The Two Pictures
Two things are true about medicine in 2026, and they do not sit easily together.
The first: science is moving faster than at any point in human history. Personalized mRNA cancer vaccines have produced sustained 49% reductions in melanoma recurrence in trials, now in Phase 3 with results expected later this year. CRISPR-based gene editing is curing diseases that were uncurable a decade ago. AI systems are designing new drug molecules in months rather than years, with the first AI-designed cancer drugs entering human trials in 2026. The horizon is genuinely vertiginous.
The second: most of this reaches almost no one. Casgevy, the first FDA-approved CRISPR-based cure for sickle cell disease, carries a $2.2 million list price. Approximately 16,000 Americans are eligible. More than two years after approval, fewer than a hundred patients globally have been treated. The GLP-1 drugs that have transformed weight management cost $936 to $1,349 per month in the United States — roughly ten times what they cost in Europe — while a recent study estimated that the underlying medication can be manufactured for less than $5 per month.
This section tracks both pictures honestly. It also tracks the question that connects them: when productive capacity in medicine increases this fast, what determines whether the abundance reaches the people who paid for it?
The Discovery Engine
For most of the modern pharmaceutical era, drug discovery has been slow and expensive. Twelve to fifteen years from candidate molecule to approved drug. Hundreds of millions of dollars per program. A failure rate above ninety percent in early trials. The bottleneck was always the same: finding molecules that bind to the right protein targets without unintended consequences took an enormous amount of trial-and-error laboratory work.
That bottleneck is breaking.
AlphaFold, developed by Google DeepMind, made it possible to predict the three-dimensional structure of essentially any protein from its genetic sequence. The 2024 Nobel Prize in Chemistry recognized this work. AlphaFold 3, released later that year, extended these predictions to interactions between proteins and small molecules — precisely the prediction that drug design needs. Isomorphic Labs, the DeepMind spin-out commercializing this technology, raised $600 million in 2025 to advance AI-designed drugs into human trials, with oncology candidates leading the pipeline.
Other companies are pursuing the same compression from different angles. Insilico Medicine used AI to identify a candidate drug for lung fibrosis and advanced it to human trials in approximately eighteen months — a process that traditionally takes four to six years. Recursion Pharmaceuticals uses automated high-throughput imaging combined with deep learning to identify cellular responses to candidate molecules. Relay Therapeutics has an AI-discovered cancer drug currently in Phase 3 trials for breast cancer.
The picture in 2026: drug discovery is being compressed by roughly an order of magnitude in its early stages. The wet-lab work, the trials, the manufacturing, the regulatory process — these still take years. But the front end, the part that used to be hardest, has been transformed.
The question of who owns what comes out of this engine is one of the defining political questions of the coming decade. Most of the foundational structural-biology data on which AlphaFold was trained was generated with public funding, accumulated over decades in databases like the Protein Data Bank. AlphaFold itself was developed at a research lab owned by Alphabet. The molecules that emerge from these tools will be patented by the private companies that develop them. The pattern that has characterized the pharmaceutical industry for half a century — public research as the upstream input, private profit as the downstream output — is now accelerating in scale and speed.
What Is Working Now
Before the horizon, the present.
Immunotherapy has changed cancer treatment dramatically in the past decade. The basic mechanism: turn the body's own immune system against cancer by removing the “brakes” that cancer uses to hide. Checkpoint inhibitor drugs — pembrolizumab (Keytruda), nivolumab (Opdivo), and others — are now standard treatments for melanoma, certain lung cancers, kidney cancer, bladder cancer, and a growing list. Some patients who would have died within months a decade ago are alive years later.
CAR-T cell therapy is more dramatic and more limited. Doctors remove the patient's own immune cells, genetically modify them to recognize cancer, grow them in the lab, and reinfuse them. For certain blood cancers — some leukemias and lymphomas — CAR-T has produced durable remissions in patients who failed every other treatment. The technology exists. The list price runs $400,000 to $600,000 for the treatment itself, and total costs can exceed a million dollars when complications are included. Even in the United States, where these therapies are approved and generally covered, only about two in ten eligible patients with diffuse large B-cell lymphoma actually receive them.
Targeted therapies attack specific molecular features of individual tumors. HER2 inhibitors, BRAF inhibitors, EGFR inhibitors, PARP inhibitors — the list is long now. Molecular testing of tumors has become standard at good cancer centers. A patient with lung cancer in 2026 may have their tumor tested for ten or more genetic markers before treatment decisions are made.
GLP-1 agonists — the class of drugs that includes Ozempic, Wegovy, Mounjaro, and Zepbound — have become a story of their own. Originally developed for type 2 diabetes, they have transformed weight management. Roughly twelve percent of Americans, around forty million people, have used one. The drugs work. They are also a case study in how the American healthcare system captures medical productivity into rents.
The list price of Ozempic in the United States is $936 per month. In Japan, $169. In the United Kingdom, $93. In France, $71. In Germany, $59. A 2024 study published in JAMA Network Open estimated that semaglutide — the active ingredient in Ozempic and Wegovy — could be manufactured for between eighty-nine cents and $4.73 per month, including a profit margin. The Senate Health, Education, Labor and Pensions Committee opened an investigation. A 2025 deal between the Trump administration and the major manufacturers lowered prices for cash-paying patients to roughly $350 per month through a new direct-to-consumer platform. The differential with European prices remains roughly threefold.
The thread running through these examples: medical technology is producing things that are real and transformative. The capture of that productivity by the system that delivers it is also real, and is the place where the falling-costs argument breaks down.
The Approved Cure That Almost No One Receives
The case that distills all of this most starkly is Casgevy.
Casgevy, developed by Vertex Pharmaceuticals and CRISPR Therapeutics, became the first CRISPR-based gene therapy ever approved by the FDA when it received approval in December 2023 for the treatment of sickle cell disease. The science is extraordinary. Patients' own blood stem cells are extracted, edited using CRISPR/Cas9 to reactivate fetal hemoglobin production, and reinfused. In trials, twenty-nine of thirty-one treated patients had no severe pain crises for at least twelve consecutive months in a twenty-four-month follow-up period. For a disease that has caused unmeasurable suffering for generations, that is a genuine cure.
The list price is $2.2 million per treatment. Bluebird Bio's competing therapy, Lyfgenia, also approved the same day, was priced at $3.1 million.
Approximately sixteen thousand Americans are eligible. More than two years after approval, fewer than a hundred patients globally have been treated. The bottleneck, as reported by STAT News in February 2026, is not principally the cost — Medicare and most commercial insurers will cover the treatment with prior authorization. The bottleneck is operational. The treatment requires extracting stem cells, modifying them in specialized facilities, and administering them after high-dose chemotherapy to prepare the bone marrow. Many patients cannot produce enough stem cells. Many cannot tolerate the conditioning. Many do not live near one of the small number of authorized treatment centers.
Sickle cell disease disproportionately affects Black Americans, and the global majority of sickle cell patients live in sub-Saharan Africa. The disease has been described in the medical literature since 1910. Hydroxyurea, the standard-of-care drug that reduces complications, was approved by the FDA in 1998 and remains unavailable to most patients globally. A modest program to provide hydroxyurea at roughly $67 per patient per year across sub-Saharan Africa would cost less than $100 million annually — a fraction of the cost of one Casgevy patient's treatment.
This is not an indictment of Casgevy. The science is real, the cure is real, and the patients who have received it have been transformed. It is an indictment of a system that produces miracles and then prices them out of the reach of the people who most need them — and whose unmet need was the moral case for developing the miracle in the first place.
The Horizon
The pipeline is full. A partial map of what is plausibly coming in the next several years.
Personalized mRNA cancer vaccines. Moderna and Merck's combination of an individualized mRNA vaccine with the checkpoint inhibitor Keytruda has shown sustained forty-nine percent reductions in melanoma recurrence in five-year Phase 2 follow-up data published in January 2026. The Phase 3 trial is fully enrolled. Results are expected later in 2026. Similar approaches are being tested for lung, kidney, and bladder cancers. If approved, this would be the first mRNA cancer vaccine in clinical use.
AI-designed drugs entering human trials. Isomorphic Labs is preparing to dose its first patients in 2026 with cancer drugs designed using AlphaFold-derived methods. Other AI-led programs are at earlier stages across oncology, infectious disease, and rare disease.
Next-generation gene editing. Companies including Beam Therapeutics are developing base editing — a CRISPR refinement that changes single DNA letters without cutting the genetic strand. In vivo editing approaches — where the editing happens inside the body, without removing cells — are in early trials. If these work, they would dramatically lower the cost and complexity of gene therapies.
Psychedelic medicine. The FDA's 2024 rejection of Lykos Therapeutics' MDMA-assisted therapy for PTSD was a setback, driven principally by trial design and conduct concerns rather than efficacy questions. Phase 3 trials for psilocybin in treatment-resistant depression are reporting out. In late 2025, the FDA granted priority review vouchers to three psychedelic drug trials, suggesting institutional movement. Whether and how these therapies become available to ordinary patients — given the intensive psychotherapy protocols some require — remains an open structural question.
Ambient AI clinical documentation. AI scribes that listen to doctor-patient conversations and generate clinical notes are now deployed at scale in major US health systems, reducing physician documentation burden and burnout. This is one of the few areas where AI in healthcare is delivering tangible benefit at the point of care today.
Multi-cancer early detection. Blood tests like Grail's Galleri claim to screen for multiple cancers from a single sample. Validation is ongoing. False positives remain a concern. If the technology matures and access widens, cancer screening could shift from organ-by-organ to single-test.
GLP-1s for indications beyond weight and diabetes. Cardiovascular benefit data has already led to expanded use. Trials are ongoing for Alzheimer's, addiction, and other conditions. If the drugs prove useful across a wide range of conditions of aging, the question of their cost and access becomes one of the defining political-economic questions of the next decade.
The honest framing: most of what is in trials will not succeed. Cancer vaccines have a long history of disappointment. Most CRISPR programs will not reach approval. AI-designed drugs are at the beginning of a long process, not the end. But a portion of what is in the pipeline will work, and what works will reshape medicine in ways that are difficult to overstate.
The Discovery-to-Access Gap
A drug discovered today may take ten to fifteen years to reach approval, even with AI shortening some of the early steps. AI does not yet compress the long, expensive trial work in actual patients, the manufacturing scale-up, or the regulatory process. These still take years and cost hundreds of millions of dollars per drug.
After approval, the gap widens further. Insurance coverage. Geographic access. Hospital networks. Physician referral patterns. The patient's ability to take time off work, to travel, to navigate prior authorization, to afford copays. CAR-T is approved. The two-in-ten access ratio is the result.
The most stark example of the gap is not in the science but in the price. The JAMA Network Open study on semaglutide manufacturing costs estimated production at less than five dollars per month, including a profit margin. The list price in the United States is $936. The differential is not principally the cost of development, recouped over the patent life of the drug — Novo Nordisk recovered its R&D investment many times over within the first years of Ozempic's launch. The differential is rent. Rent extracted from American patients because the American healthcare system has no coherent national mechanism to negotiate it down.
European systems do negotiate, and pay roughly one-tenth to one-fifteenth what Americans pay. The Medicare drug price negotiation program, introduced in 2022 and expanding through 2027, may eventually exert downward pressure on a limited number of drugs. The Trump administration's 2025 “Most Favored Nation” deal lowered prices for cash-paying patients but left insured patients largely unaffected. The structural problem — that the United States is the only wealthy nation without a national mechanism for drug price negotiation across the system — remains.
This is the gap that the technological cost curves and the political-economic cost curves do not cross. The science is doing what science does. The economics are doing what economics, in this system, does.
Reading Medical News
A short field guide for evaluating what arrives in your feed.
Relative versus absolute risk. “Cuts risk by fifty percent” almost always means halving the rate of some specific outcome in some specific population. If the baseline rate is four percent, a fifty percent relative reduction brings it to two percent. That matters, but it is not the same as halving your chance of dying.
Surrogate endpoints versus survival. Many trials measure progression-free survival or tumor shrinkage. These are useful, but they do not always translate to people living longer. When you hear “the drug shrank tumors,” ask whether it also extended life.
Phase 1 versus Phase 3. Phase 1 tests safety in small groups. Phase 3 tests effectiveness in hundreds or thousands. Most drugs that look good in Phase 1 fail later. Phase 1 and Phase 2 results are real news but very early news.
Animal studies versus human studies. Most cancer cures in mice never work in people. “Cured cancer in mice” is a step, not a destination.
Approved versus available. A drug being FDA-approved does not mean every patient who needs it can get it. Cost, insurance, geography, qualification criteria, and operational bottlenecks all stand in the way.
Press releases versus peer-reviewed papers. Press releases highlight the best numbers. Peer-reviewed papers, with full methods and data, tell the fuller story. The strongest claims should always be checked against the actual paper.
List price versus net price. Drug companies cite list prices in arguments about value and complain about middlemen capturing the difference. Insurers cite net prices in arguments about cost. The actual amount paid by anyone varies wildly. The amount paid by the system as a whole is the relevant political question.
The Question of Ownership
The pattern that runs through everything on this page is a pattern of value flow. Public dollars fund early-stage research at the National Institutes of Health, at universities, at federally supported laboratories. Public dollars fund the regulatory infrastructure at the FDA. Public dollars, through Medicare and Medicaid, are among the largest purchasers of the resulting drugs. The intellectual property that emerges from this publicly-funded ecosystem, however, is privately owned. The profits flow upward to shareholders.
This is not an accident or a hidden conspiracy. It is the explicit policy framework set by the Bayh-Dole Act of 1980, which permitted universities and other federally-funded researchers to patent and license their discoveries. The rationale was that without commercial incentive, basic research would not be translated into usable products. The framework has produced extraordinary innovation. It has also produced extraordinary capture of public investment by private capital.
The question The People's Share asks is whether this framework is the only available one. Several alternatives have been proposed and, in some cases, partially implemented.
The federal government could retain ownership stakes in drugs developed substantially with public funding, with the resulting revenue flowing to a public dividend or directly reducing patient costs. The framework would require legislation; it would not require nationalization.
Drug prices could be negotiated by Medicare across the board, rather than for a limited number of drugs as under current law. Every other wealthy nation does this.
Compulsory licensing of essential medicines — already permitted under international intellectual property law in narrow circumstances — could be expanded for medicines developed with substantial public investment.
Public manufacturing capacity for essential generics, modeled on initiatives in California and a small number of other states, could ensure availability of basic medicines at production cost rather than rent-extracted prices.
None of these alternatives is technically utopian. All are politically difficult. The technical obstacles to a different ownership framework are smaller than the technical obstacles to designing the drugs themselves.
Whether productive capacity in medicine becomes a commons or remains a private capture is not, in the end, a question for scientists or pharmaceutical executives. It is a political question, and it is being decided — by inaction as much as by action — every legislative session.
What This Section Tracks
The People's Share follows medical developments because the medical sector is where the most consequential questions about democratic ownership of productive capacity are being decided in real time.
What we watch for in this section: trial readouts that change what is possible — the mRNA cancer vaccine Phase 3 results, the gene editing pipeline, the AI-designed therapeutics arriving at trials. Pricing decisions and policy moves that determine who gets access to what works. Cases where the structure of the system has been challenged or rearranged — Medicare negotiations, public manufacturing initiatives, international price reference proposals, compulsory licensing debates. Stories of cures that work but reach almost no one, and what it would take to change that.
The technology is doing its part. Productive capacity in medicine is expanding faster than at any point in the species' history. Whether that capacity arrives at scale for the people who paid for it remains a question whose answer is being written now.