First of all, I love your style, attitude, and usually opinions. This time, however, I seem to be strongly in disagreement, especially in your comparison. The Dot Com bubble burned a lot of cash, yes, and the AI bubble is doing exactly the same. You are comparing apples to oranges, here, putting on the same line _net_ losses to a combined income which counts the same thing twice or more. Considering NVidia as an AI company is like saying that who sells bricks is in the construction business. It isn't, it's merely profiting from it. OpenAI is burning so much venture capital money that it's hard to track, but still you report it as a profitable company. I really don't know how this post came to be, but I'm confused. Still, I could totally be wrong about this: your posts are always interesting, and intellectual challenges like this post are the occasion to grow.
First of all, thank you for not scrolling past. For someone still building an audience, that kind of engagement matters, especially when the take isn’t immediately clear.
I agree that NVIDIA isn’t an “AI company” in the same sense as OpenAI or Anthropic. That’s exactly why I framed this as an industrial transformation, not a sector boom. The profits from infrastructure GPUs, data centers, and cloud services are what make this cycle sustainable; the kind of cash generation that didn’t exist in 1999.
You’re right that OpenAI itself isn’t profitable yet, but it’s generating real revenue, not speculative “eyeballs.” The ecosystem burns capital, but the core infrastructure layer prints cash.
And to be clear, I’m not an OpenAI fan. I actually hope they don’t survive the next correction. Their dominance distorts the ecosystem. The signal gets cleaner when hype-funded monopolies collapse and companies that build real systems remain.
My main point wasn’t to glorify these companies; it was to show that AI is here to stay. It’s no longer a cycle or a trend; it’s infrastructure. The question isn’t whether it survives; it’s whether we adapt fast enough to survive with it.
Appreciate you challenging the framing. This kind of debate keeps the signal clean.
I think I can agree with almost everything you are saying here, what I'm not agreeing upon is the comparison with the dotcom bubble.
You seem to think that the DotCom bubble was about pets.com, I think that it was more about people trying to get rich investing in Internet, without knowing anything about that. The average investor could not distinguish between Pets.com and Yahoo, or Google, or Amazon. Or Microsoft, which pivoted and rebranded around ".NET". It was a bubble none the less, as it is the AI one,, with very, very similar dynamics.
Worse than the dotcom bubble, the forces driving this bubble are mainly driven by very large corporation's greed, not by an explosion of new products, many of which have no future. LLMs are not suited for the vast majority of the tasks they are publicized for, and you know this better than me, given your previous content (great, btw).
Laying off the human interaction to replace it with chatbots is something that freaks me out.
This doesn't mean that AI is not here to stay, but there will be blood, when the bubble bursts.
Great article and largely confirms what I've seen and experienced. None of the problems companies are facing are new, AI just makes them worse.
(1) Making sure the right work gets done. Companies have always struggled with this but with AI making more things possible, picking the right projects to work on becomes crucial.
(2) Managing The Talent Pool. In my 30 years in the industry, only one big company had a process for taking a fresh graduate and mentoring them through their first two years as an engineer/employee. Every other company punted and would pay top dollar in the hiring market for average talent. They would also watch their best talent walk out the door over easily solved trivialities and then have to pay a 20% premium to replace them.
(3) The review stage is the bottle neck and always has been. The Pull Request is now the deliverable and review time a precious resource so getting (1) right is even more important. The tools here are primitive and there is a huge opportunity in the tooling space for someone to rethink the PR because the review tools we have just don't cut it anymore. UX/DX are the moat you can build your company on.
Companies that solved these problems 10 years ago are the ones that take advantage of the leverage that AI brings and will survive and thrive. The ones who didn't will be in trouble.
Exactly. AI is a multiplier, not a fixer. If your review process was broken, it's now broken at 10x speed.
The PR bottleneck point is underrated. We've been dealing with this for two years. The tooling hasn't kept up with the volume AI generates.
And engineering stopped being fun. You're not solving problems anymore. You're triaging AI output. 100x the information, same human brain, 10x the fatigue. That's not productivity. That's a different job.
First of all, I love your style, attitude, and usually opinions. This time, however, I seem to be strongly in disagreement, especially in your comparison. The Dot Com bubble burned a lot of cash, yes, and the AI bubble is doing exactly the same. You are comparing apples to oranges, here, putting on the same line _net_ losses to a combined income which counts the same thing twice or more. Considering NVidia as an AI company is like saying that who sells bricks is in the construction business. It isn't, it's merely profiting from it. OpenAI is burning so much venture capital money that it's hard to track, but still you report it as a profitable company. I really don't know how this post came to be, but I'm confused. Still, I could totally be wrong about this: your posts are always interesting, and intellectual challenges like this post are the occasion to grow.
Great points, Alberto, and fair pushback.
First of all, thank you for not scrolling past. For someone still building an audience, that kind of engagement matters, especially when the take isn’t immediately clear.
I agree that NVIDIA isn’t an “AI company” in the same sense as OpenAI or Anthropic. That’s exactly why I framed this as an industrial transformation, not a sector boom. The profits from infrastructure GPUs, data centers, and cloud services are what make this cycle sustainable; the kind of cash generation that didn’t exist in 1999.
You’re right that OpenAI itself isn’t profitable yet, but it’s generating real revenue, not speculative “eyeballs.” The ecosystem burns capital, but the core infrastructure layer prints cash.
And to be clear, I’m not an OpenAI fan. I actually hope they don’t survive the next correction. Their dominance distorts the ecosystem. The signal gets cleaner when hype-funded monopolies collapse and companies that build real systems remain.
My main point wasn’t to glorify these companies; it was to show that AI is here to stay. It’s no longer a cycle or a trend; it’s infrastructure. The question isn’t whether it survives; it’s whether we adapt fast enough to survive with it.
Appreciate you challenging the framing. This kind of debate keeps the signal clean.
I think I can agree with almost everything you are saying here, what I'm not agreeing upon is the comparison with the dotcom bubble.
You seem to think that the DotCom bubble was about pets.com, I think that it was more about people trying to get rich investing in Internet, without knowing anything about that. The average investor could not distinguish between Pets.com and Yahoo, or Google, or Amazon. Or Microsoft, which pivoted and rebranded around ".NET". It was a bubble none the less, as it is the AI one,, with very, very similar dynamics.
Worse than the dotcom bubble, the forces driving this bubble are mainly driven by very large corporation's greed, not by an explosion of new products, many of which have no future. LLMs are not suited for the vast majority of the tasks they are publicized for, and you know this better than me, given your previous content (great, btw).
Laying off the human interaction to replace it with chatbots is something that freaks me out.
This doesn't mean that AI is not here to stay, but there will be blood, when the bubble bursts.
Great article and largely confirms what I've seen and experienced. None of the problems companies are facing are new, AI just makes them worse.
(1) Making sure the right work gets done. Companies have always struggled with this but with AI making more things possible, picking the right projects to work on becomes crucial.
(2) Managing The Talent Pool. In my 30 years in the industry, only one big company had a process for taking a fresh graduate and mentoring them through their first two years as an engineer/employee. Every other company punted and would pay top dollar in the hiring market for average talent. They would also watch their best talent walk out the door over easily solved trivialities and then have to pay a 20% premium to replace them.
(3) The review stage is the bottle neck and always has been. The Pull Request is now the deliverable and review time a precious resource so getting (1) right is even more important. The tools here are primitive and there is a huge opportunity in the tooling space for someone to rethink the PR because the review tools we have just don't cut it anymore. UX/DX are the moat you can build your company on.
Companies that solved these problems 10 years ago are the ones that take advantage of the leverage that AI brings and will survive and thrive. The ones who didn't will be in trouble.
Exactly. AI is a multiplier, not a fixer. If your review process was broken, it's now broken at 10x speed.
The PR bottleneck point is underrated. We've been dealing with this for two years. The tooling hasn't kept up with the volume AI generates.
And engineering stopped being fun. You're not solving problems anymore. You're triaging AI output. 100x the information, same human brain, 10x the fatigue. That's not productivity. That's a different job.