Thrive Capital’s Dr. Nworah Ayogu on Why Neuralink Is More Interesting Than SpaceX
Summary
Thrive’s healthcare thesis is that AI will remake the entire value chain, not merely accelerate molecule design. Better binding predictions only move the bottleneck into biological validation, clinical trials, manufacturing, and commercialization; Nworah Ayogu’s shorthand is that “our problems haven’t gotten smaller—the tools we have have gotten bigger.”
Drug discovery’s critical constraint is shifting from generating candidates to predicting what they will actually do in humans. A molecule must bind the intended target, produce the desired effect, and avoid toxicity, yet biology lacks coding’s immediate verification loop: “We’ve cured cancer in mice 50 times,” Ayogu jokes, while human success remains much harder. Molly relays Max’s five-year “crazy goal” of 88%-90% success; Ayogu says even moving from roughly 5% to 60% success and compressing a 10-year process to four years would radically change which programs companies pursue.
Neuralink illustrates Thrive’s preference for “bilingual” companies that combine world-class science with world-class engineering. Ayogu estimates the company has over 1,000 microscopic implant threads, a purpose-built surgical robot, neurosurgeons, radiology, batteries, and an in-house pathology lab to restore computer control, robotic movement, or speech. His deliberately provocative call: “Neuralink is a more interesting company than SpaceX. The brain is vaster than space.”
Healthcare is already a $5 trillion annual US market, projected in the conversation to reach $9 trillion by decade-end, but the outcome ledger is damning. Ayogu says the sector supplies 20% of US jobs and represents a federal-budget line six times defense, yet roughly $40 trillion of spending over the last decade did not move national life expectancy. “Forty trillion for no change” is his indictment—and the opportunity case for companies that can prove hard outcomes rather than merely add workflow.
AI-native clinical products are reaching adoption speeds legacy health IT never achieved. Ayogu says two-thirds of US physicians use OpenEvidence after three years, versus fewer than half using Epic after roughly 40 years; the episode’s introduction separately cites company data of about 40%. The broader lesson is that doctors do not resist technology—they resist “the bad technology that they’re forced to use.”
Cadence offers the portfolio’s clearest evidence that AI plus human care can improve both outcomes and economics. Across roughly 7,000 care sites and more than 100,000 monitored patients, Ayogu cites 25% fewer hospitalizations, an over-80% increase in patients with controlled blood pressure, nearly a one-point A1C reduction, and $2.5 million saved for Medicare every week. Its prescribing agents reportedly achieve 98% concordance with clinicians by learning their practice patterns rather than attempting to replace them.
OpenAI may already function as one of the world’s largest points of healthcare access. Ayogu says 300 million people ask ChatGPT health questions each week and calls healthcare the mass-market “hero use case”: “For Silicon Valley, it’s coding. For the rest of America, it is healthcare.” The investable downstream effect is capacity multiplication—better triage, shorter specialist queues, and ultimately “a 10X nurse” rather than merely another 10X engineer.
Deep dive
1. AI is moving every bottleneck in the healthcare value chain
Healthcare has been part of Thrive’s DNA since its first investment, Oscar; O’Shea frames the firm as having reached 14 healthcare incubations. Ayogu sees an unusually broad scientific inflection: AI-driven drug discovery, individualized treatments, and the convergence of hardware with biology are arriving together. As with AI infrastructure’s demand for compute, memory, and energy, the effects should propagate from discovery through development, trials, manufacturing, commercialization, and care delivery.
N-of-one medicine is the sharpest expression of that shift: treatments can increasingly be designed for a single patient whose condition was previously untreatable. At the hardware frontier, companies such as Neuralink and Fortell are attacking quality-of-life losses that conventional medicine often leaves untouched—capabilities Ayogu calls “almost biblical in proportion: the lame can walk, the mute can speak.”
His operating principle is to anticipate where success creates the next constraint. Faster discovery without faster trials merely relocates the bottleneck; drugs emerging every three or four years instead of every 10 would similarly overwhelm manufacturing unless capacity, clinical processes, and commercialization are redesigned in parallel.
2. Drug discovery still lacks software’s verification loop
AI can increasingly propose a molecule that binds a target, but that is only the first layer of risk. Researchers must also determine whether altering the target produces the intended biology, how its three-dimensional shape changes, and whether off-target effects cause nausea, elevated heart rate, liver toxicity, or other harm.
The hard distinction from coding and mathematics is verification: “I know whether my code compiles or not.” Binding can be tested directly, but therapeutic effect may require cells, animals, and ultimately humans; each layer is slower and only an imperfect proxy for the next.
Ayogu keeps the field’s humbling joke intact: “If you want to cure your cancer, the best thing to do is to be a mouse.” Models cannot predict human outcomes reliably until researchers generate sufficiently rich experimental and clinical data, so data production becomes both a validation function and a standalone training infrastructure.
That changes the competitive frontier. Candidate generation may keep improving, but the more valuable system could be the one that closes the loop among prediction, experimentation, human evidence, and model retraining.
3. Isomorphic Labs is targeting the whole development system
O’Shea relays Max’s five-year “crazy goal” of roughly 88%-90% success in a process currently suffering 90%-plus failure, taking 10-15 years, and costing billions. Ayogu tempers the endpoint: even moving from about 5% success to 60% would “radically change” which programs companies are willing to pursue.
His defense of Isomorphic is that AlphaFold was the opening move, not the whole plan. He says the team has been thinking beyond protein folding since AlphaFold’s 2018 release, with the goal of improving molecular design, in-vitro cellular work, in-vivo animal testing, and clinical prediction enough to compress a roughly 10-year journey toward four years.
The company also fits Thrive’s preferred “bilingual” archetype: AI researchers held to the bar of DeepMind, Anthropic, and OpenAI working alongside chemists and biologists comparable to those at Lilly, Pfizer, the Broad Institute, or the Wyss Institute. The “ISO way” is not parallel excellence but constant crosstalk between the disciplines.
Trials and manufacturing remain outside any model’s unilateral control. Ayogu credits ARPA-H with considering clinical-process efficiency and argues that society and governmental agencies must redesign phases one through three; otherwise preclinical acceleration simply deposits more assets at an unchanged clinical bottleneck.
4. Neuralink vertically integrates everything mission-critical
Neuralink’s initial patients included people with ALS, paralysis, or traumatic injuries who had lost use of their limbs, often while still in their 30s or 40s. Its implant lets users control computers and robotic arms by thought; another trial placed the device in the speech center, the speech motor cortex, so a person unable to speak could produce speech.
The engineering stack is unusually broad. Ayogu estimates that more than 1,000 microscopic threads read neuronal activity, while their scale and required precision compelled Neuralink to build its own implantation robot—an achievement that could constitute another company’s entire business but is only one vertical here.
Clinical talent is equally embedded: practicing neurosurgeons co-develop the procedure, radiology supports brain imaging, battery specialists make an all-day device practical, and an in-house pathology lab studies potential brain effects. Most places outsource pathology; Neuralink brought it inside because safety data are “mission critical.”
Ayogu attributes its recruiting power to a compelling leader, contagious passion, and “N of 1” people wanting to work with other N-of-1 people. His hot take follows from the substrate itself: “Neuralink is a more interesting company than SpaceX. The brain is vaster than space”—and, in his framing, remains a more efficient engine of compute than every chip NVIDIA has ever made.
5. Fortell treats hearing as selective cognition, not volume
Hearing loss affects 1.5 billion people worldwide, Ayogu says, and raises dementia likelihood by one-third, making it the greatest modifiable dementia risk factor in his account. The mechanism may include both diminished sensory input and the social isolation that follows when restaurants, streets, and conversations become exhausting.
Conventional hearing aids amplify everything, although natural hearing depends on the brain suppressing irrelevant noise. Fortell instead identifies different kinds of sound and selectively adjusts them—an approach founder Matt DeJong connected to noise-processing work developed for Zoom during the pandemic.
Delivering that experience required a custom chip with fast inference, all-day battery life, and a sufficiently small form factor. DeJong’s team reportedly taped out its first TSMC chip successfully—an outcome Ayogu calls extraordinarily difficult—after roughly 15 years of founder obsession and about five years of full-time company building.
Product pull became the proof: public beta users Steve Martin and Henry Kravis recommended it to friends, while another tester refused to return the device—“Over my dead body.” A third-party hearing-aid evaluator whose prior maximum was 4.5 on a five-point scale reportedly awarded Fortell a six: “They literally broke the scale.”
6. Trillion-dollar healthcare companies will own more surface area
Ayogu’s macro call is blunt: “If you’re truly AGI-pilled…you should be actually long healthcare.” In a world with greater machine capability, people will still spend personal and societal resources on more time, better-quality time, and entertainment; care for the body, mind, and soul persists regardless of technological era.
Scientific products already support enormous businesses. He says Keytruda’s 2025 revenue exceeded that of all the major labs combined, using the checkpoint inhibitor to show how one drug can expand as researchers discover additional indications and combinations.
Lilly’s GLP-1 success supplies the operating template. Novo arrived first and shortages constrained the category, while Lilly brought supply online roughly a year sooner, leaned into manufacturing, and used LillyDirect to address commercialization and care delivery rather than treating science as the only competitive layer.
AI makes broader ambition feasible: biotech companies can pair science with manufacturing or care delivery, gaining larger addressable markets and greater defensibility. “Our problems haven’t gotten smaller,” Ayogu says. “The tools we have have gotten bigger.”
7. The system’s scale is matched by its failure to extend life
US healthcare spending is now about $5 trillion annually and, in Ayogu’s figures, is projected to reach $9 trillion by decade-end. It employs 20% of American workers, may account for 30% of newly created jobs going forward, and is the federal budget’s largest line item—six times defense.
Scale would be defensible if outcomes followed. Ayogu would happily spend another trillion dollars if everyone lived to 100, but Americans receive some of the developed world’s weakest outcomes: life expectancy is roughly 80, after moving from about 50 to 70 between 1900 and 1950 and then gaining another decade largely through cardiovascular progress.
The first leap came from sanitation, clean water, lower infant mortality, antibiotics, and vaccines; the later one from reduced smoking, statins, and blood-pressure control. By contrast, approximately $40 trillion spent over the last decade “has not moved our life expectancy an inch.”
The aggregate also conceals stark inequality: Ayogu contrasts an approximately 80-year life expectancy in Hawaii with 72 in West Virginia, plus a roughly 15-year gap between the richest and poorest 1% of men. Falling cancer mortality has been offset by forces including rising overdoses—“a tale of two cities” inside the same system.
8. Personalized therapies force an explicit debate over life’s price
The Merck-Moderna cancer vaccine illustrates individualized oncology as Ayogu describes it: remove and sequence each patient’s unique tumor, build an mRNA vaccine against its specific expression and genome, then add Keytruda to boost immune recognition and memory against microscopic residual disease.
He cites a phase-three result of 50% lower cancer recurrence over five years. The investable significance is both scientific and infrastructural: customization can produce durable immune surveillance, but sequencing, manufacture, delivery, and payment must all operate patient by patient.
“Baby KJ” is the more extreme specimen. Facing a rare genetic inability to process ammonia, with more than half of affected babies dying and liver transplant as the principal alternative, a Penn team reportedly diagnosed the condition, designed and manufactured a bespoke gene-editing treatment within six months, and treated the child at seven months.
The child is now two, walking and talking, but the one-person therapy cost about $1.5 million. Ayogu’s answer is morally affirmative—“increase my taxes, cut somewhere else”—while acknowledging the system-level dilemma: every transplant, cancer drug, and gene therapy asks, uncomfortably but unavoidably, “What is the price of a life?”
9. Ayogu’s operating history shapes Thrive’s healthcare playbook
Ayogu frames his career as alternating between making “good things big” and “big things good.” Trained in internal medicine after studying neurobiology and health policy, then medicine and business, he joined Cityblock Health’s founding team to scale integrated medical, behavioral, and social care for high-cost, high-need populations.
At Amazon, his team began with the company’s native advantage: logistics. PillPack became Amazon Pharmacy, using the ability to move medications from pharmacy to home; once customers trusted Amazon with drugs, Amazon Clinic tested whether they would trust it with low-acuity, transactional care.
One Medical extended that ladder from transactional to relational care, forming what he describes as Amazon Health. He joined Thrive roughly two and a half years before the interview, drawn by its healthcare history, its large AI bets, and the chance to combine his clinician and company-building perspectives.
10. Clinician adoption exposes the difference between imposed and loved software
AI deployment has followed money and workflow: revenue-cycle tools spread quickly in the back office because they improve collections; voice agents are entering the front office for scheduling and reminders; scribes and clinical-reference products are now penetrating the exam room.
OpenEvidence broke the usual pattern of doctors treating health IT “like Brussels sprouts.” Ayogu first heard residency colleagues praising it voluntarily, tried it himself, and recognized an N-of-1 product; the real lesson was that doctors own iPhones and Macs and never disliked technology—only “the bad technology that they’re forced to use.”
Ayogu says OpenEvidence reached two-thirds of US physicians in three years, while Epic remains below 50% after roughly 40 years. O’Shea’s introduction separately quotes company data of about 40%; the transcript does not resolve the discrepancy. Either way, the contrast is rapid, bottom-up adoption within a constituency historically difficult to move.
11. Cadence makes continuous care economically measurable
Cadence mails blood-pressure cuffs and scales to chronically ill patients, combines continuous signals with AI and human escalation, and integrates the result into health-system EHRs. It works within incumbent care rather than outside it, spanning roughly 7,000 sites and monitoring more than 100,000 patients.
The model addresses a simple mismatch: even a generous four annual doctor-hours leave more than 8,000 hours when illness evolves elsewhere. A rising heart-failure patient’s weight or an abnormal blood-pressure reading can therefore trigger intervention immediately instead of waiting for the next appointment.
Ayogu cites hard outcomes: 25% fewer hospitalizations, more than an 80% increase in blood-pressure control, nearly a full-point A1C reduction, and $2.5 million in weekly Medicare savings. He says these results have appeared in the New England Journal of Medicine, the Journal of Cardiology, and the Mayo proceedings. These are measures of admissions, strokes, diabetes, and cost—not merely better patient sentiment.
Its agents reportedly achieve 98% concordance with prescribers and increase successful patient contact by 70%. The mechanism matters: AI can interpret a signal against prior charts and call while the patient is still taking the measurement, before they leave for “bingo,” Trader Joe’s, or time with grandchildren.
12. OpenAI could reorganize care demand, while incumbents race to respond
Ayogu recalls testing pre-release o1 reasoning on New England Journal of Medicine cases: he and strong physician colleagues might solve two of 10, while the model went 10 for 10 in his informal exercise. Health-system leaders initially treated his diagnosis claim as “heresy” until live demonstrations changed the room; one skeptical biotech CEO later announced billions of dollars of spending on AI.
He now calls OpenAI one of the world’s largest healthcare providers because 300 million people reportedly ask ChatGPT health questions weekly. A collaboration with Boston Children’s and the Harvard hospitals applied the technology to roughly 300 unresolved pediatric cases and produced new diagnoses for about 20.
Better self-service and triage could redirect scarce specialists toward work only they can perform. If orthopedic surgeons stopped spending half their time on non-operative cases, Ayogu reasons, Los Angeles waits might fall from three or four months toward one while surgeons operate more and patients receive faster access.
Meanwhile, incumbents will build or buy rather than remain “flat-footed”: Lilly launched Tune Labs with NVIDIA, Oracle is moving, and McKesson has bought large oncology providers. Molly supplies Oscar’s recovery figures—from 95% below its December 2022 peak to above its IPO price and roughly 100% year-over-year growth—while Ayogu credits its decade-old modern claims stack with enabling faster AI-era movement. The final priority is national: “We need a 10X nurse a lot more than we need a 10X engineer.”