Summary

Twenty years into backing biotech, Cornell calls 2025 “probably the single most exciting year in my entire journey” — not just for the revenue but for the commercial proof. GLP-1s will be a class “easily be in excess of a hundred billion dollars a year in revenue,” and that is “the first commercial proof that we are ready” for what he frames as a once-in-a-lifetime, trillion-dollar reduction in annual healthcare spend.
The thesis that reorders where the money is: “incredible scientists have already cracked the code on most of the medicines we need to protect us from most of the diseases that will claim most of our lives.” Lifespan curves haven’t moved since antibiotics, vaccines and hygiene — so the gap is getting existing medicines to people, and the opportunity sits in complexity, cost and compliance rather than novel targets.
Price elasticity is the 2025 finding Wall Street underweighted. Compounded GLP-1s at roughly half the $400–500 branded monthly price pulled 15–20% of the market outside of scripts through Hims & Hers and similar, and oral Wegovy at $150/month has launched at ~4x the relative pace of Zepbound — taking the market from 200,000 to 300,000 new scripts a week in months.
His contrarian read on the drug race itself: more weight loss is the wrong target. “People are not solving for this massive bazooka” — they want tolerability and stayability, so the right to win is BMI 40+ and long dosing intervals (Pfizer’s acquired monthly asset, Amgen’s monthly-to-quarterly), not maximum potency.
PCSK9 is the real free lunch and should eventually out-scale GLP-1s. A population born without the protein shows an 88% lifetime reduction in cardiovascular disease risk; approved drugs cut LDL 50% and events 20–25%. It lags today only because “you don’t feel anything” — cholesterol is “a silent killer that’s just working in the background” with no acute feedback loop to drive persistence.
Screening is his cancer offense, and it is inflecting now: Guardant’s blood test is inflecting alongside Exact Sciences’ stool test because completion rates, not markers, were the scaling constraint, and he expects credible multi-cancer early detection within five years. He’s a personal abstainer on Prenuvo-style whole-body imaging — stack enough tests with material false-positive rates and a life-disrupting false positive becomes likely.
On AI he’s moved from career-long skepticism to conviction “very recently,” and the moat he underwrites is data, not models: companies that can generate proprietary “science tokens” robotically, because “a lot of the recorded literature is actually incorrect.” Model-to-molecule against known-but-undruggable targets already runs in a month versus years at firms like Lila Sciences; novel target discovery has not been cracked systematically yet.

Deep dive

Medicines exist; impact is the opportunity

Looking back over twenty years of building and backing biotech companies, Cornell’s dispatch is that 2025 was “probably the single most exciting year in my entire journey,” and the reason is commercial rather than scientific. GLP-1s will comfortably clear $100 billion a year, but what fires him up is that they are “the first commercial proof that we are ready” for a once-in-a-lifetime revolution in public health.
The framing that reorders the whole conversation: “incredible scientists have already cracked the code on most of the medicines we need to protect us from most of the diseases that will claim most of our lives.” Expected-lifespan charts “haven’t budged in decades” — the last inflection came from antibiotics, vaccines and hygiene — so the gap “is not necessarily needing more medicines,” it’s pointing the ones we have at impact.
When he says trillion-dollar revolution he means something deflationary and specific: “this trillion dollar reduction in our annual health care spend.” The evidence it’s coming is demand-side — people “voting with their feet,” saying in effect, “I’m done waiting to reactively go and treat myself for the diseases that I backward-lookingly manifest.”

The health stack has five layers

The stack splits into offense — nutrition, strength training, systematic testing, tracking the data over time — and defense, where the medicines live. His five layers: lipid optimization, cardio-metabolic health, neurocognitive health, inflammatory health, and blood pressure. Each is within an individual’s control to be proactive about, and each already has a medicine available pointed at it.
On lipids, the number he thinks people don’t internalize: most middle-aged men and women carry “somewhere between a 30 and a 50% probability of having a heart attack and a stroke sometime between the time that they turn 40 and the time that they turn 80.” Statins through PCSK9 inhibitors can take that “to sub 10%” — “and to me that’s tragic,” because the medicines already exist.
The layers compound, and he narrates it as one mechanism rather than five risks: glucose makes vasculature brittle, LDL accumulates inside it, the same overeating drives an inflammatory response, and pressure through that system climbs with age. “Oh my goodness, it’s like a ticking time bomb for all of that to go off.” Each axis is a force multiplier on the others.
The payoff claim is stated categorically, not hedged: get on these early and proactively and “they will undoubtedly add an extra decade of life to our expected lifespan,” with the potential to inflect curves that have been “dead flat for decades.”

GLP-1 commercialization proves demand

His first unlock cuts against the sell-side consensus. Wall Street is “very focused on the next GLP-1 having more weight loss and more weight loss and more weight loss,” but the data says “people are not solving for this massive bazooka” — they want to lose some weight, stabilize, stay there. Higher doses buy more efficacy and more side effects, so what patients maximize is stayability: “how do I get on these things and then stay on them?”
Lilly Direct was the second. Against the traditional model of armies of reps calling on doctors — capital- and talent-intensive — Lilly bolted on a digital front end letting patients get a script and receive the medicine straight from the company. By the end of 2025, “more than half of the new people joining are coming in directly.”
The compounded-GLP-1 “theater and dramatics” delivered the most important lesson: price elasticity is massive. Compounded versions ran about half the $400–500 monthly branded cost, and his team’s data suggested 15–20% of the market was flowing outside scripts through groups like Hims & Hers — people accepting unknown manufacturing and safety risk because $200–250 a month was affordable and $500 wasn’t.
Then 2026’s oral Wegovy, launched at roughly “4x relative launch cycle” versus Zepbound — which he doesn’t attribute to needle avoidance. At $150 a month, an $1,800 annual cost, “these medicines fly off the shelves”: the market has gone from about 200,000 new scripts a week pre-launch to 300,000 a week on data he checked “this past Friday.”

BMI 40-plus is underserved

The mechanism, as told: GLP-1 is a hormone the body releases when food reaches the small intestine, and it survives about two minutes. “We’d have to inject ourselves 30 times an hour for the rest of our lives” to get what one weekly shot now delivers. It slows gastric digestion, hits brain receptors driving satiety — together turning off the food noise — and triggers pancreatic insulin release that protects both vasculature and kidneys.
His own first exposure was 2005 at Amylin Pharmaceuticals, then “the most miraculous breakthrough,” with a twice-a-day shot. Today’s molecules are longer-acting and far more potent; Pfizer just acquired one that “might be a once a month,” Amgen is working toward monthly or quarterly. Dosing interval is the variable he weights, because “the easier it is to take a medicine, the more likely you will be to stay on” it.
On semaglutide (Ozempic, Wegovy) versus tirzepatide (Zepbound, Mounjaro), his team’s view is deliberately unromantic: “they all achieve the same goals. They were just novel IP strategies to achieve those same goals,” and they’re “pretty much interchangeable.”
The investment question is always “why does a medicine have a right to win?” From BMI 25 through 39 patients are “well served with any of the available options,” so the right to win sits at BMI 40 and above, where incredibly long titrations to high hormone levels and far more dramatic food-noise reduction are required. Longer term he’d bet low doses could protect anyone from the roughly five BMI points people pick up between adolescence and their 60s — while flagging that “they haven’t been tested in those populations.”

GLP-1s are not free lunches

Patrick’s challenge was the sharp one: if you simply ate exactly what you’d eat on the drug, would outcomes differ? The concession comes first — on weight alone, “you wouldn’t really need these medicines.” But if family history and pre-diabetes mean you aren’t producing the insulin to clear glucose, “it’s unknown as to whether that’ll be enough to be able to undo your progression to diabetes,” so “I might say you probably still need those medicines for that.”
Two benefits he thinks discipline alone forfeits. Addiction protection — data coming “over the next year or so” on drugs, alcohol and gambling, running through the same satiety signaling in the brain. And cardioprotection: Novo Nordisk data over the last year shows a north of 20% reduction in heart attack and stroke risk, “independent of weight loss” — which to him means “another biological driver” entirely.
Asked to steelman the smartest skeptic, he doesn’t dodge: “GLP-1s have real toxicity associated with them.” Nausea, vomiting and diarrhea drive early quits; the labels carry gallstones and pancreatitis; muscle decline comes with fast weight loss. His trade is explicit — those diseases in aggregate cost 5 to 10 years of life, and the liabilities are ones “you can do something about by just stopping the medicine.” “It is not a free lunch for GLP-1s.”

Three barriers block preventive medicines

The three barriers he names are “complexity, cost, and convenience” — though the third he actually describes is compliance. On complexity, his own case is the specimen: from a blood test flagging a cholesterol problem to finally getting the medicine took two years of six-week appointment waits and repeat blood draws. “The friction in our system mounts and mounts and mounts. And eventually I think a lot of people just give up.”
The line that carries it: “Why is it that I can go on my iPhone, I can go to Amazon, I click a button and the next day I can have toothpaste, but it takes me an incredible journey, hundreds of phone calls, multiple doctor visits, being pricked and prodded multiple times just to protect myself from having a heart attack?”
On cost, the structural mispricing worth underwriting: chronic preventive medicines meant to be taken for decades “are priced as similarly to acute treatments.” That’s a horizon problem, and he expects real innovation in pricing models as the system moves from waiting “until it’s almost too late” toward getting ahead of disease.
On compliance: “medicine only works if we take it,” and every extra dose, pharmacy trip and decision point raises the odds you quit — “that’s just human nature, myself included.” Persistence is “this invisible barrier to all of us realizing our best health outcome,” which is why twice-daily to weekly to potentially quarterly is, in his words, “a huge unlock.”

PCSK9 is the free lunch

The discovery story is the argument. Human genetics turned up a population carrying a mutation that stops them producing the PCSK9 protein, which normally interferes with the body’s ability to clear LDL. Longitudinal studies over 15 years showed that defect conferred an 88% reduction in the risk of ever developing cardiovascular disease — “That was miraculous” — and the industry’s whole project became conferring that advantage on everyone else.
Today’s drugs cut bad cholesterol by 50%, with over 20% risk reduction in patients who’ve already had an event and about 25% in high-risk patients, against heart disease still being “the number one killer in the US today.” The modality moved from injectable monoclonal antibodies to RNA interference — long-duration molecules that took dosing from 26 injections a year down to maybe two.
Patrick’s skeptic question — doesn’t LDL do good things? — gets answered with the genetics rather than theory: you can pin the PCSK9 protein to zero, but even in animal models the best achievable LDL reduction sits in the 80–90% zone, and people walking around without the protein “live incredibly long, healthy lives.” Conclusion: “this one is very much a free lunch.”
Why it isn’t outselling GLP-1s despite better asymmetry: GLP-1s give you side effects and visible results, so you know you’re on them. Cholesterol gives you nothing — “a silent killer that’s just working in the background” that can take you down at peak physical shape. His view is that longer term this “should be far bigger in terms of the number of people that are on it and the revenues” than GLP-1s, and that its risk-reward beats statins.

Early detection changes outcomes

Alzheimer’s was “a wasteland for decades,” and his read is that “we finally cracked the beginning of the code but not the full answer.” Anti-amyloid medicines from Biogen and Eli Lilly slow decline by about 30% in late-stage patients, where much of the damage is already done. Catch plaques earlier and you’re “turning off the faucet” — his guess, hedged, is that Lilly’s data later this year supports 40, 50% or more, the path to waking up “in a world where we live without this disease.”
Cancer he approaches from offense, partly because longer lives could increase its incidence: “as people live longer, eventually something’s going to get us.” Two things make it hard — we find it late, and cancer is “a sneaky devil”: shrink a tumor to zero and the few cells that survived now hold the growth advantage and redirect.
Screening is where he’s most excited, and the read-through is commercial. Exact Sciences’ stool test and Guardant Health’s blood test are both growing, but “the blood test is inflecting” because sample completion, not marker quality, was the scaling constraint. He grades every diagnostic on sensitivity (of 100 cancers, how many are caught) and specificity (of 100 clean samples, how many false positives), and expects genuinely good multi-cancer early detection within five years — “shame on us if we’re not getting them.”
On treatment the slope is what matters: CAR-T has moved from extracting a patient’s cells and arming them outside the body to companies like Capstan delivering it by IV inside the body, with tumor reductions of 100% in over 70% of patients held at bay for long periods. He anchors it personally — his brother-in-law’s father died of multiple myeloma nearly a decade ago and “his prognosis is dramatically different” today. On whole-body imaging like Prenuvo he’s a personal abstainer: more data is good “so long as you could put it in the right context,” but stack enough dimensions each with a material false-positive rate and a life-disrupting false positive becomes likely.

AI’s moat is proprietary data

The conceptual frame is scientific superintelligence: a superstar scientist armed with an agentic infrastructure at “an Einstein level or multiple of Einstein level.” Human scientists are bounded twice — by what a mind can retain and by what human hands can pipette. Remove both and you can enumerate “all n number of hypotheses,” pick the best next experiment, and run around the clock, compressing 3-to-5-year timelines to under a couple of years.
His long-standing objection, and what resolved it: models trained only on published research are “a paradigm of a lot of garbage in, garbage out,” because “a lot of the recorded literature is actually incorrect” and fails replication. So the winners need AI talent, capital, and “a novel way by which they can generate science tokens that don’t exist in the public domain” — token generation itself becomes the moat. At Lila Sciences he watched robotic arms move petri dishes automatically; one company they’re close to is “starting to show the bending of that curve,” suggesting scientific superintelligence follows “a pretty deterministic set of scaling laws.”
On what’s real today he draws a clean line: novel target discovery has not been cracked systematically. What has is drugging known-but-difficult targets — screens run in silico, and firms like Lila Sciences and Enabla can go “from model to molecule in what would have taken a couple years time in a month.” Asked whether the curve is now inevitable: “to me we’re on that curve,” and it flipped “very recently.”
The grassroots mirror image is the peptide subculture — Reddit groups self-experimenting outside the RCT and FDA pathway, people saying “I don’t have the time,” with a hypothesis they want tested now. His take is genuinely uncertain — “I think it’s hard to know” — but he pairs it with FDA leadership under Marty Makary trying to cut friction: an AI system combing thousands of pages of sponsor filings in record time, and challenging where animal evidence can yield to cell-based work and where redundant studies can collapse into one.

Impact reshaped his investing

The formative pair: a father who was “this relentless entrepreneur,” never successful at any of it, who died of Parkinson’s over a decade ago and told him in his last days that he “never stuck with anything long enough to see it through over the tougher moments.” And a mother who drilled in “we always commit and we never quit,” and convinced him he could do anything. “From my dad I got to experience and see what it means to take risk and that you can fail and life goes on.”
First in his family to college, MIT, then Merrill Lynch — capital markets, then the derivatives desk, where he was assigned to biotech and fell in love with it because “the risk-reward in that setting was about life and death.” Noticing that life-science investors all ran similar equity-only strategies, he spent nights and weekends on a deck arguing you could hold return and cut risk. He sent it to everyone. “Crickets.” Months later Jeff Kaplan, head of trading at a 12-person firm called Deerfield Management, called — and he joined in 2005 at 24, with the next-youngest person around 40.
The change of mind came a decade in, from Deerfield’s own patient-journey data: people prescribed lifetime medicines were staying on them about a year, then quitting. “Yes, medicine only works if people take it, but they weren’t.” He calls it “a really dark period” — married, first child, asking “are we just optimizing for maximum expected returns, but no impact on public health?” The answer wasn’t to leave investing but to repoint it from invention toward impact.
That became Bridgewell, built with Brian Kreiter, formerly Bridgewater’s COO, designing “from a first principles perspective, what would the operating system for human health look like” — the ambition being not “a once-in-a-decade $100 billion GLP-1 revolution” but the trillion-dollar cost-savings one. The process is a 9:15 morning meeting with scientists, biostatisticians, commercial and AI experts, traders and structured finance people, answering three questions: will the innovation work or fail, will it be market-relevant, and can it be financed at an attractive return — plus finding the medicines that “should exist, but don’t yet.” The kindest thing anyone’s done for him: his wife Cass, who spots the breaking point before he does.

Vol.213 Industry Watch 40 | Thinking Through Why CPUs Are Heating Up and the New AI Intelligent Terminal Market Overview April 21, 2026