as of Oct '25, is that the models continue to improve, at a rapid pace. AI slop is continuing to be produced, but that's a stupid/lazy/evil human problem.
Data centers won't have power or water.
Not enough electricity in the world to power AI say the skeptics. Power and water are indeed important factors in the location and building of data centers. Once again, though, this knowledge isn't held solely by the skeptics. Turns out that Microsoft, OpenAI, Amazon, IBM, and the rest have pretty smart people who also are aware of this need and are planning and addressing those needs accordingly. As of Oct '25, the infrastructure build out continues. Counter to this story is that AI models are taking less power. In aggregate, the power need is growing, but the Open Source world, particularly the Chinese, are showing that a lot can be done with far less power.
AI projects are failing
Yep, just as in the early web and big data waves. We are in the early Proof of Concept and roll out stage. Lots of lessons being learned. We have an education an experience gap, enterprise tooling gap, and vision gap. Back not that long ago 75% and more of Big Data projects (on premise HADOOP) were failing. Companies had to have a "big data plan" to be with the cool kids at their industry get togethers. It was the big thing, and massive money spent on failed projects. And yet - the technology matured, people gained experience and proper vision. Snowfake, Databricks, Synapse and Google Big Query are thriving today as many petabytes of data is being loaded into these cloud data providers. As of Oct '25, GenAI deployements have not reshaped the world. But neither has the momentum stopped. It's growing.
AI Can't Code
Coding is now the "killer feature" of GenAI. Improvements arriving in leaps and bounds. There is plenty of hype here centered around non-programmers "vibe coding" to hyped mythical success stories - and laughable "we got hacked" epigrams. Non-coders using AI to code is a thing and has benefits. But the REAL story is the productivity gains that REAL developers using these tools are accomplishing. As of Oct '25, there is no sign whatsoever of a bubble bursting when it comes to GenAI in coding.
AI Going to Take All Our Jobs
This was always an odd point for those who thought GenAI was all hype. If it's all hype, then there should be no worry about losing jobs. Truth is mixed. Companies are announcing that AI is allowing them to do more with less. We already can experience AI at fast food joints taking our orders. But here is where I am more aligned with the skeptics. In the NEAR term, I think companies are downsizing on the PROMISE that AI is going to reduce the need for workers. Rather than waiting for the successful implementation that reduces the need, I've seen some that are clearly jumping the gun. This too will shake out over time as the use cases and tools mature. As of Oct '25 - there is no big worry about GenAI resulting in massive human layoffs.
My Experience?
I've seen nothing to dissuade me from my original thoughts when I started playing with ChatGPT near 3 years ago. GenAI is everywhere in IT consulting. Still early days in some ways. Stil lots of concern about using only the approved enterprise safe tools, which are far away from the leading edge. CoPilot has improved over time. A year ago I thought it was near useless. But on projects where I've been able to actively use GenAI tools, the productivity gains are very real. My company is striving to get it's entire work force educated on these tools. We made a Guiness World Record by having a Vibe Coding hackathon with 50k participants, and 30k projects submitted. I made the finals out of 30k because I was already up to speed on the tools.
There will be consolidation. There are many coding tools coming out -- too many to even keep up with. I think eventually the big companies will win or purchase the winners. Lots of companies "we gotta add AI to our product" will see failure. Lots of "let's create an AI startup" companies will fail. But GenAI is now a fact of life and isn't going away.
