We have a saying around my house, muttered under your breath, in a Bronx accent: " This F-ing guy."
That's Visser to me. Because he's making me question some long-held beliefs (formed by spotting past bubbles), which is painful. From this clip, starting around 18:00, with a good graphic at 18:40:
Digital demand for tokens can, and will, scale massively; especially as token prices drop. It's also a fact that computer-driven demand can always scale faster than physical supply can grow. Fabs take 2-3 years to build. Token demand is 10X-ing in short order.
An eye-opener for me was hearing a dozen guys, 40-70 Y.O., in my men's group, gush about how much they use and love AI. Not a Luddite in the crowd. No complaining about hallucinations.
AI is not a bubble. I'm finally getting that though my thick head. Plenty of participants will indeed go bust, but the Samsungs, MU's, NVDA's could still have a long way to run from here.
All dependent on the theme that token usage continues to explode.
Later in the episode is a good piece about the acceleration that will take place once we can easily use AI on our phones.
So, God help me, I'm considering the likes of MU and Samsung AFTER they've gone through the roof.
What's the argument against this theory- 1) token demand will explode 2) physical supply to support it will take years 3) memory prices will thus continue to rise faster and stay elevated longer than the market is pricing in.