Every conversation I have with Dylan Patel, I'm really just trying to understand the supply and demand of tokens. This is a unique episode in that it's entirely dedicated to talking about both sides of that equation. We discuss: - The infinite demand for the newest models - SemiAnalysis... going from $10K on AI spend to $7M - Mythos and Anthropic's compute problem - Why TSMC spending $100B on CapEx could cause a shortage - Robotics as next demand wave - Why memory prices will double again This is my second conversation with Dylan and find myself needing to speak with him more and more often to make sense of it all. Enjoy! Timestamps: 0:00 Intro 1:00 Surging AI Spend 10:27 Token Demand 16:21 When Ideas Are Cheap and Execution is Easy 20:46 Model Hoarding 22:34 Robotics 27:03 The Compute Bottleneck 30:26 The AI Permanent Underclass 31:39 Supply Chain Reality 37:47 CPUs 42:54 Predictions: Public Backlashshow more

The day we recorded it was $10.95M but Opus 4 7 saved me lmfao

Nuts

@dylan522p Full episode on Youtube:

@dylan522p i like these types of conversations where patrick is speaking to an expert on a singular topic - in this case Dylan on AI supply/demand. very good episode. my favorite line: "If you don't use more tokens, you'll never escape the permanent underclass."

@dylan522p another banger Pat.

@dylan522p MORE invest like the best episodes.

@dylan522p Scaling laws + faster releases → NVDA, AMD, AVGO Longer GPU life + no oversupply → supports margins DRAM tight till ~2028 → MU, SK Hynix (pricing power) TSMC $100B capex → TSM + ASML, AMAT, LRCX (tailwind) CPU demand (RL + AI “slop”) → INTC, AMD Picks & shovels still win.

@dylan522p love the deep dive on compute demand and supply realities

@dylan522p $MU DRAM and HBM demand unprecedented

@dylan522p There is a very obvious inference glut coming. There are like 3 or 4 compounding aspects to supply of effective intelligence that will increase supply much faster than demand.

@dylan522p another banger form @dylan522p

@dylan522p the demand side is obvious, but supply constraints are where it gets weird. inference costs drop fast, but training the next frontier model scales exponentially. at some point, who can even afford to compete?

@dylan522p Is it on youtube?

@dylan522p Really interesting listen.

@dylan522p

@dylan522p Dylan where are the WF 1000 XM5s

@dylan522p How much of today's token demand is end-user pull and how much is teams buying ahead because access is scarce? Those look very different once supply catches up.

@dylan522p Another banger! you are a machine bro

@dylan522p Bro, thanks for this.

@dylan522p Amazing talk! Thank you and congrats!

@dylan522p Dylan is the goat

@dylan522p super uplifting. Can't wait for the mass protests. seriously though, great interview

@dylan522p The whole “escape the underclass” framing assumes individuals capture the upside. But in reality, companies instrument usage, add budgets/leaderboards, and shape behavior. End result: it stops being personal leverage and becomes an org-owned optimization problem.

@dylan522p If highly trained & experienced people help to build a better process with AI, & the need for employees is reduced, who will be the experienced people to help guide us in the future? How do you know if optimum Now is the same in the future?

@dylan522p Is this available on YouTube? Or from normal podcast providers?

@dylan522p $ASML is vital, own its supplychain, no competitors. #MOAT

@dylan522p All these forecasts of massive increases will end up in a crash. Avoid these memory companies.

@dylan522p @grok what stocks to buy then?

@dylan522p Remember. He is selling a product too.

@dylan522p De 10K$ à 7M$ en spend IA 😱 Le goulot compute est réel. Celui qui a les GPU gagne maintenant 🔥

@dylan522p Supply and demand, huh? More like an arms race in the CPU graveyard. 💀

@blknoiz06 @dylan522p @jvisserlabs

@dylan522p Compute will be the new currency

@dylan522p Dylan's point at the end around Dario / Sam; *that* is the AI bubble. Tech bros in their filter bubble, completely disconnected, scaremongering people who already feel disenfranchised / farmed by Big Tech.

@dylan522p That’s great @patrick_oshag! Thanks for your work on the pod.

@dylan522p Dylan is the goat