Why Read: The LLM boom always sounded irrational to anyone paying attention to the buildout statistics touted by its top spokespeople. The ubiquitous deployment of flawed LLMs was "projected" to cost trillions, and to achieve exactly what result was never quite clear. This week, we take a look at what we were told would happen, what actually happened, and what is likely to come. - era

I think we're in an LLM bubble, and I think the LLM bubble might be bursting next year [...]. But "LLM" is just a subset of AI when it comes to applying AI to biology, chemistry, image, audio, [and] video. I think we're at the beginning of it, and we'll see much more in the next few years[...]. - Clem Delangue, Co-Founder and CEO, Hugging Face, November 2025

No company is going to be immune, including us. If you overinvest, we'll have to work through that phase. - Google CEO Sundar Pichai on the AI Bubble, BBC (11/18/25)

Around this time last year, we published "SNS: The Trillion Dollar Promise: On Data Centers and Deception." The premise of that piece was simple: LLM CEOs were bouncing along the conference circuit raising money for data-center buildouts with the promise of a new era of AI, and their requests were ludicrous. This new age would cost us dearly, we were told, but for the ultimate tech package it was surely worth it. We needed to start spending those trillions. We noted that the sheer numbers for proposed data center buildouts lacked credulity at all, stating:

They're so costly, in fact, that US$3 trillion is already lined up to be spent between now and 2030. The McKinsey report quoted above, focused on trends for use of compute, posits that $5.2 trillion will be needed to meet those needs.

That is a lot of moolah.

Such astronomical costs put this buildout on par with the greatest endeavors of human history. World War II is estimated to have cost the United States $340 billion in 1945 dollars - around $6 trillion in today'sdollars.

The absurdity of the situation has only grown.

What started as a laughable claim has become something of a threat. Not only will these chatbots supposedly make everyone jobless, but we're now told on anear-weekly basis by LLM company executives that it is increasingly likely that they'll end humanity entirely. They are so dangerous, it is now said by those making them, that "we" may collectively have to slow down and start to take greater care. But still, please, trillions should be spent; thank you for your time.

All this, and yet it is still quite unclear what, precisely, these things are worth to us. Trillions in spend so far has led us to a point where unhinged cybersecurity breaches, AI-generated slop-code failures, and massive environmental ramifications are running head-on into the dreamy (nightmarish?) future vision of Silicon Valley's LLM darlings.

Something's gotta give.

When we examine what's likely to break first in this unhinged business model, the financial cost is the most obvious place to start. There are two major ways that this is all getting paid for: massive debt, and circular financing. Let's start with debt.

Heavy Burden

Well, I could assume, um, that the revenue will continue growing 10x a year, so it will be 100 billion at the end of 2026 and 1 trillion at the end of 2027. And so I could buy a trillion dollars, actually it would be like 5 trillion dollars of compute because it would be a trillion dollars a year for 5 years - I could buy a trillion dollars of compute that starts at the end of 2027 - and if my revenue is not a trillion dollars, if it's even 800 billion, there's no force on Earth, there's no hedge on Earth, that would stop me from going bankrupt if I buy that much compute.

And so even though a part of my brain wonders if it's going to keep going 10x, I can't buy a trillion dollars of compute in 2027. If I'm just off by a year in that rate of growth, or if the growth rate is just 5x a year instead of 10x a year, you go bankrupt. - Anthropic CEO Dario Amodei, Dwarkesh Patel podcast, February 2026 (transcribed by SNS)

The LLM industry has been given much of the money it was seeking. With trillions invested and data centers going up across the nation, the buildouts have gotten so big that companies are hoarding everything from GPUs to memory. The price for chips is sky-high. As noted in the quote from Amodei, the fervor with which these companies have been building and the immensity of the calculations for their infrastructure requirements to continue is massive.

Amodei is correct. These numbers are galling. And bankruptcy is an obvious risk here. And the LLM competition is stiff enough that there are now really only three major players.

All of the air is getting sucked out of the room.