A fund manager ran an experiment: use an AI model to design a new antivenom, then send the designs to a real lab.
Most investors assume AI drug discovery is still mostly hype. Citrini found the opposite problem — the AI worked faster than biology can test what it makes.
"There's a huge shortage of monkeys. That's probably the most important thing for preclinical testing."
Citrini runs Citrini Research and was one of the earliest public bulls on the AI hardware trade, discussing SK Hynix on this same podcast back in 2023.
The full interview is covered here so you can skip it. 87 minutes of audio, about 6 minutes of reading.
Here are the 7 numbers that matter.
Monetary Matters with Jack Farley · Oct 25, 2026 · 1h 27m · 81m saved
With Citrini, Citrini Research
Key Takeaways
Citrini rotated out of AI infrastructure and semiconductor stocks around March, after being an early public bull, because financing has shifted from hyperscaler cash flow to equity and debt issuance
The market needs roughly $50 billion a month in net new AI annualized recurring revenue to support current spending, against roughly $13.6 billion today
An Anthropic research model designed 180 antivenom candidates in 64 hours, running roughly 308 sub-agents
17 of 211 lab-tested designs actually bound to the target toxins, including 7 against a toxin specific to a Mexican scorpion species
AI has "pretty much solved" drug design, but that was never the bottleneck — preclinical testing, toxicology and clinical trials are
Monkeys are the single most supply-constrained input to preclinical testing: breeding capacity takes about 4 years to expand, and demand runs roughly 80,000 against supply near 20,000
Charles River Laboratories is his highest-conviction long from this thesis, with roughly 60% of its revenue tied to the discovery and safety-assessment work the monkey shortage bottlenecks
1. Rotating Out Of Hardware
Citrini traced his own position in AI hardware back to an early call, then explained why he's moved on from it.
Citrini has been largely out of AI infrastructure stocks since around March
"Since probably about March, we started focusing on a different part of the AI trade that has a lot more to do with the actual use of AI... the more asymmetric opportunities are definitely inside what we call agentic utilities." — Citrini
He isn't bearish on the hardware itself.
"I'm still pretty bullish on Nvidia from here." — Citrini
