“Here's a prediction market on whether people believe the Citrini report's going to come true. Something like 12% believe the Citrini scenario is going to…”
Sacks explicitly fades the AI-doom narrative ('I actually don't believe in the Citrini or the Doomer take'), which maps cleanly to NO on a market that pays out if the AI sector busts. The non-obvious step is routing his named-but-unsurfaced Citrini market to the closest faithful, liquid proxy: the broad 'AI bubble burst in 2026?' contract.
“This is why the market's getting whacked. I think that you're right, Sacks. Nobody knows. So, if you can get 5% for owning government bonds, why are we taking…”
Nobody knows if AI cash flows survive, so 5% risk-free govt bonds beat paying up for equity risk.
“I don't think AI is going to get rid of Salesforce, but it could eat into their growth opportunity. We just don't know. What if it changes the pricing model?”
Sacks expects Salesforce to survive but sees AI compressing its growth and possibly breaking the per-seat pricing model, which is a mild bearish lean on the growth multiple rather than a solvency short. The link is that a slower-growth, AI-pressured CRM should trade at a lower forward multiple even if revenue holds.
“if there's a lot of competition at both those layers of the stack, do the chip companies get it all? I think it's a unclear question”
If model and app layers commoditize each other, the chip layer captures the value; demand stays token-constrained.
“I think you're going to see a 10x in the demand for tokens, but I also think you're going to see a 90% price reduction in the cost of an output token,…”
10x token demand against 90% cheaper output tokens, with $10B revenue per gigawatt, scales the leading model lab.
“I think Trump's going to lose the midterms and we're going to get to more chaos again.”
Jason stakes a clean directional call that Trump's party loses control in November, which maps to Democrats flipping the House. The market already prices this near-certain at 80c, so the edge is in the few points of remaining gap rather than the direction.