Summary
Martin Shkreli joins the All-In podcast for a wide-ranging conversation that covers technology investing, pharmaceuticals and his time in prison. The tech portion centers on his skeptical view of quantum computing and his outlook on the AI compute buildout.
Shkreli argues that quantum computers are not simply faster computers. He says the field was founded on a single algorithmic advantage, Peter Shor’s approach to integer factorization and elliptic curve discrete logarithms, and that quantum machines run at far lower clock speeds, have no quantum RAM, and cannot beat Nvidia GPUs on most workloads. The hosts push back with the case that the field is at an early transistor-like stage.
On AI infrastructure, he suggests that today’s H100, B100 and B200 chips could end up obsolete within 20 years, points to photonic computing as a possible successor, and questions the push to power data centers with space-based or nuclear energy. He remains long semiconductors, citing Jevons paradox, while noting that far more efficient transformer architectures could reduce demand for Nvidia hardware.
The episode also covers how he made money in biotech, including a drug for focal segmental glomerulosclerosis, royalty-style pharma investing, and how he manages his own portfolio today.
📺 Source: All-In Podcast · Published October 12, 2026
🏷️ Format: Podcast
