I originally joined the workforce just when the dot-com bubble was about to burst in Spring 2000. While I have some recollection of what took place during those months, I have zero digital souvenirs from those days. Because people didn’t have blogs, there was no social media, our phones didn’t have internet nor cameras. Most importantly: it was a pre-meme era.

This time it’s different. No, not the bubble. I’m talking about how tech has given us the means to preserve a piece of it. Yet if we just do occasional social posts into the endless feeds that are guided by algorithms, ain’t nobody gonna find them. Which is why I decided it’s worth my time to collect some artifacts into this newsletter issue. At least I can export the HTML of it and store it on a USB drive thrown into a box. Then possibly discover it before the next bubble emerges.

So, listen up, future me: here’s what it looked like in August 2025.

Was the Sam Altman dinner on Thursday, August 14th, the thing that put the big dominos in motion? You know, that casual chat where he said some folks are going to lose a phenomenal amount of money with AI:

One week later, the news articles about other tech CEOs making statements we hadn’t seen from then before had already become a daily event. Gone were all the earlier news about $100M compensation to AI engineers. Instead, the earlier AGI evangelists now seemed to say, “you know, let’s not focus too much on that silly AGI thing”.

Internet sickos like Ed Zitron of course were more than happy to spread these news. After all, the theory of AI subprime crisis had been first presented at least a year earlier. If things would now really start to fall apart, there was a lot more at stake than just the valuation of companies like OpenAI.

The only way Big Tech knew how to join the AI race was to spend Big Money. Because when you can’t yet show any significant revenue, let alone profit, at least the massive AI capex spend will signal that you are not ignoring this revolutionary moment in technology.

In his article “Honey, AI Capex is Eating the Economy”, Paul Kedrosky came to the conclusion that capex spending for AI contributed more to growth in the U.S. economy in the past two quarters than all consumer spending. So, rather than just being tech giants splashing their money around, there was now a real-world economic dependency to the continued spend on AI.

What did the AI capex translate into, besides NVidia GPUs? Construction. In the US, there was more money being spent on building data centers than almost any other non-residential construction category in 2025.

What this also meant was putting all the eggs of market growth into fewer baskets. FT’s illustration of the market capitalization concentrating on the top ten largest US companies made anyone old enough to remember year 2000 see familiar signs of a certain event starting with the letter “B”.

We’re used to technology getting better and much cheaper over time. Betting on this sounded like a believable strategy for AI startups building services on top of foundational models. Except the scaling laws of LLMs ended up biting those companies in the behind.

Not only was training the next foundational model requiring exponentially more computing resources. Making AI services “think” before responding, in order to reduce hallucinations and make them useful, was consuming tokens at an ever-higher rate.

Read the take by Ethan Ding on this token spend to learn why there is no flat subscription price that works in this new AI world. Unless you’re at the scale of Microsoft and have huge enterprise contracts that can lock the customers into your ecosystem for eventual indirect monetization.

It was tough for Microsoft in the Summer of 2025, too. Posting record results while doing more layoffs than ever before. Even Satya said he felt bad about having to do job cuts. Yeah, those GPUs aren’t going to do the work for us, but we cannot choose NOT to invest in them, now could we?

The looming stock market reaction was likely to be much quicker than the reaction of the average C-suite. If Big Tech CEOs like Satya Nadella were grappling with how to find enough money to set on fire in the data center construction business, the CEOs in less techy companies were feeling some serious AI FOMO.