What if the AI bubble does not exist

So many are talking about the “AI bubble” — about Capex, EBITDA, and all those economic metrics. But only few talk about the causes of the “AI bubble”.

Moore’s Law was finally retired in September 2022, and just three months later, ChatGPT went live. Coincidence?

For a good 50 years, Moore’s Law coordinated the innovation cycles of the chip industry. That could only work because companies agreed that Moore’s Law was a good idea. And lo and behold: the market rewarded them with predictable depreciation cycles and performance incentives for cooperation.

Since Moore’s Law’s retirement, the global computer has been operating under non-deterministic innovation cycles. The AI trio — consisting of OpenAI, Anthropic, and Google — has each taken advantage of scaling laws they developed themselves. Examples:

NVIDIA wants to use “Hyper Moore’s Law” to optimize the entire chip architecture at once. Google DeepMind, however, insists on Chinchilla scaling. OpenAI scales its LLMs using its in-house Neural Scaling Laws, while Anthropic couldn’t care less about Huawei’s Tau Scaling Law — let alone ANY company of any stature or significance showing interest in Huawei’s Tau Scaling Law (Huawei launched its proprietary scaling law in response to Trump’s erratic tariff policy).

Long story short: Everyone does what they want. And since this forum is populated by individuals who are neither Dario Amodei, nor Greg Brockman, nor Larry Ellison, nor Mark Zuckerberg, it’s safe to assume that decisions are being made at the management level of the Big Tech companies that we don’t see. Here’s an example:

Circular financing indicates cooperation. What’s wrong with cooperation? Maybe Larry Ellison and Sam Altman are just accomplices. Of course accomplices shake hands. Meanwhile, a few college-educated tech bros are standing in the ring, waving their hands around angrily and shouting, “OI! That’s against the rules!”

What rules?

In my view, the “accusation” of circular financing is mainly leveled by business analysts who, when asked about Moore’s Law, recite a memorized phrase but struggle to explain what implications an industry-wide, coordinated pace of innovation actually entails.

I also can’t shake the suspicion that many proponents of the AI bubble are envious of tech elites and simply want to witness the winners finally losing for once. The idea that the “AI bubble” doesn’t even exist is a theory they would never even consider. After all, that would be unfair.

In the meantime, I dismiss as incompetent anyone who tries to compare the “AI bubble” to the Big Short of 2007 — we’re neither in 2007 nor dealing with real estate; anyone looking for the “causes” of the “AI bubble” would do best to look at computer science in 2022 to today.

It’s starting to get annoying: Social media is flooded with the message that the “AI bubble” is about to burst any moment now. As long as NVIDIA’s stock doesn’t crash, nothing is going to burst here. If you want to influence the market, you’re better off doing it on the market itself: Short their stock if you’re so sure.