The Unbundling of the PBM

This isn’t just about pharma pricing. It’s about what software will eat next

Thanks for opening. And welcome to new subscribers. Here’s what I have for you this week. If you don’t want to read the essay, scroll down to some of the things I found.

The Big Idea

• Unbundling of the PBM. I saw lots of change happening in this area, so I thought I’d unpack it a bit. I hope you find it interesting.

Ideas and Signals | Important and interesting things I found

• Maybe AI and Humans shouldn’t work together.

• Is the review paper dead?

• The future of AI interfaces in healthcare

Interesting News

• The billion dollar scribe

• Dr. Prasad goes to Washington. Then leaves.

• Making health data great again.

• “I’m a therapist. ChatGPT is eerily effective”

---

The Big Idea | The Unbundling of the PBM

Everyone pays for them but few of us understand them. Pharmacy Benefit Managers (PBMs) have ruled the drug pricing world for decades, but their grip is slipping.

What is a Pharmacy Benefit Manager?

PBMs are middlemen between drug manufacturers and insurers (or employers). They sit at the center of the drug supply chain, deciding what medicines people can get and what they’ll pay. As a vertically integrated industry, they run most elements of the drug supply chain, often operating their own plans and pharmacies.

PBMs wear two hats in how we get our meds. First, they act as intermediaries between drug manufacturers and insurers, negotiating rebates and drug placement on formularies. Second, they function as suppliers to insurers, who then serve the end consumer. Market concentration gives PBMs leverage to negotiate bigger rebates. But those savings often don’t trickle down. Even if passed to insurers, there’s no guarantee they reach patients. The result: PBMs may lower costs on paper, but the benefits rarely show up at the pharmacy counter.

PBMs have become extractive gatekeepers controlling access and affordability with zero accountability to the public. They now manage nearly 80 percent of all prescriptions filled in the United States.

With that control comes real impact: These practices contribute to U.S. drug prices that can be more than double those of identical drugs in other high-income countries. Nearly 1 in 3 Americans say they’ve skipped or rationed medications due to cost.

The Spread Pricing Racket | A page from the PBM Playbook

Spread pricing is when a pharmacy benefit manager (PBM) charges the health plan more for a drug than it pays the pharmacy, and keeps the difference—the “spread”—as profit.

Here’s the breakdown:

1. The PBM pays the pharmacy $100 for a med

2. It then bills the employer or insurer $130

3. The $30 difference is the spread — and the PBM pockets it

What once made sense has overstayed its welcome

At one point, PBMs served a purpose. They were designed to help health insurers contain drug spending. And for decades, the healthcare industry relied on PBM contracts for simplicity and efficiency—a one-stop shop under a single agreement.

But that model is under pressure.

Rising drug costs, regulatory scrutiny, and growing consumer expectations have led many to question whether PBMs are still up to the job. And if someone can do it better, PBMs will be displaced.

That’s starting to happen. Technology is quietly mediating the role of the PBM, and the unbundling is underway.

Where the Unbundling Has Already Begun

• Major insurers are going direct. In an early declaration of independence, Blue Shield of California made headlines by dropping CVS Caremark and assembling its own pharmacy supply chain using Amazon Pharmacy, Mark Cuban’s Cost Plus Drugs (see below), and Abarca. (For a brilliant argument for why health plans should negotiate directly with pharma, read this Healthcare Dive by Kathy Chang, Head of Trade Relations at Blue Shield of California.)

• Looking for a la carte services. A recent survey from Pharmacy Strategies Group (PSG) found that nearly one in three organizations are evaluating unbundled PBM models. These allow plan sponsors to separate services and choose partners based on performance and value—giving stakeholders more control.

• Retail giants are rethinking the model. Walmart and Amazon have made strategic bets by offering simplified pricing and bypassing traditional intermediaries.

• States are pushing back. Ohio and West Virginia removed PBMs from managing their Medicaid drug programs due to spread pricing abuses and saved tens of millions by bringing the work in-house.

• Regulatory pressure is accelerating unbundling. A U.S. federal judge recently blocked an Arkansas law that would ban companies from owning both a PBM and a pharmacy—raising deeper questions about vertical integration. Meanwhile, bipartisan scrutiny and FTC investigations are forcing PBMs to defend their role in rising costs and limited access.

The New Model Is Here

We’re seeing early signs of disintermediation. Employers are experimenting with direct-to-manufacturer deals powered by better data and AI-driven utilization modeling. Startups are using software to rebuild the pharmacy stack with flat fees and transparent contracts.

Two standouts: Cost Plus and Capital Rx.

Mark Cuban and the Evisceration of the Old Model

Mark Cuban’s Cost Plus Drugs is driving a full-frontal assault on the drug pricing maze that PBMs intentionally created. Instead of hiding costs behind rebates, middlemen, and opaque negotiations, Cost Plus offers a dead-simple model: the cost of the drug, a 15% markup, a pharmacy fee, and shipping. That’s it. No spread pricing. No rebates. No more ‘hide the weenie.’

This isn’t just Cuban delivering cheaper meds—it’s a dismantling of the entire pharmacy benefit structure. By stripping away the bureaucratic layers PBMs added, Cost Plus turns drug pricing into something predictable and transparent. It’s a software-enabled supply chain wrapped in a consumer-friendly experience. And it sends a message: if we can reprice generics with a spreadsheet and a Shopify site, maybe the whole PBM edifice was just complexity theater.

Capital Rx | Like Uber for Pills

Another example of new intermediaries is Capital Rx. They’re reshaping pharmacy benefits from the ground up with technology. At the core is its Clearinghouse Model®—a real-time pricing system that anchors drug costs to a national index (think: how airline tickets or commodities are priced). This eliminates the hidden markups, inconsistent rates, and opaque rebate deals that have traditionally polluted PBM contracts.

Instead of relying on spread pricing or complex negotiations, Capital Rx offers fixed pricing to employers and health plans and a platform for managing things. And it’s all transparent. It’s not just unbundling the PBM—it’s rebuilding it as a software layer. Their pitch is beautiful: if you can make drug pricing work like e-commerce, why tolerate slippery middlemen?

Capital Rx is effectively a platform, a little like Uber. It connects payers (employers, health plans) with pharmacies and pricing infrastructure. It uses software to intermediate these relationships, replacing messy human negotiations and paperwork. It relies on real-time data to match buyers with transparent pricing—much like Uber matches riders with drivers and fare estimates.

————

The Operating System is Being Rewritten

We operate in a system flush with exclusive protections and stale anticompetitive practices. As it turns out, PBMs weren’t necessarily the root of the problem. They were the operating system for a dysfunctional drug pricing market. But now that the market is becoming legible to software, that OS is being rewritten.

Technology isn’t attacking PBMs directly—it’s eroding the complexity moat they sat behind.

And PBM isn’t dying all at once. It’s being taken apart brick by brick. Employers, states, health systems, and startups are realizing they don’t need the whole PBM ... just the pieces: contracting, claims, formulary management, pricing intelligence. And they’re finding smarter, more aligned ways to get those pieces.

---

Ideas & Signals

Maybe AI and humans shouldnt work together

A provocative new perspective from Rajpurkar and Eric Topol challenges the traditional “AI-assist” model in radiology, where machines simply support humans. Instead, they argue for role separation giving AI and radiologists distinct responsibilities based on their strengths.

🔺Whether AI goes first, doctors lead the process, or cases are dynamically routed based on complexity, early evidence suggests clearer boundaries may improve accuracy and reduce both overreliance and neglect.

As diagnostic AI grows more capable, the future may be orchestration rather than collaboration. | Link

Is the review paper dead?

This LinkedIn post from Dr. Gary Nakhuda got me thinking. Long a cornerstone of medical publishing systematic reviews may be the next casualty of AI. Just a year ago, trusting a language model to synthesize literature might have seemed crazy. But tools like OpenEvidence, Elicit, and even ChatGPT now produce amazing summaries. Dr. Nakhuda recently compared an AI-generated review of AI in assisted reproduction to a human-written paper on the same topic. The AI’s version? Cleaner, more complete, and more publishable. We may soon need to find a new academic chore for the next generation.

The future of AI interfaces in healthcare

This article got me thinking about how AI interfaces might evolve.🤔 Right now it’s just a blank search box. But how could the front page interface evolve in healthcare? My thoughts on LinkedIn:

1️⃣ Type 1 — Conversational Interfaces
This is where most AI in healthcare is today: chatbots, GPT wrappers for documentation. Standalones with no context...These systems are adjacent to care, not part of it.

2️⃣ Type 2 — Co-inhabited Interfaces
LLMs embedded into the EHR or ambient scribe tools that understand what you’re doing and where you are in the workflow. A Type 2 scribe might “know” you’re in a follow-up visit for asthma and that you tend to write long patient instructions, so it adjusts. This offers huge promise for reducing clinical friction.

3️⃣ Type 3 — Generative Interfaces
This is where it gets crazy😳. Imagine an EHR that reshapes itself based on your past documentation, your patient panel, even your tone of voice. Type 3 systems could triage before triage and feel like a clinical partner.
Somebody build this with/for me....👆

Interesting news

The billion dollar scribe

Ambience Healthcare, a leading AI scribe company that helps with clinical notes and billing announced raising an additional $243 million infusion putting its valuation at over a billion — all AI scribe funding announcements for the year collectively approach nearly $1 billion. Here is a summary of capital raised this year:

Alot of eyebrows are being raised about a commodity platforms reaching these levels. It's beginning to smell like the dotcom around here. | Link

Dr. Prasad goes to Washington. Then leaves

In last week's letter I suggested that the single arm clinical trial is dead. Well, apparently it’s back on. Vinay Prasad, the controversial head of the Food and Drug Administration office that oversees vaccines and gene therapy abruptly left the agency after less than three months on the job. From StatNews, “In the end, Prasad wasn’t able to outrun his firebrand persona. His fearlessness is what brought him into Makary’s orbit, and ultimately to the FDA. But it also seemed to be his undoing.”

Lots about Prasad's fugue state. Alot of it suggesting that we might be worse off after his departure.

• Stat | Are we sure Vinay Prasad’s ouster is positive for biotech?

• Stat | Vinay Prasad was not good for the FDA, but his ouster makes things worse.

• Science/The Pipeline | Vinay Prasad: That was fast

Making health data great again

President Trump announced the development of a health care record system that he said would allow Americans to more easily and broadly share their personal health information with health care providers. It's powered by partnerships with Google, Amazon, Apple, and OpenAI, and overseen by CMS. I would love to help you understand this but the deets are paltry. And where are the hospitals in this discussion? How about patients? And Epic was apparently not available for comment. | Link

"I'm a therapist. ChatGPT is eerily effective"

This New York Times essay by a therapist got some traction.

One day, I wrote to it about my father, who died more than 55 years ago. I typed, “The space he occupied in my mind still feels full.” ChatGPT replied, “Some absences keep their shape.” That line stopped me. Not because it was brilliant, but because it was uncannily close to something I hadn’t quite found words for. It felt as if ChatGPT was holding up a mirror and a candle: just enough reflection to recognize myself, just enough light to see where I was headed.

Conversational AI is very much a mirror, as this quote suggests. And I suspect that that This essay captures it nicely. | Link

And by the way, Sam Altman, CEO of OpenAI has made it clear that your personal therapeutic discussions are not subject to any level of privacy protection.

---

Thanks again, Bryan