Unraveling AI's Trillion-Dollar Contradiction

October 2025 saw the largest compute deals in history and the rise of models that may not need them. This is the new fault line that will define the market.

Every month, the Chocolate Milk Cult reaches over a million Builders, Startup Founders, Investors, Policy Makers, Leaders, and more. If you’d like to meet other members of our community, please fill out this contact form here (I will never sell your data nor will I make intros w/o your explicit permission)- https://forms.gle/Pi1pGLuS1FmzXoLr6

---

Every month, our team reads thousands of pages of AI research, market reports, and policy papers. We rigorously test tools, interview experts building in the forefront of technology, collaborate with multiple top investors, and independently verify research claims. We compile the most pressing developments in our monthly AI Reports to help builders, investors, and tech leaders stay ahead of the curve. For example, our Market report predicted the fracturing of transformer dominance in favor of other architectures. Specifically, we predicted the rise of Diffusion Models and Latent Space Reasoning as two major avenues for massive short-medium growth.

Since then—

1. Inception Labs raised a 50 Million USD round for their work on Diffusion Models. Their work drew attention from several notable figures like Elon Musk, who said they were excited about Diffusion Models since their generation was not token by token, leading to a higher performance ceiling (which is bar for bar what we laid out many months ago).

2. Samsung’s TRMs showed the breakthrough that can be accomplished by latent space reasoning, beating cutting edge model with a 7M param one.

However, October was different. There was one theme that dominated everything else. Nvidia 5 Trillion? OpenAI committing 100s of Billions it doesn’t have? AMD and Oracle all seeing spikes because of that? And all the major data center buildouts?

Was all this sustainable? And how will AI valuations change when we consider the rise of highly intelligent, cheap edge models that remove the need for Data Centers altogether? This last question becomes even more pressing when Kimi-K2-Thinking showed China creating cutting-edge AI on older hardware through better architectural decisions. If China can keep doing this, then are the massive valuations around hyperscalers bound to pop?

That is exactly what this report will answer.

To access the full article—and all premium breakdowns going forward/written prior—upgrade to a premium subscription below.

Each piece is rigorously researched, built from firsthand signals, and written to make you sharper than the noise. It takes a long time for me to compile, analyze, and verify research. If you believe deep insight deserves support, become a premium subscriber to allow me to keep doing the same.

Flexible pricing available—pay what matches your budget here.

Most companies offer learning or professional development budgets. You can expense this subscription using the email template linked here.