Mark Cuban on Healthcare Transparency, AI, and Why Trust Is the Product

Julie Yoo opened by calling healthcare an inefficient market with broken economics. She said one of its many ills is that it completely lacks competitive dynamics and free-market forces.

Yoo argued that the mere presence of Cost Plus Drugs has forced incumbents to change, something she said has been intimidating to avoid for many decades. She framed Cuban's willingness to enter the field as the reason she invited him onto the show.

Cuban described UnitedHealth as having more than 3,000 subsidiaries and about $150 billion in intercompany transfers annually. He said the company often increases premiums and deductibles while steering patients into relationships with companies it owns, and that patients often do not know about that ownership.

Cuban said his ultimate mission is to disrupt healthcare so that it becomes affordable again.

Cuban said his co-founder Dr. Oxmanski's original cold email proposed manufacturing generic drugs on the FDA shortage list. These include pediatric cancer drugs and widely used hospital medicines such as pitocin and epinephrine.

Cuban said AI and robotics shrank what would have been a football field or more of production into 5,000 square feet, and that the line can turn drugs every four hours. A six-month supply can be made before the line moves to the next drug.

The team has also built pods about the size of a tractor trailer that can be shipped anywhere and configured for specific drugs. Cuban said the government was the first customer, that a second order has been received, and that disaster response such as FEMA is a possible use.

Cuban said the pods cost about $20 million and could be configured for sterile injectables. Multiple pods can be linked together to scale production.

Cuban said Cost Plus Marketplace takes sterile injectables the company makes, along with products sourced from other manufacturers, and sells them at cost plus 10 or cost plus 12 percent to clinics, hospitals and providers. He said the company can save hospitals millions of dollars a year.

He said the biggest challenge is that the company can convince CEOs but cannot always convince purchasing staff to adopt it.

Cuban said the price a self-insured employer pays rarely matches the amount charged or the amount paid to the hospital, creating what he called spread pricing. He said administrative services only (ASO) arrangements owned by the carriers make it impossible to reconcile the numbers and see whether the employer is paying more or less.

He pointed to 835 and 837 files, which carry the data sent to hospitals and the payments made in response to claims. He said employers have no access to them and often cannot access their own claims.

Cuban said removing these intermediaries is the key to what is happening in healthcare now. He also described his own experience as a self-insured employer and his talks with hospital CEOs and CFOs about their relationships with large vertically integrated carriers.

He said he proposed to hospitals a direct contract in which he would trust the doctor and avoid pre-authorization and denials, and pay the contracted price on time, in exchange for terms they wanted. Cuban said hospitals dislike delays, denials, late payments, clawbacks and rising deductibles that turn providers into subprime lenders.

Yoo said the current bot war between payers and providers is a case of AI implementing the existing fee-for-service billing system to its full extent, showing how broken that system is. She asked whether direct contracting that reduces information asymmetry is what will be needed.

Cuban called the claims upcoding and downcoding fight a short-term play. He said AI is a great equalizer, so competing AI systems will net out at some level, and he did not see dramatic long-term change from that fight.

Cuban said insurers that grew through acquisitions may have a problem, and that their data may remain siloed across thousands of subsidiaries. He said the change has to begin with employers, who can run their existing benefit contracts through AI to see where they are being overcharged.

Cuban said he uses a prompt that asks a single question: where are they ripping me off, and what should I do about it. He said results appear within about 30 seconds when he runs a contract through AI.

He described using several AI agents to locate every state entity or organization with an open-records opportunity, find contracts and RFPs, and draft a records request. He said he personalizes each draft and then sends it.

Cuban said the documents go into a shared Google Drive that multiple language models can access. When new files load, he asks the models for good news, bad news, recommended actions and a simple summary he can send to a company or school.

He said the result is a view of how large PBMs and ASOs are signing contracts with school districts and cities from Denver to Dallas and Fort Worth.

Cuban said the ultimate mission is to disrupt healthcare so it becomes affordable again. He said he is a fan of universal healthcare, but that it cannot be reached until people have correct data and can make decisions based on it.

Cuban said trust is still the company's primary product. Cost Plus spends nothing on marketing or advertising, and he expects customers to tell others how much they saved. He said a fixed 15 percent markup is meant to earn that trust through full transparency.

Cuban projected that the company could sell a couple hundred pods a year within two years and a couple thousand a year within five. He said there is good margin on pods that sell for about $20 million each.

Cuban said the healthcare workforce looks the same as any other industry, and that companies still need to hire entry-level staff. He said every model falls behind the moment it is released, so people are needed who understand context and consequences, work with models and act as the human in the loop.

He said AI does not replace jobs, but people who know how to use AI replace the jobs of people who do not. He said the first step is to teach executives how AI works, including agents, models, the difference between open-source and closed models, and model latency.

Cuban said he tested this by having ChatGPT design a GLP-1-friendly competitor to a food company's flagship cookie, including a full bill of materials. He said the answer came in about three minutes.

Cuban said there is no better time to be an entrepreneur than right now, and predicted there will be a trillionaire in the basement building an agent swarm on an entrepreneurial basis.

He said agents cannot see social media posts or watch ads, and transcripts are hard to get in any material amount. If a future agent-driven economy develops, he said commerce will need to change how it communicates with those agents.

Cuban demonstrated agentfeeds.dev, which lets users paste the transcript of a post so followers' agents can pull updates. He said it is not truly active yet and that trust still has to be addressed before an explosion of such tools.

Cuban said the cost of designing a drug personalized to a person, after reviewing their DNA, will start approaching zero as the capabilities of models go up. He said algorithms, not GPU speed, will change everything and make the process more efficient, though the speed of algorithmic improvement is unknown.

He said personalization could start with over-the-counter drugs such as Advil, Tylenol and ibuprofen, since people respond differently to them. Cuban said a personal dose could be identified without much trial and error, and that the same logic could later apply to prescription drugs such as Eliquis, Xarelto and warfarin.

Yoo argued that much of healthcare is the job of doctors helping patients figure out why a drug does not work for them, calibrating and titrating doses, and changing behavior so life fits the drug. She said better individualized drugs could be among the most disruptive forces for care delivery.

Cuban countered that the latency of models means doctors are still needed if he tried a new personalized medicine and became ill. He said he would rather have a human who can use all five senses, look, ask questions and notice a crying mother or child than rely on telehealth or a robot.

Cuban said the old norm of publish or perish has given way to patents being copied with small changes. He said the IP needed to train, post-train or retrieve for a model may not be made available, even though it is required.

Yoo responded that the role of the doctor is changing. She said demand for real healthcare services may increase as AI advice becomes more abundant, and that physicians could become more like physician scientists collecting novel data about how individuals react.

Cuban said he does not know whether the electronic health record market will be disrupted. He said the EHR market is dominated by Epic and that it is Epic's world.

He said his company partners with Epic to offer pharmaceutical pricing at the point of prescription, where doctors can offer patients different options. He called that a valuable step even though he is not qualified to judge where Epic's business is disruptible.

The hosts discussed whether healthcare spending will remain a fifth or sixth of GDP. Cuban said that if people are healthier, they will be more productive, which could more than account for the spending.

Cuban argued that insurance premiums are a debt people never pay off until they die, and that in current systems people must fight for coverage when they need it. He proposed a program in which participants deposit what they would pay as a premium under an ACA silver plan in their state.

He gave an example of $2,100 for a family of five in Texas. If a medical or pharmaceutical expense exceeds the money in the account, a bank would loan funds up to the point where stop-loss coverage applies, which he gave as about $40,000. Part of the deposit would fund stop-loss coverage, and the federal or state government would guarantee the loan so providers are paid.

Cuban said the loan would be repaid through payroll deduction. If a person never has a claim, they would get the full balance and accrued interest at age 65, when Medicare begins. He compared the guarantee to existing federal programs for student loans, SBA loans and home loans, and said such a program would end the role of insurance companies because the money belongs to the individual.

Cuban said spending would need limits, such as not paying a relative for a toe fix.

Cuban said there will be two types of companies: those great at AI and those that go out of business. He said there are also two types of users, those who use AI to learn anything and those who use it so they do not have to learn anything.

He said the biggest challenge for entrepreneurs, including 50 to 70-year-olds, is comfort with talking to AI by voice or typing. He compared the change to early spreadsheets, when people checked the results on a 10-key adding machine, and said output should still be checked.

Cuban said the hardest part is simply going for it, and that this is where Silicon Valley has an advantage because its culture encourages it. He called this the greatest transitional technology he has seen.