Originally published in Second Opinion, a newsletter for healthcare founders, operators, and investors.
If the United States’ healthcare sector were its own country, it would be the third largest economy in the world. By any rule of capitalism, a market that size should have minted a titan. Software gave us multiple trillion dollar platforms. Retail, search, advertising, even payments each produced a generational tech giant. And yet healthcare has not produced a single $100 billion healthtech company.
Why? The short answer is that technology, for the most part, hasn’t actually created value in healthcare. It shuffled it, bureaucratized it, and in many cases destroyed it. The longer answer is that might all be changing now.
The money is chasing the wrong layer
In 2025, US digital health startups raised $14.2 billion, up 35% on the year, a roaring rebound that should, in theory, be hunting the biggest prize in the room. But follow the marquee checks and you find them clustered around the administrative layer of medicine. AI scribes that transcribe the visit, payer and provider-ops tools waging agent-on-agent warfare, and workflow automation that shuffles tasks between humans a little faster than before. These are real businesses and some of them are good businesses. But all of them orbit the edges of care rather than advancing care itself.
Meanwhile, AI built to actually diagnose, treat, and manage patients isn’t highly valued. This is the strange part, because roughly half of that $5 trillion of annual US healthcare spend is clinical labor; doctors, nurses, therapists, and the hours they spend with patients. That time is the single most expensive and most rationed resource in the entire system.
Most of the capital is chasing the smaller prize.
Technology never solved healthcare’s core problem
Investors aren’t irrational for avoiding the clinical layer. Healthcare has spent decades teaching them a painful lesson: technology rarely translates into productivity gains.
That’s not because healthcare hasn’t innovated. Few industries have seen more technological progress. Advanced imaging, robotic surgery, precision medicine, and electronic health records all transformed aspects of care. But they shared a common limitation: none fundamentally changed the relationship between clinician time and patient demand.
A surgeon equipped with better tools is still one surgeon. A therapist with a better workflow is still one therapist. More patients ultimately required more clinicians, which meant costs continued to scale with labor.
This is healthcare’s version of Baumol’s cost disease. While software, manufacturing, and communications became dramatically more productive over time, healthcare remained tethered to clinician time. A doctor can only see so many patients in a day, which meant costs continued to rise even as the rest of the economy became more productive. The technology improved the quality of care, but it did little to increase the amount of care a clinician could deliver.
The result was that while healthcare became more technologically sophisticated, it simultaneously became more expensive. Productivity stagnated, spending climbed, and investors learned to view claims of healthcare efficiency with skepticism.
That skepticism was rational. Until now.
Why AI is actually different
AI is the first technology that can actually alter the production function. Not by replacing clinicians but by decoupling clinician time from patient time in a way that wasn’t previously possible. A skilled therapist can see eight patients per day. With well designed Clinical AI delivering evidence-based interventions, tracking progress, surfacing red flags for human review, that same therapist potentially supports hundreds of patients. The human remains in the loop, but the human is no longer the rate-limiting variable for every unit of care delivered.
This matters most in the fields where the labor shortage is most acute: behavioral health, primary care, chronic disease management. The waitlist for mental health services in the US is six to twelve months in most cities. The shortage of primary care physicians is projected to reach 25,000 within a decade. These aren’t distribution problems that can be solved by matching supply and demand. They’re fundamental capacity constraints and capacity constraints are exactly what AI was built to solve.
Solving Jevons paradox in medicine
For the first time, it’s possible to imagine healthcare capacity scaling faster than healthcare labor, deliverable at near-zero marginal cost. But a breakthrough this powerful is also a loaded weapon, and the trigger is the reimbursement model. Cheaper care drives more consumption. That’s Jevons paradox, arriving in medicine.
Build Clinical AI on top of fee-for-service and you don’t bend the curve, you detonate it. The same technology that could collapse the cost of care could just as easily explode spending by generating more billable activity. Whether AI lowers costs or raises them depends entirely on what you pay for.
The solution is outcomes-based pricing: you get paid for health produced, not activity generated. If the patient’s depression remits, you get paid. If the chronic condition is controlled, you get paid. If the intervention fails, you don’t.
That single shift changes everything. Every productivity gain accrues to the provider. AI no longer optimizes for visits, claims, or utilization; it optimizes for outcomes. You’re no longer building a billing system with a clinical wrapper. You’re building a health production system.
If AI can reduce the cost of delivering a therapeutic outcome from $5,000 to $500, the economics become extraordinary. You can serve patients who were previously uneconomic to serve, scale without hiring linearly, and compete on results rather than labor.
The $100B health-tech company
I’m reminded of Stewart Butterfield’s observation that “in the long run, your market cap is proportional to the amount of value you create for your customers.”
The first $100 billion health-tech company will be a vertically integrated clinical AI provider that does three things simultaneously: delivers measurable clinical value, materially bends the cost curve, and gets paid for outcomes rather than activity. Its moat will not be the model itself, but the proprietary data, infrastructure, evidence, and actuarial capabilities required to consistently produce and price health outcomes at scale. Building that foundation will take years.
Outcomes-based healthcare requires control of engagement, intervention delivery, measurement, and pricing. That naturally favors integrated platforms over point solutions. Producing health outcomes reliably is not a single-product challenge. It requires owning the entire system that generates, measures, and captures value.
Once such a platform emerges, choosing not to partner with it will become a competitive disadvantage, much like a consumer startup deciding to ignore Google Ads or Meta. The economics will simply be too compelling to ignore.
Healthcare has never produced a trillion dollar company because no technology could create healthcare output without healthcare labor. Every healthcare company of the past was, at its core, a labor company.
AI changes that equation. For the first time, it creates the possibility of a healthcare company that scales like software while producing real-world health outcomes. That is the opportunity investors have been searching for, and the reason healthcare’s largest company may still lie ahead.
Like autonomous driving, the technical challenge is immense and the evidentiary bar will be unforgiving. There will be many contenders, but only a handful will survive. Healthcare has never been easy to transform. For those who succeed, however, the prize will be extraordinary.