Mark Cuban’s Plan to Make Healthcare Affordable

Breaking the Economics of Healthcare: Transparency, Direct Contracts, and AI

Mark Cuban argues that healthcare is not an efficient market. The economics are broken, and no amount of technology layered onto the existing structure will fix that on its own. Cost Plus Drugs, he says, has already forced change in incumbents that policy and decades of market pressure never did, and much of that pressure comes from the simple fact that prices are public and anyone can pull the data through the API.

Cost Plus Drugs and the Expanding Business

The company launched in January 2022 with 111 purely generic medications. Demand grew quickly, manufacturers expanded their portfolios, and the catalog now runs to roughly 4,000 SKUs. The growth curve is “up and to the right,” and keeping up with it is the central operational challenge.

Cuban describes a discount card that lets patients pick up prescriptions at roughly 17,000 affiliated independent pharmacies. The pharmacy fill fee of $12 makes the cheapest generics more expensive than mail order, but for specialty generics, biosimilars, insulin, and cold-chain products, local pickup is cheaper.

Manufacturing: From Dallas to Modular Pods

The Dallas facility was built around generics on the FDA shortage list, including pediatric cancer drugs and hospital staples like oxytocin and epinephrine. Using AI and robotics, the team compressed what would have been a football field of equipment into roughly 5,000 square feet, and the line can turn over to a new drug every four hours. Products are sold to hospitals and clinics through Cost Plus Marketplace at cost plus 10 to 12 percent. Cuban says the team can save hospitals millions a year, though the harder problem is convincing purchasing departments even after the CEOs are on board.

The more ambitious piece is a manufacturing unit about the size of a tractor trailer, which Cuban calls a pod. Each costs around $20 million, can be configured for sterile injectables or tablets, and can be daisy-chained for scale. The government is the first customer, with interest in disaster response and military deployment. Cuban is most enthusiastic about configuring pods for N-of-one therapies, shipping them to research hospitals so that individualized treatments can be produced and iterated quickly.

Direct Contracts and Radical Transparency

Cuban’s approach to employers is built on direct contracting. He offers hospitals a deal: he will trust the physician’s judgment and skip prior authorizations and denials for anything that is not clearly fraudulent, and he takes responsibility for patient out-of-pocket costs, so the hospital does not have to act as a subprime lender collecting deductibles before care. In exchange he wants a better price, reflecting lower administrative overhead, and the right to publish the contract on cost plus wellness so that it becomes an open template any employer or provider can use.

Julie Yoo frames this against the scale of employer-sponsored coverage, which covers well over half the country, and against the entrenched national carriers that have made it nearly impossible for startups to enter. Cuban adds the structural problem: UnitedHealth’s 3,000-plus subsidiaries and roughly $150 billion in intercompany transfers mean that premium increases, deductible increases, and redirection to affiliated providers can all happen at once, often without the employer knowing the affiliation exists. Third-party administrators owned by the big insurers, known as ASOs, make the picture worse. Employers rarely see the terms of the relationships their plans depend on, and the 835 and 837 remittance files that would allow reconciliation are often inaccessible even for an employer’s own claims. Cuban argues spread pricing on the medical side is far worse than on the pharmacy side, and that the medical side is the much larger share of spend.

Why AI Alone Cannot Fix It

Cuban is a strong AI proponent but insists that AI cannot repair a broken financial system. The core issue is who takes the risk. Contracts run to hundreds or thousands of pages, and the insurers’ legal and contractual complexity, which he describes as contractual terrorism, outpaces any team’s ability to read everything. Once a contract is signed, he argues, AI cannot protect the buyer from terms the buyer did not understand. The lowest-hanging fruit is therefore to use AI before signing.

He is also skeptical that AI will produce lasting change through the claims arms race. Upcoding and downcoding between payers and providers is short-term, and because AI is an equalizer, the advantages on each side tend to net out. He does argue that the insurers, having grown by acquisition into siloed data estates, will struggle to become AI-native companies over the long run.

His concrete recommendation is a single prompt run against every benefits contract, broker agreement, TPA contract, and PBM agreement: where are they screwing me over, and what should I do about it? He describes running contracts through several LLMs in parallel, using agents to file open-records requests against public entities, and using the results to expose how school districts, cities, and unions are being overcharged. If the employers representing a large share of the insured population did this before signing, he argues, the cost structure of the system would collapse. He ties it to a broader point about governance: companies should hire a healthcare CEO or CFO whose sole job is managing contracts, rather than waiting for legislation.

Trust as the Core Product

Cuban is clear that Cost Plus Drugs does not sell drugs so much as trust. The company spends nothing on marketing, and its pricing model is fully transparent with a fixed markup. Early critics predicted the model would fail, but he argues that customers who save money tell others, and trust accumulates from there. He sees trust as the missing element in healthcare economics, and believes a transparent markup will protect the business against competitors who undercut on percentage fees.

His longer-term vision is that pods can scale to thousands of units, with real margin in each one, and that domestic tablet production could undercut Indian manufacturing even after paying for better manufacturers. The constraint is regulatory: each new generic requires a new drug application costing about $365,000 in fees, so onshoring at scale has a substantial fixed cost. He expects a U.S.-made label on the site within a couple of years once FDA work is complete.

He also distinguishes regulatory capture from scale capture, arguing that the current problem in healthcare economics is scale capture. Vertically integrated companies talk about technology in their earnings calls, but they do not say it will lower costs for employers and patients. Cuban says he does not deal with the large insurers directly because doing so would pull him back into the structures he is trying to avoid.

Workforce, Literacy, and the Executive Problem

On talent, Cuban rejects the idea that entry-level positions can simply be eliminated. Every model is already behind the moment it ships, so organizations need people who understand context, can keep humans in the loop, and have watched the implementation history evolve. AI does not replace jobs, he says; people who know how to use AI replace people who do not.

He sees the more pressing gap at the executive level. Many C-suites do not understand how models, agents, open versus closed weights, or latency work, and that gap is where organizations stall. He recounts asking an LLM to design a GLP-1-friendly competitor to a food company’s flagship cookie, complete with a bill of materials and sourcing, and getting an answer in about three minutes that the executive had not expected.

He also points to a consumer opportunity that he thinks is underexploited. Roughly 40 percent of Americans cannot cover a $400 emergency bill, and nonprofit hospitals are required to offer financial assistance programs, some very generous and reaching up to 600 percent of the federal poverty level. Cuban argues that consumers do not know these programs exist and that a good AI assistant could navigate them. He says hospitals are poor at marketing and are overly dependent on insurer networks for patient flow, and he suggests that building agents to help patients use financial assistance would be a valuable business.

Cuban compares the current fear of AI to the early resistance to spreadsheets, when people insisted on adding machines and paid hundreds of dollars for Lotus 1-2-3. Agents also face a structural gap: they cannot see social media posts or ads easily, which raises questions about how commerce will be communicated to agent-driven consumers.

Personalized Medicine and the Future of Doctors

Cuban expects the cost of designing a drug personalized to an individual, based on DNA and the full body of medical knowledge, to fall toward zero as model capability and algorithms improve. He imagines that process starting with over-the-counter drugs, where individual responses are notoriously variable, and eventually reaching prescription therapies where clinicians currently choose among near-equivalent options with little rationale. He compares the eventual subscription model to consumer DNA testing and expects personalized medicine to become part of a normal subscription once the marginal cost is low.

Julie Yoo notes that this shift would be disruptive to care delivery. Much of what physicians do today is helping patients make drugs work for them, titrating doses and changing behavior, and personalized medicine would reduce much of that work. Cuban agrees that the physician’s role will change but argues that doctors will remain essential because of latency in models and the need for human judgment, presence, and the ability to read context. Medical education will need to teach doctors how to extract and translate information from models, and he warns that IP ownership will complicate the picture, since a single individualized model may be patented or licensed in ways that make access expensive.

Yoo argues that the role of physicians may shift toward physician-scientists who collect novel data and generate insights at the individual level, and that as advice becomes more abundant, demand for real-world care could rise. Cuban adds that longitudinal monitoring via wearables will replace yearly blood pressure readings with continuous data, and that translators who connect novel individual data to AI outputs will become a new job category.

Policy Prescriptions and the AI Industry

Asked what he would do as president or adviser, Cuban proposes a set of concrete steps: make all contracts with taxpayer-funded entities public without redaction, let consumers apply savings from better prices toward their deductibles (eight states already do something similar), require nonprofit hospitals to publish their general ledgers and procurement records, and reconsider the insurance model entirely. He argues that roughly 72 percent of major carriers’ revenue comes from taxpayers, and that the insurance premium is the one debt most people never pay off. His alternative is a deposit-based program in which individuals fund a savings account matching a silver ACA plan premium, with bank-backed loans guaranteed by government and stop-loss coverage above a threshold. Money would be spent under defined constraints, removing the insurer as the decision-maker.

On AI, Cuban argues that the industry’s public relations problem stems from social media algorithms, which shape what people see and believe. He sees data center opposition as a proxy for a broader failure to understand AI, and he argues that AI companies need to become part of people’s feeds rather than publishing position papers. He also expects that communities which negotiated poorly for data centers, like employers that negotiated poorly on healthcare, have fallen for the same kind of contractual terrorism.

Advice for Founders

Cuban’s closing advice is that AI will sort companies into those that are excellent at using it and those that go out of business. The more important divide is between people who use AI to learn and people who use it to avoid learning. He encourages founders of all ages to talk to their AI, compare outputs across models, and check the results, and he argues that the discomfort of using AI for unfamiliar tasks is often the path to something groundbreaking.

• Tags:
• AI Agents
• Personalized Medicine
• Feature
• Founder Advice
• Healthcare Economics
• Cost Plus Drugs
• Direct Contracting
• Pharmacy Benefit Managers
• Self-Insured Employers
• Drug Manufacturing
• Healthcare Transparency