Quantum computing's only trick: factorization—everything else loses to a laptop
He says quantum computers have just one trick to beat classical computers—breaking factorization and elliptic curve encryption—while all other operations are slower than a laptop, which is why he's shorting the entire industry.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Quantum computing's only killer app is factorization
Shkreli says quantum computing is commonly equated with ‘faster computation,’ but globally there's only one algorithm that quantum computers can compress from exponential time to polynomial time—integer factorization and elliptic curve discrete log problems, the mathematics for breaking encryption. For everyday algorithms like sorting and merging, quantum machines are actually slower: existing quantum computers run at only hundreds to thousands of hertz, single-core single-threaded, nowhere near GPU performance. Unless your business is factoring large numbers, quantum computing has nothing commercially exciting to offer.
— Martin ShkreliWithout quantum memory, he's betting on optical computing
His argument goes further: quantum computers lack ‘quantum memory’—input size must be tiny, can't fit real-world data. He says whenever someone pitches quantum computing to OpenAI, the response is always ‘not interested.’ He and his wife tested every quantum computer in the world; they couldn't even compute 300 times 200. He's been shorting quantum computing companies, losing millions of dollars, yet says he got ‘billions of dollars' worth of insight’ in return. What he really backs is optical computing: a Microsoft paper mentioned matrix multiplication at near-zero latency in light. He thinks in twenty years today's H100, B100, B200 will all be e-waste.
— Martin ShkreliPharma treats billion-dollar drugs as waste
At 16 he worked at Jim Cramer's hedge fund, which returned 23% annually after fees. Later he discovered the ‘molecular market’ is far less efficient than the public market: he bought a drug from Bristol Myers for $2 million; that company is now valued at $6 billion. Pfizer too has been willing to dump billion-dollar drug rights for millions. With Sparsentan, his play was to take a drug originally destined for a ‘moribund’ hypertension market and redirect it toward FSGS kidney disease, for which no drug had ever been approved. FDA approved it, and thirty thousand patients got their first treatment option.
— Martin ShkreliHe co-founded EAC as effective altruism's counter-movement
He says he co-founded the EAC (effective accelerationism) movement as a counter to EA (effective altruism). He calls Dario Amodei a ‘fear-monger’ and questions whether OpenAI's breakaway to found Anthropic was legally defensible—comparing it to stealing trade secrets (distinct from the Waymo engineer who went to prison for stealing physical documents). He believes Nick Bostrom's *Superintelligence* ‘scared the hell out of’ Elon Musk and seeded this wave of AI doomsaying, while Meta, the President, and Jensen Huang ‘have the backbone’ to call it what it is: computer science, nothing to lose sleep over.
— Martin ShkreliAI will deliver nothing to drug development
His claim is stark: ‘I don't believe any AI-related drug will exist in the next decade. AI doesn't help drug development.’ He knows this from having done it—built Cray clusters to analyze drugs, ran an NLP company in 2015 that fed Watson tens of millions of papers. The hard part of drug development isn't compute; it's biology. He also takes a swipe at something Dario Amodei said two years ago: AI would make humans ‘completely unemployed’ in one to five years and ‘cure all diseases.’ He says none of that happened, and none of it will.
— Martin ShkreliDon't set the AI market ceiling at white-collar work
Faced with the claim that ‘$400 billion in venture capital is flowing to AI,’ he doesn't think that's a valuation trap. He pushes back against capping AI demand at ‘knowledge workers’—that severely undercounts total addressable market because demand for intelligence is unlimited. His example: as long as intelligence isn't ‘asymptotic’—i.e., more tokens still bring better answers—McDonald's could easily spend $10 billion renting ten million GPUs from OpenAI just to compute ‘how should the next burger be designed?’ The NPV math works out.
— Martin ShkreliThis AI wave mirrors the internet: consumers and Meta are the winners
His analogy for the AI market is the internet: the real winners aren't infrastructure providers but consumers and companies getting ‘near-tax-free’ intelligence windfalls, just as Cisco only got a sliver of the pie. He singles out Meta as most likely to ‘make too much’ because Meta has both distribution and has shown product chops (he mentions Muse). His read is we're still on ‘day one’ of AI; in hindsight, today's chatbots will look childishly primitive.
— Martin ShkreliRefusing to admit he wasn't the best cost him his freedom
On Sam Bankman-Fried: he lumps himself in the same category—people branded ‘genius’ as kids who carry the burden of having to prove they deserve the label, an obsession that pushes you toward disaster. His real lesson, he says, wasn't the facts of what he did but ‘refusing to accept I wasn't the best hedge fund manager.’ That unwillingness to fold is what sent him to prison. Prison showed him that regardless of background or fortune, everyone at heart wants roughly the same things—like being a good father.
— Martin ShkreliIn their own words · checked verbatim
The modern quantum computers run at a clock speed of hundreds of Hertz or thousands of Hertz, not at the gigahertz that we're used to. They're single core and they're single threaded when it comes to doing a job like a merge sort or something like that.
Martin Shkreli3:02
Chelsea and I have tried every single quantum computer in the world. We cannot multiply 300 by 200 on any of these.
Martin Shkreli6:04
Pfizer would sell you a drug for a million bucks that was worth a billion dollars. They didn't care.
Martin Shkreli9:14
I think that I want to pay deference to him creating a trillion dollar company, but I kind of hate the guy.
Martin Shkreli13:27
I don't think there'll be a single AI-relevant drug in the next 10 years. AI does not help drug discovery.
Martin Shkreli15:28
McDonald's can afford to spend $10 billion and say, OpenAI, give me 10 million GPUs. I want to know what the best next sandwich we should launch is. And that's an NPV positive.
Martin Shkreli17:29
I had a hard time admitting I was not the best hetero manager in the world, but I was a wonderful entrepreneur. That simple refusal to admit that put me in prison.
Martin Shkreli22:36
Figures
| Shkreli's prison sentence | 7 years | 0:00 |
| Price increase on life-saving drug | 5000% | 0:00 |
| Jim Cramer hedge fund annual return (after fees) | 23% | 8:11 |
| Price paid for drug from Bristol Myers | $2 million | 9:14 |
| Market value of that company today | $6 billion | 9:14 |
| Venture capital flowing into AI (per Jensen Huang) | $400 billion | 16:29 |
| Hypothetical McDonald's AI compute purchase | $10 billion for ten million GPUs | 17:29 |
Glossary
- ECDLP (elliptic curve discrete log problem)
- The mathematical problem underlying modern encryption that only quantum computers can quickly solve.
- Jevons paradox
- When efficiency improvements increase total demand, used to counter the idea that computing power will become scarce.
- EAC (effective accelerationism)
- A movement countering effective altruism that advocates rapid advancement of transformative technologies like AI.
- NPV positive
- When discounted future cash flows exceed initial costs, making an investment economically justified.
How to listen
Investors and entrepreneurs focused on hard-tech valuation bubbles, innovative-drug commercialization pathways, and AI industry pricing logic.
4:03–5:03: Guest caught in repetitive hedging language with no substantive content; skip.