Epic Took Share in a Year Hospitals Bought Less Software

The US acute care electronic medical record market added share to its largest vendor in 2025 while the number of hospitals actually choosing an EMR fell. KLAS Research’s survey of purchase decisions, reported in May 2026, put Epic at 43.7 per cent of the acute care market, up 1.4 points from 42.3 per cent a year earlier, with Oracle Health at 21.9 per cent, Meditech at 14.7 per cent, TruBridge at 7.6 per cent and Altera Digital Health at 2.9 per cent. The same research found that acute care purchases declined significantly during the year as market uncertainty cut buying decisions.

That combination is the whole story of this market, and it is not a story about demand. It is a story about what happens to a software category when the installed base stops turning over.

Growth without installations

Pheonix Research’s forecast for electronic medical records runs from USD 31.8 billion in 2025 to USD 58.6 billion in 2033 at a stated 7.9 per cent. The endpoints compound at 7.94 per cent, so the headline holds. Set that against a year in which purchase activity fell and the source of the growth becomes obvious: it is not new hospitals. It is price per bed, module attach and hosted infrastructure inside accounts that already exist.

The prior year tells the same story with a number attached. KLAS’s 2025 report on decisions made during 2024 recorded 272 hospitals affected by an EMR purchase decision, against 319 in 2023 — a fall of 14.7 per cent. Even in a year with more activity, the affected population was a rounding error against the roughly six thousand hospitals in the United States. An EMR is not a product a hospital buys repeatedly. Most of the market is, at any given moment, not buying.

When the buyer is captive and the purchase is once in a generation, revenue growth inside the base is the only growth available. That is a different economic object from a growth software market. It behaves like a subscription with a renewal floor, which is why the category rate looks more like regulated utility growth than like enterprise software.

The bed-count map is the competitive map

The aggregate shares hide something more useful. KLAS’s category results put Epic first in large health systems above ten hospitals, first in small systems of two to ten, first among standalone hospitals above 200 beds, first in academic medical centres and first in children’s hospitals. Meditech leads standalone hospitals of 26 to 200 beds. Oracle Health leads the smallest standalone facilities, one to 25 beds.

Read as a map of switching cost, this is exactly what you would expect. The hospitals that change systems are the ones small enough for a migration to be a project rather than a decade. The hospitals that do not change are the ones where the record is entangled with a medical school, a transplant programme, a research enterprise and a payer contract.

So the only genuinely contestable share sits at the bottom of the bed count, and that is precisely where the two challengers are strongest. The strategic problem for a challenger is that winning there is a volume business with the worst unit economics in the category, while the profitable tier is structurally closed.

Europe is about to create the same effect by statute

The next leg of demand is not American and it is not clinical. Regulation (EU) 2025/327 establishing the European Health Data Space was published in the Official Journal on 5 March 2025 and entered into force on 26 March 2025. Its timetable is set: the Commission must adopt several key implementing acts by March 2027; from March 2029 the first group of priority data categories — patient summaries and ePrescriptions — must be exchangeable across member states under the primary-use rules, with secondary-use rules applying for most data categories; from March 2031 the second group, covering medical images, laboratory results and hospital discharge reports, follows.

That is not a research programme. It is a statutory specification of what a medical record must contain and how it must move, applied to every member state on a fixed calendar. European EMR spending from 2027 is therefore substantially a compliance retrofit: interface work, terminology and coding alignment, audit logging, patient-access plumbing.

And every euro spent on making the incumbent’s record conform is a euro that raises the cost of replacing it. Compliance retrofits entrench installed bases, because they are sunk expenditure attached to a specific system. The European regime will do to the European market what meaningful use did to the American one: create the category, then freeze it.

What 7.9 per cent does not price

There is a real risk in this forecast, and it is on the demand side rather than the technology side.

An EMR’s pricing power rests on being the system of record. If the record becomes an exchangeable object with legally specified content — which is precisely what the health data space is designed to produce — then the value of holding the record falls relative to the value of reading it. In that world the EMR becomes a data-entry terminal behind a national or hyperscaler platform, and terminal software does not earn subscription multiples.

None of that is imminent, and no current contract reflects it. But it is the asymmetry worth naming: the company that wins the compliance retrofit wins three years of revenue and may be paying for the erosion of its own franchise.

For anyone buying into the sector, four variables are worth watching rather than the topline share number. The number of hospitals in an active selection process, which is published annually. The share of revenue coming from modules rather than core licences. The proportion of installed base on vendor-hosted infrastructure, because that is where the price escalator lives. And whether the health data space implementing acts, due by March 2027, specify an exchange architecture that leaves room for proprietary front ends or one that does not.