GLP-1 drugs out-earn OpenAI and Anthropic combined, pulling labs into biotech

1. Immortality is a headline — ▶ 0:37

   Even if biological aging is solved, "the world is inherently dangerous": someone living tens of thousands of years can be taken out by something "really stupid." Colville would have said no, five or 10 years ago, to anyone alive getting that potential. He is now "more 50/50," partly from "feeling the singularity" at the intersection of AI and bio. His reasoning is that extrapolation breaks once you hit a really steep exponential, so the old no is no longer safe.

2. Waiting for aging's penicillin — ▶ 2:43

   Today's preventive tools are "a trillion-dollar wellness industry of things that don't really work": multivitamins, aspirin, most peptides. Colville's grandfather was among the first US soldiers treated with penicillin, around 1940, for near-fatal pneumonia, and lived into his 90s. He expects the aging equivalent within 10-15 years, with statins for cardiovascular disease as the early template. The goal is daily interventions that finally enable preventive medicine, beyond heart disease.

3. Podcasts came before products — ▶ 4:37

   Colville expected interventions first and the field's ChatGPT moment after. It ran the other way: Andrew Huberman, Peter Attia and Bryan Johnson brought the concept to the masses. The real step function from day-to-day interventions is still 5-10 years off, with first products 3-4 years out. The first test is Loyal's daily oral treat for dogs. Efficacy and safety packages are cleared under the RXE pathway. Manufacturing remains, likely by end of next year, "with error bars."

4. Four buckets of intervention — ▶ 8:33

   The measurement era is drawing backlash: people don't want to be told they slept poorly, they want something to dial. First-gen interventions slow or prevent aging, and nobody has proven that works in humans; he expects proof within 10 years. Second-gen means bigger effects, where AI matters: replace every cell and tissue over 50 years, restore cell age, or pause the clock. The logic for replacement is that swapping in a fresh cell is easier than repairing "hundreds of thousands" of forms of damage.

5. Replacement took 20 years — ▶ 11:27

   Shinya Yamanaka showed a skin cell can be wound back to a pluripotent stem cell. Everyone expected fast translation. It took about 20 years. Only in the past year and a half is there human data that neural cells derived this way survive, engraft and function for over a year and a half. Where replacement tissue comes from is open: a patient's own cells, or synthetic embryos, which labs have only pushed days to weeks while racing to grow transplantable organs.

6. The immortal germline — ▶ 14:30

   Parents in their 40s-60s do not pass their age to their children, or each generation would live shorter. Nature therefore has some way to rejuvenate cells or select protected ones. Colville calls this among the least studied yet most impactful phenomena in biology. It is the principle behind the $3 billion put into Altos Labs by Yuri Milner, Jeff Bezos and, by rumour, Bill Gates. Turning an aged cell back to a stem cell is thought to reverse most aspects of aging biology.

7. NewLimit tests reprogramming — ▶ 16:18

   Epigenetic reprogramming resets a cell's age without changing its identity. It was the subject of Colville's Stanford PhD, and the evidence is "very early." He rates NewLimit the leader among it, Altos and others. The company has a $3.1 billion valuation and over $600 million in cash. Its first target is liver: old mice handle alcohol badly, but an mRNA lipid-nanoparticle therapy lets them cope like young ones. Immune cells and kidneys follow, with three or four human therapies funded. Its open blog contrasts with secretive Calico and Retro.

8. Cryo is an engineering problem — ▶ 21:08

   Reversible cryopreservation differs from cryonics, which freezes tissue with ice-crystal damage and hopes for future repair. Laura Deming's approach avoids ice crystals on freezing and thawing. Colville's argument is that aging is a biology problem with thousands of poorly understood damage types. Cryo is closer to engineering, with a short, known list of obstacles. Her staging: reversibly preserve organs for transplant, a multi-billion-dollar market, on the same tech tree as whole humans. She won't sell until a path to solving ice crystals exists.

9. No single winner yet — ▶ 28:48

   It is too early to say whether restoring cell age or replacing cells wins. Genome instability might be the most upstream driver, which would allow stopping aging outright. Colville says that is possible but not likely. Even if NewLimit reprograms age inside cells, extracellular matrices accumulate damage that this may not fully fix. His conclusion is that it will take a coordinated, multi-pronged effort across approaches, and he doesn't expect one shining fix, "sadly."

10. He takes no peptides — ▶ 30:51

    Colville takes none. Data are thin, and an n of one proves nothing unless the effect is dramatic. He likes pioneers generating new data but objects to big claims from weak evidence. BPC-157 is the test case: it acts via VEGF and angiogenesis, which worries cancer researchers, yet raising VEGF in mice extended lifespan 20-25% in a well-powered study. Still, "I will refuse to tell a single soul to take it until I see it proven." Supply is a further problem, as contents often don't match labels.

11. AI isn't the bottleneck — ▶ 34:59

    AI has made molecule generation cheaper and faster and cut day-to-day work like filings and competitor research by 30-95%. Zero-shot design at Nabla Bio mostly yields new versions of existing medicines. The cost sits downstream in trials. The bigger change to drug supply over six years was China biotech: cheap skilled labour, a "996" work ethic, faster human testing. Big pharma licensed 3-5% of drugs from China then, about half last year. AI won't close that gap for 3+ years, and a zero-person biotech would still need millions.

12. Why labs want biotech — ▶ 41:19

    Semaglutide and tirzepatide out-earn OpenAI and Anthropic combined and are growing fast. Labs with huge valuations need revenue, and one drug can command more than their whole company. Colville calls biotech "very AI-resilient": patent-protected, regulated, and unlikely to be wiped out by Anthropic in a day or a year. He cites Marc Andreessen's point that regulated prices rise while unregulated ones commoditize. Anthropic's life sciences team has said it may be its second-biggest market focus; he places it in the top five.

13. Moonshots are de-risked — ▶ 56:06

    Colville used to be incrementalist. Watching Laura Deming and Loyal's Celine Halioua changed that. Dog lifespan once seemed insane and now no one blinks. Ambition recruits top talent and funding. His filter is founders with both a solid 5-year plan and a solid 15-year plan. Deming's organ preservation funds her whole-human goal. Max Hodak's Science targets 5-10 year markets en route to 20-30 year aims. Failure is built in: "if everything is succeeding, you're almost like not swinging big enough." Pharma now employs people with longevity in their titles, including at Eli Lilly.

14. A tiny market needs a Manhattan Project — ▶ 1:03:43

    Colville calls for a Manhattan Project for aging, citing Rina Ramani's analysis that underinvestment costs the US trillions. The combined market cap of aging-focused companies is probably about $15 billion, roughly 1/50th the companies and investment of cancer. Against AI funding it is invisible on a bar chart. Some big checks were not deployed efficiently, so observers ask where the progress is. His answer is that it is so early, and money in the space is extraordinarily high leverage.

15. Hype pockets in aging — ▶ 1:07:04

    Cellular senescence, the zombie-cell story, is real in a dish. In vivo, after 15-20 years, it is still the wild west, with barely any markers to identify these cells. One weak paper can spawn a thousand imitators, and some researchers chase the next grant rather than the answer. NAD supplementation may help energy, but strong human longevity data does not exist while clinics profit from the narrative. Non-GLP-1 peptides are entertaining but "not moving the needle" toward 10 or 30 extra healthy years.

The aging field has no commercial product with proven benefit in humans. Alex Colville says so flatly, and his whole outlook is built around that gap. Nobody has shown that any intervention slows the rate of aging in people. The past three years of noise did not change that. The noise came from podcasts, not drugs. Colville expected the opposite order: a working intervention first, then a cultural explosion. Instead Andrew Huberman, Peter Attia and Bryan Johnson brought the idea to the masses, and capital followed. An Uber driver in Istanbul asked him about Johnson two and a half years ago. The real step function, the point where day-to-day medicine changes how people age, is still 5-10 years away by his estimate.

That gap explains why the money looks bigger than it is. Headlines about billionaires and crypto founders suggest a crowded field. Colville puts the combined market cap of aging-focused companies at probably about $15 billion. Cancer has roughly 50 times the companies and investment. Set against AI funding, he says, "you wouldn't even see it on a bar chart." Some of the early big checks were not deployed efficiently. Calico took over $4 billion over 13 years, matched in part by AbbVie, and told almost no one what it was doing. Outsiders see billions spent and ask where the progress is. Colville reads the same facts the other way. The field is so early that each dollar is high leverage. He wants a Manhattan Project for aging and cites Rina Ramani's analysis that underinvestment costs the US trillions.

What would first proof look like? Not exotic. The first generation of interventions simply slows or prevents aspects of aging. Today's preventive stack, multivitamins, aspirin and most peptides, belongs to a wellness industry of things that don't really work. Statins are the template: a daily intervention people take once they reach a certain age. Colville wants that beyond cardiovascular disease. His grandfather was among the first US soldiers treated with penicillin, near death from pneumonia around 1940, and lived into his 90s. Colville is waiting for that moment for aging, and expects it in 10-15 years. The first concrete test is a dog. Loyal has cleared its efficacy and safety packages under the reasonable-expectation-of-effectiveness pathway. Only manufacturing remains, likely by the end of next year, with error bars. A human version would follow a few years later. Asked to bet the fund on one category for the next decade, he picks this one. The field needs a first product people love. Five years out he would lean to replace or restore, but he is torn.

The second generation is where effect sizes get large, and where he expects AI to matter. Replacement is the plainest idea. Swap in a fresh young cell or tissue, because that is sometimes easier than repairing hundreds of thousands of forms of molecular damage. The catch is how slowly such ideas reach patients. Shinya Yamanaka showed that a skin cell can be wound back to a pluripotent stem cell. Everyone expected fast clinical use. It took about 20 years. Only in the past year and a half is there human data that neural cells derived this way survive, engraft and function for over a year and a half. Where replacement tissue comes from is still open. It could be a patient's own cells. It could be synthetic embryos, which labs have pushed only days to weeks while racing to grow transplantable organs.

The second route restores age inside a cell without changing what kind of cell it is. Its motivation is reproduction. Parents in their 40s, 50s and 60s do not pass their age to their children, or each generation would live shorter. Nature has some way to rejuvenate cells or select protected ones. Colville calls it among the least studied and most impactful phenomena in biology. It is the principle around which Altos Labs raised $3 billion from Yuri Milner, Jeff Bezos and, by rumour, Bill Gates. The evidence for partial reprogramming is still very early. He rates NewLimit the leader of the companies working on it. NewLimit has a $3.1 billion valuation and over $600 million in cash. Its first target is liver. Old mice handle alcohol badly and young mice don't. An mRNA lipid-nanoparticle therapy lets old mice cope like young ones. Immune cells and kidneys follow, and the cash covers three or four human therapies over about five years. The company also publishes openly, which is rare in a field that Calico and Retro made secretive.

The third route is to pause the clock. Reversible cryopreservation differs from cryonics, which freezes tissue with ice-crystal damage and hopes for future repair. Colville's case for it is about problem type. Aging is biology, with thousands of damage types that are poorly understood. Cryo is closer to engineering, with a short, known list of obstacles. Laura Deming stages the business sensibly. Organs for transplant are a multi-billion-dollar market on the same tech tree as whole humans, and she will not sell a service until there is a path to solving ice crystals. None of the three routes is declared the winner. Genome instability could prove the most upstream driver, though he doubts it. Reprogramming a cell's inner age might leave damaged extracellular matrix untouched. He expects a multi-pronged effort, not one shining fix.

AI changes less of this than the frenzy implies. It has made molecule generation cheaper and faster, and it cuts everyday work like filings and competitor research by 30-95%. Companies such as Nabla Bio generate antibodies and small molecules zero-shot, though mostly new versions of existing medicines against known targets. Moving to new targets is, he says, inevitable. But the cost and time sit downstream, in taking an optimized molecule through safety and efficacy to approval. AI has barely dented that. So a million zero-person biotechs would each still need millions of dollars. What actually reshaped drug supply over six years was China, with no new technology. It had skilled cheap labour, a "996" work ethic and the ability to test in humans faster and at a fraction of US cost. Big pharma licensed 3-5% of its drugs from China five or six years ago and about half last year. AI will not close that gap for at least three years, and probably much longer.

That makes the labs' appetite for biotech a commercial story, not a scientific one. Semaglutide and tirzepatide out-earn OpenAI and Anthropic combined. A lab with a vast valuation needs revenue, and one drug can command more than its whole company. Biotech is also AI-resilient. It is patent-protected and regulated, and Anthropic will not wipe it out in a day or a year. Colville cites Marc Andreessen's point that prices in regulated industries rise while unregulated ones commoditize. Anthropic's life sciences team has said it may be its second-biggest market focus, and he places it in the top five. Labs that are seen curing disease also, he notes, tend not to get restricted.

The same discipline governs what he refuses to believe. He takes no peptides. A sample of one proves nothing unless the effect is dramatic, and most peptide effects are not. He likes pioneers who generate new data. He objects to big claims built on weak data. BPC-157 shows the difficulty. It acts through VEGF, which makes cancer researchers nervous. Yet raising VEGF extended mouse lifespan 20-25% in a well-powered study. Still, "I will refuse to tell a single soul to take it until I see it proven." Supply is a separate problem, because what is sold often does not match the label. Senescence has a similar shape. Zombie cells are real in a dish, but after 15-20 years there are barely any markers to find them in living animals. One weak paper spawns a thousand imitators, and some researchers chase the next grant. NAD supplements may help energy, yet the human longevity link does not exist while clinics profit from it.

His investing logic follows from this. He used to be an incrementalist. Watching Deming and founders like Celine Halioua at Loyal changed his mind. Extending dog lifespan once looked insane and now no one blinks. Ambition recruits top talent and funding. He looks for founders with a solid 5-year plan and a solid 15-year plan, such as Max Hodak at Science, who aims at 5-10 year markets en route to 20-30 year goals. Failure is expected: "if everything is succeeding, you're almost like not swinging big enough." The market is forming around them. Until two years ago no one would acquire an aging company. Now pharma employs people with longevity in their titles, including at Eli Lilly.

The obvious objection is that his calendar is more confident than his evidence. He gives first products in 3-4 years, proof within 10 and a preventive step change in 10-15. His own examples cut against that speed. iPSCs took 20 years to reach humans. Reprogramming evidence is very early. Senescence has gone 15-20 years without good in vivo markers. He concedes AI will not help with trials for years, and says first products were on track without AI. So the compression has to come from elsewhere, and he does not say where. He also never says what result would count as proof that aging slowed. His change of view on immortality is explicit: no, five or 10 years ago, and roughly 50/50 now, partly from a feeling of acceleration. A feeling is not a measurement. He admits extrapolation is hard once a curve turns steep, and that cuts both ways.

The next test is concrete. If Loyal's manufacturing clears by the end of next year, the field gets its first product people can hold. Whether that opens the way to humans, or stays a dog treat, is unresolved. Even a cure for biological aging would leave the rest of the list. Colville's own reminder is that "the world is inherently dangerous."