Epic Systems (MyChart)
Summary
Epic became healthcare’s dominant software vendor by compounding one architecture, one culture, and one customer promise for 47 years. Its single Chronicles database connects clinical records, scheduling, billing, MyChart, and hundreds of modules without the seams created by competitors’ acquisitions. The result is perhaps “the most durable software company in history”: 607 customers, only one temporary six-month defection, $5.7 billion of 2024 revenue, and roughly 30–35% EBITDA margins.
The Kaiser Permanente contract converted Epic from a promising roughly $50 million vendor into the industry’s default low-risk choice. Kaiser wanted to split inpatient work to billion-dollar Cerner and ambulatory work to Epic, but Judith Faulkner refused: one integrated system was necessary for patients, operations, and revenue collection. Epic then modeled Kaiser’s transaction load overnight, declined warrants even after Cerner reportedly offered 10% equity, won a project with roughly $400 million accruing to Epic, and reached $500 million of annual revenue by 2007.
HITECH’s $36 billion stimulus did not create Epic’s advantage, but it pulled Epic’s victory forward and locked the installed base in place. EHR adoption rose from 9% of hospitals in 2009 to 95% in 2014 as providers received $44,000–$64,000 per physician and later faced penalties for noncompliance. That favored the expensive, integrated vendor most likely to work, while “meaningful use” also legislated cumbersome workflows and produced digitization without the hoped-for “digital transformation.”
Epic’s organization is engineered for reliability rather than conventional corporate efficiency. It recruits largely from college, tests candidates systematically, trains them in an “Epic way” of coding and communication, makes developers repair their own bugs immediately, and sends employees into clinical settings. Its commandments begin “Do not go public,” “Do not acquire or be acquired,” and “Software must work”; the trade-off is an intense, up-or-out culture where implementations resemble “military-level logistics.”
The company maximizes lifetime customer value by charging less aggressively than its position might allow. It has almost no outbound sales or marketing, often rejects prospective customers, raises incumbent pricing by about 2% annually on average, and can bundle new modules into enterprise relationships before outside point solutions become entrenched. Epic’s governing idea resembles Bezos’s framing that, over the long run, customer and company interests converge: make choosing anything else feel “irrational and irresponsible.”
Epic’s next opportunity is to turn its provider dominance into a healthcare data and automation platform. Cosmos contains anonymized information from 295 million patients and 15 billion encounters, while Epic’s network reaches 325 million patients worldwide and exchanges 20 million records daily through Care Everywhere. Payer automation, pharma research, prior authorization, ambient-AI partners such as Microsoft/Nuance, Abridge, and Suki, and an eventual AI operating system could expand Epic beyond the EHR without displacing its core.
The investor case is extraordinary durability offset by the fact that investors effectively cannot participate. The hosts estimate Epic could merit roughly $100 billion, well above a Forbes-implied $15 billion and conservative software comparisons near $50 billion, yet Faulkner’s trust will prohibit an IPO or sale. International pricing, a possible Sherman Act antitrust event, mandated information access, and an AI-driven interface shift are credible risks, but switching costs, scale, network effects, brand, and process power make outright displacement difficult to imagine.
Deep dive
1. A quiet Wisconsin company became American healthcare’s nervous system
Epic is unusual before examining a single product: no conventional marketing, almost no outbound sales, no discounts or negotiation, no venture capital, no acquisitions, no remote work, and a stated commitment never to go public or be acquired. Prospective customers can be told they are not ready.
The company approached $5.7 billion in 2024 revenue with more than 14,000 employees and over 600 health-system customers. It has lost only one customer, for six months, before that customer returned—an almost unmatched retention record across 47 years.
MyChart is Epic’s visible consumer surface, but the consequential product is the integrated operating infrastructure beneath it. The hosts describe an EHR as too narrow a label: Epic combines the medical record, ERP-like administration, revenue cycle, scheduling, permissions, and applications into a health system’s “central nervous system.”
That position sits inside an American healthcare economy consuming about 18% of GDP. They do not attempt to solve that entire system; their question is how Epic became indispensable within it.
2. Judith Faulkner arrived at computing as the discipline itself was forming
Judith Greenfield was born in 1943 in New Jersey. Her father ran a pharmacy and soda fountain; her mother became a peace activist and led an organization that shared in the 1985 Nobel Peace Prize awarded to Physicians for the Prevention of Nuclear Wars.
Faulkner’s mathematical instinct appeared early. A seventh-grade problem about why a number divisible by three has digits summing to another number divisible by three convinced her that “my future is in math,” leading her to major in mathematics at Dickinson College.
A summer job in the University of Rochester’s particle-physics laboratory required Fortran. Given a manual, she taught herself the language in one week and became one of the lab’s best programmers, later describing programming as being “like a kid playing with clay”—mathematics, language, and art together.
Stanford and Wisconsin moved her graduate-school applications into their newly forming computer-science departments. She chose Wisconsin, never completed the PhD, and never left Madison; the hosts treat Stanford as a genuine sliding door that might have produced a very different company culture.
3. Medicine supplied the problem, but third-party payment made records existential
At Wisconsin, Dr. Warner Slack’s Computers in Medicine course led Faulkner to write an on-call scheduling program. Paid $5 an hour and later $10, she worked so quickly that even the raise generated little income while psychiatry, obstetrics, rehabilitation, inpatient care, and intensive care requested more applications.
Those departments shared a problem: each treated overlapping patients without visibility into what other departments had done. They wanted one longitudinal database containing the patient’s entire journey, not separate paper files scattered through a hospital.
Standardization efforts had begun at Massachusetts General Hospital in 1912, produced a standardized treatment diary in 1919, and expanded through a health-record librarians’ association in 1928. Paper still imposed a hard limit: records could be inconsistently written, physically misplaced, and nearly impossible to move between institutions.
Payment policy transformed the stakes. Employer insurance grew from roughly 10% coverage in 1942 to 30% in 1946 and 80% in 1964; Medicare and Medicaid followed in 1965. Once providers billed insurers or government rather than patients directly, standardized documentation became the evidence required to receive payment.
4. MUMPS made a shared clinical database technically possible
In 1966, Massachusetts General began COSTAR, a computerized ambulatory record funded by the National Institutes of Health. Its programmers needed high-concurrency transactions, efficient storage, and many simultaneous users, capabilities poorly served by existing languages and mainframes.
Their answer was the Massachusetts General Hospital Utility Multi-Programming System, or MUMPS. The language integrated database operations directly into the programming environment and handled concurrent updates, reducing both programmer friction and the risk of collisions in safety-critical patient records.
Lead programmer Neil Pappalardo left in 1968 to found Medical Information Technology, later MEDITECH. Dr. Slack moved to Harvard and connected Faulkner to this community, giving her access to both MUMPS and a programmer-led model of company building.
Epic still uses MUMPS’s modern descendant, Caché, around Chronicles. Many competitors use other technologies; MUMPS and its descendants became a distinctive part of Epic’s architecture.
5. Chronicles put the patient at the center of one permanent data model
Faulkner recalled sitting in her living room in the mid-1970s when the architecture arrived whole: “The sun was shining. I was disattentive. I was just sitting there and suddenly it all came to me. Here’s how you build it. The integrated system.”
She ran to the kitchen and wrote “code, code, code, code, code.” The result became Chronicles, named for the chronicle of a patient’s care journey and still the conceptual core beneath Epic’s modern applications.
Early Chronicles was simply a database running on a mainframe or minicomputer. Hospital departments built text-only screens and queries on terminals, but every screen reached into the same patient-centered data rather than reconciling separate departmental systems.
Faulkner says her original production code has long since disappeared, yet the architectural choice endured: EpicCare, Resolute, MyChart, Cosmos, and specialty modules all communicate with one core database. That continuity became more valuable as the product surface expanded.
6. Epic started with $140,000 and spent a decade looking like a small business
Wisconsin physicians described Chronicles at academic conferences, creating inbound demand from other health systems. Faulkner repeatedly resisted starting a company before conceding that she and several colleagues could operate one part-time.
Human Services Computing began in 1979 with a $70,000 bank loan for a Data General Eclipse minicomputer and approximately $70,000 from friends, physicians, and programmers at a $70,000 pre-money valuation. Those were Epic’s only primary equity investors; later secondary shares reportedly even reached Sequoia.
The operation initially consisted of roughly three half-time people in a Madison basement shared with the future American Girl company. Faulkner’s visit to Pappalardo supplied manuals for HR, payroll, benefits, college recruitment, and internal promotion—not a conventional sales-and-marketing business plan.
Renamed Epic Systems in 1983, the company had nine customers then, only 24 by 1988, and $1.5 million of revenue. Its deliberate target was the most complex academic hospitals that possessed computing infrastructure; it later pursued integrated delivery networks and children’s hospitals.
7. Resolute turned a clinical record into a revenue-producing system
Epic’s first eight years addressed clinical documentation without the hospital’s most economically urgent need. Resolute, launched in 1987, connected clinical activity to billing and revenue-cycle workflows using the same Chronicles data.
Ben’s framing is blunt: if a hospital’s goal is to turn care into legitimate reimbursement without losing information or creating legal exposure, the integrated clinical-and-billing architecture “is the product.” A dropped field can mean lost revenue, fraud allegations, or patient harm.
Competitors frequently combined separately developed or acquired clinical and financial products. Epic’s single database avoided fragile handoffs and gave decision-makers one accountable vendor, unified workflows, and documentation linked directly to the bill submitted to a payer.
Reliability followed from architecture as much as engineering discipline. Epic implementations might remain expensive and complex, but the company could more credibly promise a defined cost and schedule, then deliver software that “does the thing that they say it’s going to do.”
8. Graphical EpicCare completed the suite as personal computers spread
Falling computer prices enlarged Epic’s market in the 1990s. EpicCare, launched in 1992, replaced direct terminal interactions with what Epic describes as the industry’s first Windows-based graphical ambulatory EHR.
Epic now combined Cadence scheduling, Resolute billing, and EpicCare clinical workflows for outpatient care. Revenue rose from $1.5 million in 1988 to $18 million in 1995—strong vertical-software growth, though nowhere near contemporary Microsoft’s $6 billion revenue.
EpicCare Inpatient arrived in 2001, bringing overnight hospital care onto the same foundation. Inpatient work mattered disproportionately because it contained the most complicated and expensive episodes and therefore much of a hospital’s revenue.
The hosts call this the “holy grail”: ambulatory care, inpatient care, scheduling, and billing on one longitudinal patient database. Epic increasingly concluded that hospitals wanted breadth from one vendor, not a collection of better-looking point solutions.
9. MyChart made patients participants in Epic’s network
MyChart grew from EpicWeb, a 1997 project giving physicians remote record access. A young programmer, Sumit Rama, asked president Carl Dvorak for “something hard to do”; work began around 1998 and produced an integrated patient portal in 2000.
The timing was genuinely pioneering: consumers could access regulated medical data online, manage relatives’ care, communicate with clinicians, and eventually schedule appointments. Ben compares the irreversible change to Zillow—once people could inspect the information directly, they would not willingly return to phone calls and faxes.
Self-scheduling, waitlists, and cancellation filling also created direct hospital economics. A missed appointment had been lost capacity and revenue; MyChart could move another patient into the opening while reducing scheduling calls and administrative labor.
Epic now reports 191 million active, deduplicated MyChart users. That patient-side scale reinforces hospital adoption because families increasingly expect a familiar portal and access to records across Epic institutions.
10. Kaiser needed one system after internal development and IBM both failed
Kaiser Permanente’s 2003 RFP covered 30 hospitals, more than 400 clinics, 11,000 physicians, and 8.5 million patients. Because Kaiser combined insurance and care delivery, the replacement system had to coordinate an unusually broad enterprise.
Northern and Southern California had operated almost like separate organizations, each developing proprietary systems with consultants. Physicians transferring from Kaiser’s smaller Pacific Northwest region kept reporting that its ambulatory Epic implementation worked better than the California projects.
Kaiser abandoned the competing internal efforts and initially selected IBM for a system-wide project. That implementation failed, forcing another bid and giving the previously small Epic a chance against Cerner, a public company approaching $1 billion of annual revenue.
The decision-maker distinction matters: Epic’s commercial customers are principally hospital CEOs, CIOs, and CFOs, even though doctors and nurses provide the underlying value. Those administrators needed implementation certainty after experiencing the institutional damage of a failed project.
11. Faulkner refused Kaiser’s preferred compromise and still won
Kaiser proposed Cerner for inpatient care, its established strength, and Epic for ambulatory care, where Epic had proved itself. Faulkner rejected the split: Kaiser should choose either vendor for everything because discontinuity between settings would damage patient records, operations, and billing.
During late-stage technical diligence, Epic initially planned a theoretical architecture presentation. Dvorak instead led an overnight effort modeling Kaiser’s exact transaction flows in Excel, demonstrating excess system capacity while Cerner could not provide comparable evidence.
According to the account the hosts heard, Kaiser then requested equity. Cerner reportedly offered 10% of its company; Faulkner refused warrants for Kaiser or anyone else, arguing that granting them to one large customer would oblige Epic to repeat the mistake and ultimately hurt all customers.
Kaiser still chose Epic. The public “$4 billion deal” included years of implementation, Kaiser employee work, and lost productivity; Epic’s portion was closer to $400 million. The Los Angeles Times called it a potential “Model T of its industry”—not the first system, but the first reaching masses.
12. A successful Kaiser go-live converted reliability into brand
Epic’s annual revenue reached roughly $162 million after signing Kaiser and $500 million by 2007. Hiring reportedly shifted from about ten college graduates per month to hundreds flowing through the doors as other hospital systems waited to see whether Kaiser worked.
The decisive validation came several years later when the enormous system went live without collapsing. Buyers could now say America’s largest health system had tested Epic at scale, turning an architectural claim into a reference that de-risked every subsequent RFP.
Kaiser established the healthcare equivalent of “nobody gets fired for buying IBM,” except IBM’s own Kaiser project had failed. Epic could charge a premium because a functioning implementation was cheaper than years of delay, disruption, and abandoned software.
13. A failed Philips partnership taught Epic to control every dependency
Philips proposed jointly developing radiology software and selling a Philips-branded version of Epic through its global customer relationships. Epic hired people in the Netherlands, invested for at least a year, and may have launched the initiative before it folded.
The financial and organizational detour created lasting scar tissue: partnerships meant risk, external dependencies, and outcomes Epic could not control. The company returned to direct customer relationships, internally built products, and ownership of the full implementation promise.
Ben treats that experience as part of the explanation for Epic’s historically closed posture. Patient privacy and reliability supplied legitimate reasons, but one failed partnership became evidence inside Epic that outside integrations could undermine precisely what customers paid it to guarantee.
14. Verona turned a software factory into a recruiting environment
Faulkner visited Microsoft’s Redmond campus while her son worked there and copied its low-rise, collegiate layout, private offices, paths, and shared spaces. Epic acquired roughly 1,700 acres near Verona, with about 410 developed as campus and the remainder kept as farmland.
The aesthetic came from an earlier headquarters conference room decorated like a Wisconsin lodge after receiving a fireplace. Because customers loved that room, Epic expanded the idea across buildings inspired by Oz, Harry Potter, Alice in Wonderland, treehouses, barns, and Grand Central Station.
Two architecture firms associated with Microsoft’s campus and the Disneyland California Adventure renovation helped design the campus. It now covers 89 buildings. Four indoor auditoriums hold about 18,000 seats; Deep Space sits 74 feet underground and holds 11,400, closer to a basketball arena than a corporate meeting room.
The extravagance serves a practical purpose: attract ambitious graduates to rural Wisconsin and preserve a self-contained culture. Wedding music plays when a new customer signs, presenting the contract as a lifelong marriage rather than a quarterly sales event.
15. Epic’s commandments promise permanence before they mention product
Faulkner calls Epic a “software factory”: it takes developers in and produces medical software. The campus, recruitment system, training, workflows, and customer conferences are infrastructure for that factory rather than ornamental corporate perks.
Its first four commandments are “Do not go public,” “Do not acquire or be acquired,” “Software must work,” and “Reality equals expectations.” Others include keeping spoken and unspoken commitments, refusing mediocrity, being fair, having courage, teaching culture, and avoiding operating debt.
Ben’s observation is revealing: none explicitly mentions healthcare or placing the patient first. They are Faulkner’s general rules for running a durable company, posted throughout bathrooms and break rooms so employees and visiting customers repeatedly encounter the same promises.
16. Epic hires for raw ability and then standardizes how people work
Faulkner found interviews poor predictors of programming skill, so in the late 1980s she asked her seventh-grade son, then participating in coding competitions, to design a test. Variants of that test were reportedly used for 18 years before answers spread online.
Today every role takes structured assessments; even culinary applicants reportedly face logic testing. Programming exercises, a “Rembrandt” assessment, and other instruments can carry enough weight that some candidates historically received offers without conventional interviews.
New hires undergo months of intensive training, learning an Epic method for everything from yellow-legal-pad notes and email writing to software development. The premise is that tested, repeatable practices create trust among people who can then operate with few middle managers.
Titles remain light and the organization relatively flat, but performance expectations are not. Epic hires enormous graduate cohorts, trims aggressively, and operates “up or out”; once someone survives for years, colleagues infer that the person can be trusted inside the machine.
17. Quality control and implementation intensity are parts of the product
Employees complete immersion trips into clinical settings, including operating rooms, to observe actual workflows. That institutionalizes the startup advice to “talk to customers” across the workforce, even though many recruits studied neither medicine nor healthcare.
Developers are trained to minimize time between writing and testing code. When a defect appears, the original developer drops other work and fixes it while the context is fresh; the reported rule is that developers repair their own bugs.
That process prioritizes robustness over fashionable interfaces or maximum shipping speed. A medication-dose bug can kill a patient, while a revenue-cycle bug can erase hospital income or create federal billing exposure—two distinct reasons for an unusually low tolerance of defects.
Implementations resemble “military-level logistics,” with young project managers tracking dependencies across what Peter Drucker called the most complex human organization ever attempted. Ten- or twelve-hour days are common, but high performers often value receiving consequential responsibility immediately after college.
18. Epic organizes almost the entire company around existing customers
The hosts reduce Epic to three principal roles: software developers, project managers handling implementations, and technical specialists supporting installed customers. Roughly eight “salespeople” respond to inbound demand and generally come from implementation or support rather than conventional enterprise sales.
Technical specialists are the largest group. Each customer receives dedicated teams for every product it uses—EpicCare, Resolute, MyChart, Cosmos, and specialty modules—plus a named “BFF,” or best friend forever, accountable for the customer’s overall success.
Epic sends separate report cards to customer CEOs, CIOs, and CFOs, scoring performance from one to five and benchmarking each institution against comparable systems. The unusual message is that buying Epic does not automatically mean using it well.
Customers submit and vote on product ideas at annual gatherings. This is customer obsession aimed at hospital leadership: Epic listens closely because solving those administrators’ problems increases retention, product breadth, and its chance of winning the next institution.
19. Standardization, moderate pricing, and bundling deepen the relationship
Epic strongly recommends standard implementations, partly because uniformity makes updates, new modules, and interoperability easier. Customers can customize, but the hosts believe standard or mostly standard deployments receive better economics, and Epic may decline buyers unwilling to adopt its operating model.
Existing-customer price increases average about 2% annually, below many software vendors. Several sources said MyChart’s price had never changed, though Ben cautions that one unchanged component does not reveal the full economics of a complex enterprise agreement.
The bundling playbook is Microsoft-like. During telehealth’s sudden rise, Epic could recommend a third-party product, develop its own module, and later include that module in an existing agreement; customers then had little reason to procure a permanent standalone vendor.
This is not altruism detached from self-interest. Epic’s thesis is that preserving customer value, avoiding rebids, and winning expansions maximizes its own long-term outcome—the same logic behind Bezos’s claim that Amazon and its customers ultimately want the same thing.
20. Care Everywhere solved Epic-to-Epic exchange by making it mandatory
The hosts separate interoperability into three problems: Epic-to-Epic exchange, Epic-to-another EHR, and third-party applications accessing Epic data. Each becomes harder because technical complexity, patient privacy, hospital competition, and Epic’s own incentives compound.
Care Everywhere addresses the first category and now moves approximately 20 million patient records daily. Ben used it to connect three Seattle MyChart accounts, illustrating why Epic can credibly call itself the healthcare sector’s largest sharer of medical-record data.
Faulkner personally insisted that participating hospitals could not choose which Epic customers received records. One early customer unknowingly accepted the feature and later admitted he would have refused; Faulkner called the approval “pure luck,” then made exchange mandatory and retrofitted older versions.
Hospitals remain businesses reluctant to help local rivals. One CIO’s analogy was that United would never hand Delta its customer data merely because passengers might benefit; healthcare creates sharper moral discomfort because patients reasonably regard medical information as their own.
21. Third-party interoperability exposed a convenient alignment of safety and strategy
Epic historically limited application developers more aggressively than ordinary enterprise-software vendors. HIPAA risk and the absence of any “Cambridge Analytica situation” support that caution: a careless integration could expose exceptionally sensitive data at enormous cost.
The same restrictions conveniently protected Epic’s bundle, reduced competition for the user interface, and preserved its ability to build an internal version of an emerging module. Ben and David keep both truths intact rather than treating privacy either as pretext or as the entire explanation.
Epic-to-non-Epic exchange also suffered from inconsistent standards and different architectures. Even willing institutions faced fragmented databases and ambiguous responsibility when information failed to transfer correctly.
Access has widened through multiple programs, standards, revenue shares, and occasional warrants. Developers can now build substantially more around Epic, though the process remains unlike plugging into a typical SaaS API.
22. HITECH paid an entire industry to digitize at once
George W. Bush called for wider electronic-record use in 2006, when one estimate put EHR adoption near 13% of healthcare facilities. The hosts stress that the digitization push was bipartisan before becoming associated with the Obama administration.
After the 2008 crisis, the 2009 HITECH Act became part of the American Recovery and Reinvestment Act. It offered $27 billion in direct EHR incentives and roughly $36 billion including broader health-IT programs, combining fiscal stimulus with a “shovel-ready” modernization project.
Incentives amounted to roughly $44,000–$64,000 per physician over several years, paid largely through health systems. “Meaningful use” began as a carrot, then became a stick when providers faced financial penalties for failing to satisfy prescribed usage requirements.
One vendor summarized the response: “I’ve got this check dangling in front of me, and I have to check these boxes to get there.” A roughly $2 billion EHR industry accepted extensive product requirements because the government’s $36 billion injection was too large to ignore.
23. The subsidy favored the safest vendor and froze today’s winners in place
When government effectively made every qualified system affordable, hospitals did not need the cheapest option; they wanted the one most likely to install successfully and unlock payments. Ben’s analogy is a subsidy covering either a Target handbag or a Birkin—the premium, reliable product becomes the obvious choice.
Adoption moved from 9% of hospitals in 2009 to 95% by 2014, a five-year shift comparable only to pandemic-era software adoption. Epic was already winning after Kaiser, but HITECH pulled years of RFPs into one concentrated period.
The consequence was path dependence. EHR replacements carry enormous switching costs, so vendors that won during HITECH gained decades-long installed bases while more innovative future entrants lost the greenfield opportunities from which they might have grown.
The hosts call this a regulatory tailwind, not clear regulatory capture. Faulkner served on an Obama health-IT council, but Cerner, Allscripts, and other competitors also had government representation; people within Epic themselves appear ambivalent about rules that made their product more cumbersome.
24. Digitization improved access while failing to deliver clean cost reduction
The patient gains are concrete: online records, messaging, self-scheduling, family-care management, drug-interaction checks, and transfer between providers. A cited study found 45% of patients perceived improved care quality after EHR adoption, while 6% perceived a decline.
Cost evidence runs both ways. One statistic suggested hospitals cut costs by about 10%, yet critics argue EHR prompts generate extra tests, more billing codes, and higher-coded versions of the same visit—the revenue maximization hospitals partly bought the systems to achieve.
A doctor described supervisors asking, “That appointment was a two. Don’t you think it might be a three?” Structured software can identify legitimate missed charges, but critics say it can also facilitate upcoding and encourage care that increases revenue without proportionate patient value.
HITECH’s interoperability aspirations largely failed because meaningful use carried direct financial incentives while common data standards did not. Epic’s comprehensive architecture therefore became even more attractive: customers could satisfy immediate requirements without depending on industry-wide exchange.
25. Meaningful use digitized bureaucracy instead of removing it
By defining “meaningful use,” government effectively specified software features and clinical workflows. Mandatory fields and tracked clicks optimized legal compliance rather than the spirit of helping doctors and patients, while adding permanent operating overhead after stimulus payments expired.
A 2016 study found clinicians spent about two hours entering EHR data for every hour of direct patient care. Digitization also enforced rules previously handled informally—for example, doctors could no longer as easily delegate certain orders to assistants and sign later without leaving an auditable trace.
David’s pushback preserves the baseline: a 1970 study estimated communication activities such as managing physical records already consumed 35–39% of hospital operating costs. EHRs redistributed and increased some burdens, but paper was never an efficient alternative.
Obama conceded in 2017 that “mountains of paperwork” remained and clinicians still spent excessive time on administrative entry. A CIO’s sharper verdict was that HITECH “wildly succeeded” at digitization and did “absolutely nothing” for digital transformation—although digitization created the prerequisite for future transformation.
26. Competitor consolidation strengthened Epic’s integrated architecture
Epic crossed $1 billion in revenue in 2011 while winning Johns Hopkins, Cedars-Sinai, UCSF, and other major systems. Its competitors increasingly combined acquired products: Allscripts merged with Misys in 2008 and Eclipsys in 2010, while MEDITECH acquired LSS Data Systems in 2011.
Cerner’s $1.3 billion acquisition of Siemens Health Services in 2014 added another collection of previously merged systems. Even in 2025, some users reportedly still identified as “Siemens customers” or “Cerner customers,” evidence that legal consolidation did not produce one coherent platform.
Each transaction widened Epic’s relative advantage. Rivals could offer more modules on paper, but the modules came from separate databases and engineering histories; Epic continued building internally against Chronicles and could make one organization accountable for the whole workflow.
Meaningful use accelerated consolidation among both vendors and health systems. Epic did not need the legislation to invent its edge, but the compressed buying cycle and competitors’ integration burdens made that edge harder to overcome.
27. Losing the military contracts spared Epic a decade of implementation risk
The Department of Defense awarded a $4.3 billion EHR contract in 2015; the Veterans Administration followed with a $10 billion contract in 2017. Epic bid through government-contracting partners, but Cerner teams won both.
The DoD rollout took until late 2024 to become fully live. The VA system remains only partially deployed and is now expected across all sites “as early as 2031,” implying at least a 14-year project even under the latest optimistic schedule.
The hosts avoid assigning all blame to Cerner: it operated beneath government prime contractors amid military, procurement, and EHR bureaucracy. Yet the incentive problem is stark—contractors can continue billing when implementations run long, inspiring the saying that failed government contracts make more money than successful ones.
Epic customers told David they were “down on their hands and knees thankful” it lost. Cerner simultaneously absorbed Siemens, lost founder Neil Patterson to cancer in 2017, and cycled through leaders; Epic remained focused on commercial health systems rather than the government “shitshow.”
28. Oracle’s Cerner acquisition validated Faulkner’s permanence promise
Epic subsequently won Partners HealthCare, Mayo Clinic, Cambridge in the UK, Intermountain Health, and CommonSpirit Health. By 2018 it had $2.7 billion in revenue; by 2019, $3.2 billion.
Oracle acquired Cerner for $28 billion in December 2021. Cerner had roughly $5.5 billion of revenue but had been flat to declining since 2018, and Oracle later stopped reporting it separately while describing Oracle Health as a growth and profitability headwind.
Layoffs and Oracle’s effort to rewrite the system made Faulkner’s message more credible: Epic would remain one internally developed platform, never be bought, and never subject customers to an acquirer’s strategic priorities. Every rival transaction made Epic “more different” over time.
The lesson is not that Epic avoided growth. It accepted slow, organic momentum for two decades, then harvested the reliability and trust that shortcuts could not manufacture—“building strength” before the market demanded the full suite.
29. Academic dominance and health-system consolidation extend the runway
Epic announced in 2018 that all 20 top-ranked US academic hospitals used its software, and approximately 90% of medical students now train on Epic. That creates a labor-market network effect: physicians arrive at future employers already familiar with one system.
International operations contribute only about 10–15% of business, but the UK is becoming meaningful. London’s Guys and St. Thomas’ NHS Trust went live in 2023 in what the hosts describe as perhaps Epic’s largest single implementation.
US provider consolidation is a less attractive social outcome but a business tailwind. HIPAA, HITECH, the Affordable Care Act, and broader compliance burdens make independent practice harder, pushing care into large systems—the upmarket customers Epic deliberately chose decades ago.
Ben and David initially expected hospitals to be obvious villains but found many operate on thin margins. They merge and construct new facilities to gain leverage against equally scaled insurers, producing local quasi-monopolies without necessarily creating large hospital profit pools.
30. Epic is expanding from providers into the entire grid of care
Epic calls the next layer a “system of connectedness” or “grid of care”: products linking providers with insurers, pharma companies, home health, rehabilitation, and other post-acute services. The most immediate economic targets are payers and drug companies.
Prior authorization is the cleanest example. Hospitals and physicians dislike manually requesting permission, insurers dislike processing it, and Epic already contains much of the necessary clinical data; automating the exchange could create value for both sides while embedding Epic more deeply.
The playbook remains consistent: sell one indispensable product, observe adjacent vendors, build competing modules when Epic has an integration advantage, include them in enterprise agreements, and only then use the accumulated provider asset to reach entirely new customer categories.
31. Cosmos converts 15 billion encounters into shared clinical intelligence
Cosmos aggregates anonymized information from 295 million patients and 15 billion individual encounters. The source data is about patients and held by institutions, and may reside on-premises or in different clouds, but participating Epic customers contribute it into a queryable collective resource.
Institutions contributing data can access Cosmos without a separate charge. A physician treating an unusual condition can find “lookalike” patients, inspect prior outcomes, and identify clinicians who handled comparable cases rather than relying solely on personal experience.
The hosts invoke Flint’s water crisis as the archetype: patterns invisible across paper files can become obvious when structured records are examined horizontally. Cosmos extends that possibility to rare diseases, clinical trials, population research, and signals spanning hundreds of millions of people.
Epic says its systems prevented 66 million potential adverse drug interactions and 250,000 potential surgical errors in 2023. Those are company figures, but they illustrate the promised endpoint of digitization: computational assistance that would be impossible with isolated paper charts.
32. Epic’s economics are smaller than its systemic importance suggests
Epic serves 607 customers encompassing 3,200 hospitals, 590,000 physicians, 495,000 staffed beds, and 325 million patients worldwide—280 million in the United States. It adds only about 10–25 new health systems annually.
Revenue reached $5.7 billion in 2024, up 16% from $4.9 billion, after averaging roughly 13% annual growth over five years. Epic employs about 14,000 people and holds 42% of US hospitals, 58% of ambulatory physicians, and is used in some way by 79% of Americans.
Dividing revenue by customers gives only around $10 million per health system annually. One midsized customer reportedly employs 100 people internally to administer Epic, while consultants, implementation work, downtime, and optimization absorb much more of the system’s total cost than Epic captures.
Estimated EBITDA margins range from 30% to 35%, implying roughly $1.7 billion at the low end. UnitedHealth Group, by comparison, generates around $400 billion of revenue and $35 billion of EBITDA: its profit alone is about six times Epic’s revenue despite Epic’s infrastructural leverage.
33. A $100 billion valuation is plausible but intentionally unmonetizable
Forbes estimated Faulkner’s 2021 net worth at $7.6 billion, implying an Epic valuation near $15 billion given her roughly 50% economic ownership. The hosts call that “ludicrous” for a growing business with nearly permanent customers.
Applying the roughly 30-times EBITDA multiple Oracle paid for Cerner gives about $51 billion; a nine-times revenue software multiple produces a similar figure. Both understate Epic’s superior growth, integration, margins, and revenue durability.
Ben and David believe public investors might value Epic around $100 billion, and strategic buyers could rationally pay at least that for its industry position, profit stream, and Cosmos data. Both say they would buy at that valuation if shares produced distributions, even without a liquidity event.
The exercise remains hypothetical. Epic buys back shares, will not list, and will not sell; its scarcity is a governance choice, not an overlooked investment opportunity.
34. Faulkner designed succession to prevent later managers from selling
At 81, Faulkner retains approximately half the economics and 100% of voting control. She has signed the Giving Pledge and transfers non-voting shares to the Broots and Wings Foundation, which sells them back to Epic to finance philanthropy.
Upon her death, voting shares will enter a purpose trust governed by family members, five long-serving Epic managers, and roughly three customer CEOs. The structure embeds employees, owners, and the institutions most exposed to Epic’s continuity.
Trust rules prohibit an IPO or sale. They also require future CEOs to be long-tenured Epic employees and software developers, preserving the idea that the company’s leader must come from its core craft rather than finance, sales, or outside professional management.
Founder continuity has already lasted 47 years, longer than Jensen Huang’s Nvidia tenure or Mark Zuckerberg’s at Meta and comparable only to exceptional cases such as Warren Buffett. That duration allowed Faulkner’s personality to become operating infrastructure rather than a set of slogans.
35. Epic possesses nearly every durable software power at once
Switching costs are the most obvious: replacing Epic can cost hundreds of millions or billions, disrupt every department, impair revenue, and put patient safety at risk. The hosts consider it potentially the highest-switching-cost software product ever built.
Scale economies let Epic amortize hundreds of applications across a vast installed base and include modules that smaller competitors must sell separately. A new EHR entrant would need to reproduce decades of reliable functionality before matching the incumbent’s baseline offer.
Network effects now operate through Care Everywhere, familiar patient access, and a labor force trained on Epic. Each additional hospital makes the network more useful to patients and physicians and makes remaining on Cerner or another system more conspicuous.
Brand and process power complete the moat. Prestigious institutions signal that Epic will work, while college recruitment, proprietary training, MUMPS/Caché expertise, customer support, and bug discipline are difficult to recreate by hiring a few experienced executives.
36. Epic reinvested in code while refusing artificial growth
Epic spends about 35% of operating expenses on R&D, versus cited figures of 10% at athenahealth, 23% at Oracle, 28% at Amazon, 36% at Apple, and 45% at Google. With almost no sales expense or acquisitions, internal product development absorbs the resources.
R&D compounds in software in a way that transient selling expense does not. Epic could fund that compounding because Faulkner’s programming ability, modest capital needs, and lack of venture deadlines allowed a small profitable business to wait for computing and healthcare demand to mature.
Faulkner describes growth as climbing the next hill without seeing the whole mountain. Chronicles led to billing, ambulatory care, inpatient care, MyChart, specialty modules, Cosmos, and payer workflows.
37. Antitrust, information access, and international economics form the bear case
International systems generally spend less on healthcare administration than US institutions, limiting willingness to pay even if Epic wins more customers. A domestic market distorted by exceptionally high healthcare spending may not reproduce abroad.
The hosts identify a possible Sherman Act antitrust event around Epic’s control of information and competing health-IT services as a major risk. A material antitrust judgment could alter product access, contracting practices, or the company’s ability to bundle adjacent modules.
The Cures Act restricts “information blocking” and can permit methods such as browser extensions, robotic process automation, or screen scraping. A third party might build the interface clinicians prefer across much of Epic’s base, though replacing the transactional backend would remain a much harder leap.
Greater interoperability could favor best-of-breed applications, while a new care-delivery or AI paradigm could make parts of the traditional EHR obsolete. The hosts find these risks real but hand-wavy because Epic sees them too and still controls the system into which challengers must write.
38. Ambient AI could make Epic more valuable by making the EHR disappear
Physicians who use ambient-listening products reportedly favor them. Microsoft/Nuance, Abridge, and Suki integrate with Epic, letting doctors focus on patients while software prepares records rather than forcing constant typing and clicking.
The near-term bull case is straightforward: Epic becomes the distribution choke point for successful clinical AI, earning partnership economics or building its own bundled version. A healthcare startup reaching hospital physicians increasingly needs Epic’s cooperation.
One CIO offered the more radical possibility that the visible EHR fades into the background. Ambient AI could capture the encounter, structure clinical data, prepare billing, transmit documentation to payers, adjudicate claims, and reduce the interface to exceptions requiring human judgment.
The hosts hedge hard—administrative bloat and the jobs attached to it rarely vanish smoothly—but if healthcare becomes an AI operating system, Epic owns the deepest data, workflows, customer relationships, and permissions from which to build it.
39. Epic won because healthcare rewards one vertical system, not a horizontal platform
Horizontal software must avoid overfitting one customer category; vertical software wins by understanding one industry more deeply than anyone else. Epic listened to hospital leadership, then built every workflow required to run those institutions rather than inviting an ecosystem to complete the product.
Healthcare intensified that logic. A broken interface between applications can kill a patient, leak protected data, lose reimbursement, or create fraud exposure, so buyers value one accountable vendor and one database more than modular elegance.
Most vertical software companies remain limited by their niche, but Epic’s niche is 18% of the US economy and touches nearly every person. That makes it plausibly the world’s most valuable vertical-software company, with Bloomberg the only comparison the hosts could readily identify.
The quintessence is patience under founder control: Epic used time as a resource most capital structures cannot tolerate. It built slowly enough to avoid architectural shortcuts, then let regulation, consolidation, network effects, and 47 years of accumulated code make the integrated model nearly impossible to challenge.