Digital health, clinical AI and the trade war: the view from this side of Lake America

Digital healthcare in Canada has long accepted that the three largest hospital EMR vendors have almost no Canadian presence: no Canada-based employees at all in the case of Epic and MEDITECH, and a handful in the Cerner part of Oracle. Yet our spend on these systems is well over a billion dollars a year. In 2026, living on the shores of Lake America, are we still OK with this?

Two oligopolies, six vendors. The physician EMRs belong to Canadian healthcare giants. TELUS Health is a $2 billion business with 7,800 employees; its EMRs claim more than 40,000 clinicians and plausibly generate $150 to $250 million a year. Accuro is a Kelowna subsidiary of 400 to 500 people serving about 20,000 clinicians, sitting inside Loblaw alongside Shoppers, Lifemark (300 clinics) and Maple virtual care.

WELL Health, a $1.4 billion clinic operator with 3,200 employees, serves 10,000 physicians through WELLSTAR, which reported revenues of $68 million last year. Together: a $300 to $450 million market, headquartered in Toronto, Kelowna and Vancouver inside larger healthcare companies.

The hospital EMR market is three to five times that size. My bottom-up estimate for total costs – licences, hosting, implementation – is $1.5 billion to $2.2 billion a year. Ontario’s Auditor General found that province alone averaged more than half a billion a year on EHR initiatives over 13 years.

Several hundred million goes straight to the vendors. That figure is rising: Newfoundland’s Epic install, several new Ontario sites, and Quebec’s Epic program – contracted at $1.5 billion – ramping to $3 billion or more.

The employee counts. Epic: no Canadian office and no Canadian entity. Its contact page lists a phone number and no address. MEDITECH, in Canada since 1981 lists a 781 number and names three Canada-dedicated staff in Westwood, MA. Oracle does have thousands of people in Canada, but that’s Oracle, not Cerner. On the public evidence, none of the three maintains more than a token Canadian presence for their EMRs.

This is not normal, and it’s not okay. Compare Microsoft: 2.3 percent of its 228,000 employees work in Canada – about 5,300 people in 11 offices, plus two Azure regions and a $19 billion investment commitment. Compare Pharma and Life Sciences where we insist that companies invest in exchange for access to our public market.

Even Heidi, the Australian scribe start-up, has built a Canadian team of more than 30 in three years and hosts Canadian data in Quebec. A three-year-old scribe company appears to employ more Canadians than the three vendors that hold hospital records for over 80 percent of Canadians.

In a year when every cabinet table is arguing about U.S. trade in services, one of our largest health-IT expenditures has no entry on the ledger. And this spend is controlled by the public sector.

Co-pilots are prying this open. On September 1, OpenAI announced that ChatGPT for Healthcare can connect directly to Epic patient records through the hospital’s own FHIR interface, in a chat window beside the chart or embedded inside it.

Microsoft expanded Dragon Copilot into a full clinical assistant this year. Canadian scribes Tali and WELL’s Nexus AI, and Heidi from Australia, all have similar products in market. With every announcement like this it becomes clearer that the system of work is separating from the system of record. Clinical copilots are a welcome development that allows for increased Canadian competition in these hospital systems.

From a trade perspective, that separation is exactly what we want. Tight integration was the oligopoly’s moat: one vendor, one system, one tightly controlled environment, and every innovator waiting at the gate of the castle. A loosely coupled co-pilot layer is a market Canadian firms can actually enter, and they already have or are trying to.

League’s Spring 2026 release moved into agentic care navigation. Cohere launched health agents in the U.S. with an American partner and says Canadian healthcare, with data resident in Canada, is next.

More is coming, from Canadian vendors and from international vendors like Heidi that actually employ Canadians and host here. Will we allow the existing record vendors to bundle their own, proprietary co-pilot and close the gate again?

Epic switched on native AI Charting in February; Oracle expanded its Clinical AI Agent in August. They intend to raise the drawbridge. And leave the peasants outside.

Ottawa has the power levers. On September 3 the Prime Minister created Digital Transformation Canada with an explicit mandate to use federal purchasing power to help Canadian tech and AI firms scale and to strengthen digital sovereignty. That is close to what the Munk School’s Sovereign by Design report asked for in March: one body with the authority to buy, build and invest.

That report places health records in its second-highest sensitivity tier, recommends “juridical sovereignty” as the minimum for the provinces that hold them, and warns that Buy Canadian’s definition of a Canadian supplier still lets a firm qualify with “a Canadian mailbox and non-Canadian personnel.”

A phone number and no address is that loophole, exactly. We should prefer Canadian-owned and -operated infrastructure with a legal air gap from the U.S. CLOUD Act.

Legislation before Parliament (S-5) will give Health Canada new regulatory levers to fill in this moat. The Competition Bureau can help, much like the FTC does in the US.

Be precise about the target. Credit where it is due: Epic’s APIs are genuinely open – the OpenAI integration happened precisely because connecting does not require Epic’s permission. Oracle and MEDITECH are further behind. But an open interface is not an open market. The line is not whether Epic may build its own co-pilot; of course it may. The line is whether a hospital can run someone else’s on equal terms. Firms should not be allowed to favour their own AI products in the name of integration, then sell the integrated system on the basis that they are integrated and no one else is.

The obvious objection. Push the record vendors and Washington pushes back – and the Canadian health-tech firms we are trying to help are the ones who get hurt, because they need access to a U.S. market ten times the size of ours. Non-hospital vendors like PointClickCare, Alayacare, and League may be collaterally damaged. Meanwhile, hospitals under real pressure will buy the best product for their clinicians, not the most Canadian one.

Both points are fair. But we already run this experiment everywhere else in healthcare. No drug or device maker sells into Canada without a licensed Canadian establishment answerable to Health Canada.

They have a real address, a named importer, an inspectable presence. We tollgate pharma and life sciences for investment and have for decades. We require that of the company selling the pacemaker. We require nothing of the company holding the record of the patient it was implanted in.

Premiers and hospital executives can help stop these practices and repatriate much of the spend:

• Make a Canadian office a procurement requirement, not a preference, and tie it to the next competition rather than ask for it in existing contracts.
• Score Canadian employment and Canadian hosting in every procurement; data residency is not data sovereignty when the provider answers to a foreign court. Score down AI training data leaving the country.
• Break the oligopoly apart at the AI co-pilot layer, where switching costs are low and the incumbents’ advantage is smallest.
• Contract for open, non-exclusive APIs, and refuse AI bundles that hand the record vendor the work layer by default. Do not allow contract extensions into AI without a fresh competition.
• Sign shorter contracts.
• Be careful about U.S.-only solutions. Favour companies, even U.S. companies, that have a more global footprint and posture.

On August 30, following a presidential decree, Google Maps renamed Lake Ontario “Lake America” for American users; within a day Ontario agencies and, reportedly, Elections Canada were showing it on their own sites.

If a dependency buried in a U.S. platform can push a politically driven change onto Canadian public websites overnight, we should at least know which dependencies we are accepting in clinical infrastructure.

Are we really okay with most of our hospital data being in these hands?

Keep the innovation layer in Canada. Move the clinical copilot home.

Will Falk is a retired management consulting partner who spends his time with start-ups and as a fellow at four institutions. He is a contributing editor at Canadian Healthcare Technology magazine where his conflicts declaration is on file.