Alex Karnal - The Trillion-Dollar Health Revolution - [Invest Like the Best, EP.468]
What this covers
Patrick talks with Alex Karnal, co-founder of Braidwell and a former Deerfield Management investor, using the Health Stack as the organizing frame. The stack is a set of defensive layers, and the conversation keeps returning to three barriers that keep people from existing medicines: complexity, cost, and compliance. The first half is concrete pharmacology. GLP-1 is explained through its two-minute natural half-life, which engineering stretched to weekly dosing, and through its effects on stomach emptying, satiety, and insulin, along with toxicities Karnal thinks are worth accepting. The PCSK9 inhibitor case rests on a gene-mutation population with about 88 percent lower lifetime cardiovascular risk, roughly 50 percent LDL reduction, and RNA interference that could cut injections from about 26 a year to two.
The range then widens, and so does the disagreement. On the commercial side, Karnal points to Lilly Direct's growing share of new patients and to compounded GLP-1s sold at about half price through Hims and Hers as evidence of price elasticity. On screening, he warns that every added test raises the odds of a disruptive false alarm, and he argues that Guardant Health's blood test scales better than a stool test. He also forecasts a world without Alzheimer's if anti-amyloid medicines are given earlier, and he describes how cancer adapts to targeted therapy.
The last stretch turns to AI in drug discovery. Karnal says much of the published literature cannot be replicated and omits informative failures, which makes it a poor training set. He describes agentic, robotic labs running around the clock, and expects model-to-molecule design to shrink from years to about a month. He concedes that AI has not yet systematically found novel targets.
Karnal argues that most diseases that will claim most lives already have existing medicines, and that GLP-1s and PCSK9 inhibitors show people will proactively take preventive drugs once access, tolerability, and price are fixed, while AI-driven automated labs will accelerate drug discovery.
- Most lethal diseases already have medicines, so the gap is deploying them proactively rather than inventing more
- GLP-1 uptake exceeded expectations because consumers want tolerable, cheap, convenient access
- AI-driven agentic labs with robotic experimentation can compress discovery timelines and generate proprietary data
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Much of the recorded scientific literature is incorrect, and many published experiments cannot be replicated, so AI models trained on that literature risk garbage-in, garbage-out outcomes.
“a lot of the recorded literature is actually incorrect. And there's been tons of studies that show if you go try to replicate the experiments that are in the literature, you don't even get the same results.”
The body's naturally produced GLP-1 hormone lasts only about two minutes, so using it directly would require about 30 injections an hour; molecular engineering extended it to a weekly or daily dosing schedule.
“it only lasts for about two minutes before it's gone”
A population with a PCSK9 gene mutation that prevents production of the protein had about an 88 percent lower lifetime risk of cardiovascular disease over 15 years of study, which led drug developers to replicate the advantage with PCSK9 inhibitors.
“they had an 88% reduction in the risk of ever developing cardiovascular disease”
Each additional screening dimension carries a material false-positive rate, so the odds of a disruptive false alarm rise with every test, and imaging data is only valuable when placed in the right clinical context.
“the odds that you come out of that experience with a false positive that can be disruptive in life becomes material”
The published literature is biased toward successes and omits failed experiments, which are often more informative than successes for scientists trying to solve a problem.
“what's often missing is all the failures, which would be even more informative than the successes”
Cancer adapts when a targeted therapy eliminates its most vulnerable cells, because the remaining resistant cells gain a growth advantage and can regrow aggressively even after a tumor had shrunk to zero.
“the cells that were left behind – even if it's just a few – now had the advantage in growing rapidly”
GLP-1 slows stomach emptying, signals satiety in the brain to quiet food noise, and triggers insulin release from the pancreas, producing effects that span diabetes, cardiovascular, kidney, and inflammation benefits.
“The combination of those two mechanisms of action are what allow us to turn off the food noise.”
GLP-1s carry real toxicity, including nausea, vomiting, diarrhea, gallstones, pancreatitis, and muscle loss, but the speaker judges the cardiometabolic and lifespan benefits worth the risk, with obesity-related diseases costing five to ten years of life.
“We're talking about the aggregate of those diseases costing us 5 to 10 years.”
LDL cholesterol accumulates in the vasculature over time until it causes a blockage, which produces a heart attack or stroke; most middle-aged adults carry a 30 to 50 percent lifetime probability of one between ages 40 and 80, which statins and PCSK9 inhibitors could cut to under 10 percent.
“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”
Three barriers, complexity, cost, and compliance, prevent people from getting the maximal benefit of existing medicines, and the compliance barrier is the invisible one that drives people to quit.
“Those break up into, in my mind – complexity, cost, and convenience.”
Guardant Health's blood-based colorectal screening test scales better than Exact Sciences' stool test because completing and returning a stool sample is a friction that limits adoption.
“could never have the scale as a simple blood test”
Scientists have already cracked the code on most of the medicines needed to protect people from most diseases that will claim most lives, so the gap is pointing existing medicines at impact rather than needing more invention.
“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”
Colonoscopy remains the gold standard for colorectal screening and colorectal cancer grows slowly, so widespread regular colonoscopy could make low incidence achievable.
“we could very much live in a world today with very low incidence of colorectal cancer”
High blood glucose makes the vasculature more brittle and visceral fat compounds risk, so when lipids accumulate alongside brittle vessels the combined effect multiplies the chance of a heart attack or stroke.
“you're only force-multiplying the risk that you ultimately have one of those horrific heart attacks or strokes”
Eating less can help with weight, but for people with a family history of diabetes whose bodies no longer produce enough insulin to clear glucose, medicines are still likely needed to undo progression to diabetes.
“you probably still need those medicines for that”
RNA interference medicines, a newer modality that acts in the liver, have very long-duration effects so that PCSK9 dosing could move from about 26 injections a year to maybe two.
“we didn't have to inject ourselves 26 times a year, but maybe only twice”
AI is still early at discovering novel biological targets; it has not yet systematically found a target that did not previously exist, but it is breaking codes for known targets that were previously hard to drug.
“we have not yet, in a systematic way, been able to discover a target that didn't exist yet”
Compounded GLP-1s cost about half as much per month as branded versions, and roughly 15 to 20 percent of demand that could not be captured through prescriptions flowed through telehealth channels like Hims and Hers, showing massive price elasticity.
“Compounded GLP-1s cost about half as much per month as the traditional approved GLP-1s by Lilly and Novo.”
AI-driven discovery could shorten breakthrough timelines from three to five years to under a couple of years, and model-to-molecule design that once took years can now take about a month.
“can go from model to molecule in what would have taken a couple years' time, in a month”
Eli Lilly's direct-to-consumer digital channel, Lilly Direct, grew so fast that by late 2025 more than half of new GLP-1 patients were coming in directly rather than through traditional prescriptions.
“more than half of the new people joining are coming in directly”
Late-stage anti-amyloid drugs from Biogen and Eli Lilly slow cognitive decline by about 30 percent, and earlier treatment could reduce decline rates by 40 to 50 percent or more, potentially enabling a world without Alzheimer's.
“we can slow that decline by about 30%”
Longer dosing intervals, such as monthly or quarterly injectables from Pfizer and Amgen, matter because the easier a medicine is to take, the more likely patients will stay on it and get long-term benefit.
“the easier it is to take a medicine, the more likely we'll be to stay on these medicines”
Novo Nordisk data show GLP-1 medicines cut heart attack and stroke risk by more than 20 percent, and this effect held independent of weight loss, implying a separate cardioprotective biological driver.
“that was independent of weight loss. That speaks to me that there's just another axis and another biological driver”
PCSK9 medicines lower bad cholesterol by about 50 percent, cut heart attack and stroke risk by over 20 percent in previously affected patients, and by about 25 percent in high-risk patients.
“they today can lower our bad cholesterol, that LDL cholesterol, by 50%.”
PCSK9 lowering is nearly a free lunch because LDL cannot be pushed to zero in animal models beyond 80 to 90 percent, and people with natural PCSK9 deficiency live long healthy lives, suggesting low LDL carries little harm.
“this one is very much a free lunch”
The speaker expects Lilly's upcoming data later this year to show that clearing amyloid plaques earlier, before significant accumulation, dramatically protects against developing Alzheimer's.
“getting at those plaques earlier, before they accumulate significantly, is going to show dramatic effects on protecting us from developing Alzheimer's”
Early detection combined with anti-amyloid treatment could stop new plaques from accumulating, analogous to turning off a faucet, because new plaques are cleared before damage builds.
“you can think about it as turning off the faucet”
Next-generation CAR T approaches, including in vivo methods such as those from Capstan, show potential to reduce tumors by 100 percent in more than 70 percent of patients and keep them controlled for long periods.
“they have the potential to reduce tumors by 100 percent and to do that in 70 percent plus of patients”
Clear data from Eli Lilly show GLP-1 medicines produce a 94 percent reduction in the risk of moving from prediabetes to diabetes.
“show a 94% reduction in your risk of moving from being pre-diabetic to diabetic”
Scientific superintelligence in biology is likely to follow a fairly deterministic set of scaling laws, and at least one company is starting to show this curve bending.
“the idea of getting to scientific superintelligence is one that is likely to follow a pretty deterministic set of scaling laws.”
Semaglutide and tirzepatide achieve the same goals and differ mainly in novel intellectual property strategies, so from an investment perspective they are largely interchangeable.
“They were just novel IP strategies to achieve those same goals. They're all pretty much interchangeable.”
PCSK9 and LDL management suffer from lower adherence than GLP-1s because high LDL produces no felt symptoms, so patients fail to stay on drugs that protect them.
“you don't feel anything if you have a high level of bad cholesterol. You could think of it as a silent killer”
Oral Wegovy has launched at roughly 4x the relative pace of the most recent injectable launch, Zepbound, and at a $150 monthly price point it is driving weekly new-script growth.
“an almost 4x relative launch cycle for the oral Wegovy versus the most recent injectable launch, Zepbound”
FDA leadership under Marty Makary is trying to reduce friction by using AI to review sponsor filings and questioning animal studies and redundant trials that repeat the same finding.
“In what settings do we no longer need multiple redundant studies that show us the same thing? We can cut through it with one study.”
AI drug discovery companies that can generate novel scientific data, not just train on public literature, will build durable moats and follow deterministic scaling laws toward scientific superintelligence.
“the generation of those tokens becomes a big driver of their moat”
Upcoming addiction data may show GLP-1 medicines protect people from alcohol, drug, or gambling addiction, a benefit that diet alone would not provide.
“You would lose the free gift of protecting yourself from being addicted to drugs, alcohol, or gambling.”
Investors focus on the next GLP-1 delivering maximum weight loss, but patients want a tolerable drug they can stay on, so higher doses that add side effects may be the wrong focal point.
“people are not solving for this massive bazooka. People are solving for something that helps them to lose some weight, stabilize and be there.”
GLP-1 medicines are the first commercial proof that people are ready to be proactive about their health, which the speaker believes will become a once-in-a-lifetime trillion-dollar revolution in public health.
“it's actually the first commercial proof that we are ready for”
A decade into his Deerfield career, the speaker realized that most patients stopped their medicines within about a year, so the industry was optimizing returns without delivering public health impact, which motivated founding Braidwell.
“they're taking it for a year on average and then quitting, are we just optimizing for maximum expected returns but no impact on public health?”
On average people gain about five BMI points from adolescence to their 50s and 60s, and the speaker believes low-dose GLP-1s could eventually protect nearly everyone from that drift.
“on average, people are going to pick up about five BMI points”
Patients who cannot afford $400 to $500 a month will still take compounded GLP-1s that have had no clinical trials, because they value the benefit enough to accept unknown manufacturing and safety risks at about $200 to $250 a month.
“they're willing to take the risks – not only the safety risks of the medicine, but all these risks of these unknowns”
Companies such as Lila Sciences and Nabla can now use agentic systems to generate a molecule against a specified target in about a month, compressing work that once took a couple of years of in silico screening.
“can go from model to molecule in what would have taken a couple years' time, in a month.”
Because PCSK9 has an asymmetric benefit and risk profile, its long-run patient count and revenue should exceed GLP-1s, but adherence is harder since high LDL produces no felt symptoms.
“this medicine should be far bigger in terms of the number of people that are on it and the revenues that are associated with it than GLP-1s”
In the future discovery process will be largely automated, with agentic AI directing robots to run most experiments, freeing scientists from bench work that consumed about two-thirds of their careers.
“are no longer going to spend two-thirds of their career at the side of a bench pipetting all day long”
Backing innovation requires answering three questions: will the innovation work or fail, will it be commercially relevant, and can capital accelerate the company enough to produce an attractive return.
“we essentially need to be able to answer three questions”
The obesity market splits into overweight (BMI 25 to 30), obese (30 to 40), and morbidly obese (above 40); existing drugs serve the 25 to 39 range well, so the unmet need and investable right to win lies at BMI 40 and above.
“I think where we still need some help, and I think where there's a right to win, is really in that BMI zone of 40 and above.”
A defensive health stack has five layers: lipid optimization, cardiometabolic health, neurocognitive health, inflammatory health, and blood pressure, each with available medicines to control risk.
“the five key layers are lipid optimization, it's cardiometabolic health, it is neurocognitive health, it is inflammatory health, and our blood pressure.”
At the lower end of the price curve, roughly 150 dollars a month or 1,800 dollars a year, GLP-1 medicines sell rapidly, which the speaker says is the real driver of uptake.
“if we get to an $1,800 a year cost, these medicines fly off the shelves”
Multi-cancer early detection tests will likely become widely available within the next five years, and once they exist, failing to get and act on them will be a failure of public behavior.
“I think we're going to wake up within the next five years with really awesome multi-cancer early detection tests”
Medicines that must be taken for decades for chronic prevention are priced similarly to acute treatments, creating a timeline and horizon mismatch that pricing innovation should address.
“Those medicines today are priced as similarly to acute treatments, yet they need to be taken over decades.”
Karnal believed life-science portfolios could use derivatives rather than just equities to deliver the same return at lower risk or higher return at the same risk, a pitch that initially drew no response.
“a better way to build these portfolios and select securities, ways in which we might be able to either maintain the type of return”
Cancer is hard to treat because after a drug shrinks a tumor, the surviving resistant cells can regrow rapidly, and cancers are often identified too late.
“The cells that were left behind – even if it's just a few – now had the advantage in growing rapidly”
Weekly new GLP-1 prescriptions rose from about 200,000 to about 300,000 within a few months after the oral Wegovy launch.
“that has now moved from 200,000 a week to 300,000 a week in just a few months time.”
A diagnostic should be judged on sensitivity, the share of true cancers detected, and specificity, the share of cancer-free samples correctly cleared, since both determine real-world value.
“the sensitivity, and the other is specificity”
Modern overeating of calorie-dense food drives an inflammatory response that contributes to cardiovascular events and to inflammatory diseases such as atopic dermatitis, ulcerative colitis, and Crohn's disease.
“We overeat to a point where it drives an inflammatory response in our bodies.”
High-stress living, inactivity, and obesity combined with brittle vessels and rising LDL raise blood pressure over time, which is another force multiplier of cardiovascular risk.
“the pressure going through that system is going higher and higher as we're aging and doing nothing about it”
The speaker needed about two years to go from recognizing a cholesterol problem to accessing the medicine, illustrating how fragmented health system navigation discourages people.
“to two years later finally getting access to the medicine that could help protect me from that problem”
Life expectancy charts have not budged in decades, and the last major inflection came decades ago from antibiotics, vaccines, and better hygiene.
“they haven't budged in decades. The last major advancement we had, and inflection, was decades ago on the back of antibiotics, vaccines, and better hygiene.”
After animal studies, clinical development of a new drug typically takes another five to seven years to prove in people that the idea has the intended effect, making drug development a marathon.
“now you've got another five to seven years worth of clinical studies you have to do”
Ordering toothpaste on Amazon takes a day, while getting a cholesterol medicine takes hundreds of calls, multiple visits, and repeated blood tests, which shows the health system is far more complicated than consumer retail.
“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”
A health stack pairs offensive measures such as nutrition, strength training, monitoring, and data tracking with defensive medicines, and AI can help track that data to reach real impact.
“It's everything from proper nutrition, strength training, proper monitoring and testing”
Online communities testing peptides on themselves reflect a shared mission among people who want answers now rather than waiting for slow traditional pathways to test their hypotheses.
“I'm not willing to wait for scientists to go take on this problem independently, for it to go down the traditional pathways”
Booking a medical appointment can take six weeks and repeated blood tests, which is absurd compared to ordering toothpaste on Amazon, and this friction causes many people to give up.
“I think a lot of people just give up. It is really hard to navigate that system.”
Karnal learned risk and reward early by investing small amounts alongside his entrepreneurial father, such as putting in two dollars toward a rocking chair, which shaped his later comfort with capital allocation.
“I could put $2 into a rocking chair that we bought. And if we sold it for 200 bucks, I thought that was amazing.”
Heart disease remains the number one killer in the United States, which is why cardiovascular protection matters so much in the health stack.
“heart disease is still the number one killer in the U.S. today”
Working on Merrill Lynch's derivatives desk, Karnal was assigned to biotech and found the life-or-death risk-reward setting compelling enough to pursue a career in life sciences investing.
“the risk reward in that setting was about life and death.”
Grassroots peptide communities share experiments online, and while this shortens feedback loops, it may increase danger because the experiments lack controlled trials.
“If you shorten it, you probably increase the danger, but you decrease the time to learning or something.”
GLP-1 medicines will easily be in excess of $100 billion a year in revenue, and their adoption has exceeded almost everyone's expectations, including the speaker's own.
“I think it's easy now to conclude that this as a class of medicines will easily be in excess of $100 billion a year in revenue.”
Breakthroughs in medicine require scientific instinct converging with tremendous capital, and both must converge for a discovery to reach patients.
“You need scientific instinct married with tremendous amounts of capital.”
The kindest thing anyone has done for Karnal is that his wife Cass recognizes when he is at the breaking point and distracts or makes him laugh to restore balance.
“my wife Cass has this amazing ability to just know when I'm at that, like just on the precipice of that breaking point.”
Through most of 2025 the average monthly GLP-1 cost was over 400 to 500 dollars, and he attributes recent price declines to the administration.
“Through most of 2025, the average monthly cost of a GLP-1 was north of $400 or $500.”