Kevin Warsh takes the reins at the central bank, why the AI memory chip boom faces a cyclical reality check, and Evercore’s formula for decoding prediction markets.
Big fan of your work! I am a new ish writer on Substack, (in economics, finance, geopolitics), and would really appreciate if you subscribed. It would mean the world to learn and grow with a writer I highly respect :)
Really interesting. I think the distinction between structural goldmines and cyclical winners is an important one, especially with insane capex dulling the Mag 7s glow - but ultimately it’s the tech companies at the centre, with the hardware companies riding the waves around them, eaves that are very much subject to change depending on the moves made in AI…
Besides the fact that they spam advertisements, I really dislike these prediction markets like Kalshi and Polymarket, and I feel that there’s way too much noise and it’s not always accurate because of the sheer number of bots and insiders (like how that Navy Seal was betting on geopolitical events that he was literally doing a while back and got caught). The main issue for analysis is it’s very hard to differentiate whose a trigger happy gambler versus an informed insider, so, although I’m no expert, stay skeptical whenever you see people referencing surface level observations (like Blah Blah has a 54% chance of happening according to Kalshi).
The signal-to-noise problem is real, but Nevada's First Judicial District Court already flagged a different angle in March: it issued a fourteen-day TRO against Kalshi, calling the contracts unlicensed gambling. When the state built on games of chance says you're running a casino, the "wisdom of crowds" framing gets harder to defend. Insider trades and bot noise are downstream of that unresolved question.
The AI infrastructure buildout is real. The assumption that hardware demand scales linearly forever is where the risk begins. One efficiency breakthrough and the market suddenly remembers semis were cyclical the entire time.
Every time Wall Street declares a cyclical industry has permanently changed, its usually right before it hasnt.
The market already ran this experiment in March: Google's TurboQuant cut KV-cache memory 2.6x in practice, and memory stocks crashed on the spot. What struck me was that Micron had just tripled revenue with 75% gross margins on a former commodity, and the stock still fell. The cyclical reflex you describe doesn't even wait for the breakthrough to ship. It front-runs the rumor of one.
Really interesting breakdown of how Fed signals and tech momentum are starting to diverge. Feels like the market is balancing between liquidity optimism and valuation caution right now. The part on momentum indicators especially stood out, a reminder that trends can stay strong longer than most people expect, until macro shifts force a reset.
The momentum point is the live one here. The Valuation entry has a number that fits: Kalshi pulled $263M in fee revenue last year, and 89% of it came from sports. So the prediction markets people treat as a crowd-sentiment crystal ball are mostly running on game-day betting flow, which is its own kind of momentum that holds until it doesn't.
Prediction markets around Fed expectations are interesting because the number moves with crowd anxiety as much as with new information. The post's framing as a mirror, not a crystal ball, holds up. Context tells you whether a 70% cut probability reflects fresh data or just everyone reading the same Warsh headline and piling in.
Excellent post! I think most investors don't do much risk management analysis and are treating Polymarket like a casino. You have to be very careful because very clever traders know how to profit from retail. That said, I love the idea of Polymarket since it provides us asset managers with a lot of information, but you have to know how to distinguish valuable information from price manipulation on prediction platforms.
The retail-vs-clever-trader framing might understate it. Fourteen of the twenty most profitable wallets on Polymarket are bots, and one turned $313 into $437K in a month. The "crowd wisdom" being aggregated is mostly automated execution reading other automated execution. https://thesynthesisai.substack.com/p/the-price-machine digs into what that does to the signal.
Big fan of your work! I am a new ish writer on Substack, (in economics, finance, geopolitics), and would really appreciate if you subscribed. It would mean the world to learn and grow with a writer I highly respect :)
https://ananyaa2108.substack.com?utm_source=navbar&utm_medium=web
Subscribed!
Really interesting. I think the distinction between structural goldmines and cyclical winners is an important one, especially with insane capex dulling the Mag 7s glow - but ultimately it’s the tech companies at the centre, with the hardware companies riding the waves around them, eaves that are very much subject to change depending on the moves made in AI…
Besides the fact that they spam advertisements, I really dislike these prediction markets like Kalshi and Polymarket, and I feel that there’s way too much noise and it’s not always accurate because of the sheer number of bots and insiders (like how that Navy Seal was betting on geopolitical events that he was literally doing a while back and got caught). The main issue for analysis is it’s very hard to differentiate whose a trigger happy gambler versus an informed insider, so, although I’m no expert, stay skeptical whenever you see people referencing surface level observations (like Blah Blah has a 54% chance of happening according to Kalshi).
The signal-to-noise problem is real, but Nevada's First Judicial District Court already flagged a different angle in March: it issued a fourteen-day TRO against Kalshi, calling the contracts unlicensed gambling. When the state built on games of chance says you're running a casino, the "wisdom of crowds" framing gets harder to defend. Insider trades and bot noise are downstream of that unresolved question.
Agreed. I’m inquisitive to see how it’ll all pan out.
The AI infrastructure buildout is real. The assumption that hardware demand scales linearly forever is where the risk begins. One efficiency breakthrough and the market suddenly remembers semis were cyclical the entire time.
Every time Wall Street declares a cyclical industry has permanently changed, its usually right before it hasnt.
The market already ran this experiment in March: Google's TurboQuant cut KV-cache memory 2.6x in practice, and memory stocks crashed on the spot. What struck me was that Micron had just tripled revenue with 75% gross margins on a former commodity, and the stock still fell. The cyclical reflex you describe doesn't even wait for the breakthrough to ship. It front-runs the rumor of one.
Great article
Really interesting breakdown of how Fed signals and tech momentum are starting to diverge. Feels like the market is balancing between liquidity optimism and valuation caution right now. The part on momentum indicators especially stood out, a reminder that trends can stay strong longer than most people expect, until macro shifts force a reset.
The momentum point is the live one here. The Valuation entry has a number that fits: Kalshi pulled $263M in fee revenue last year, and 89% of it came from sports. So the prediction markets people treat as a crowd-sentiment crystal ball are mostly running on game-day betting flow, which is its own kind of momentum that holds until it doesn't.
I think the prediction markets section is the most useful part here.
They are becoming hard to ignore as a signal, especially around Fed expectations, but I agree that the context behind the number matters a lot.
Prediction markets around Fed expectations are interesting because the number moves with crowd anxiety as much as with new information. The post's framing as a mirror, not a crystal ball, holds up. Context tells you whether a 70% cut probability reflects fresh data or just everyone reading the same Warsh headline and piling in.
Excellent post! I think most investors don't do much risk management analysis and are treating Polymarket like a casino. You have to be very careful because very clever traders know how to profit from retail. That said, I love the idea of Polymarket since it provides us asset managers with a lot of information, but you have to know how to distinguish valuable information from price manipulation on prediction platforms.
The retail-vs-clever-trader framing might understate it. Fourteen of the twenty most profitable wallets on Polymarket are bots, and one turned $313 into $437K in a month. The "crowd wisdom" being aggregated is mostly automated execution reading other automated execution. https://thesynthesisai.substack.com/p/the-price-machine digs into what that does to the signal.