HLTH Europe in Amsterdam was buzzing with rooms full of pharma execs, hospital chiefs, frontline HCPs and VCs. That’s one reason I get such value from this meeting, the mix of different stakeholders and how accessible they all are. Anyway this is what I heard with my views woven in:
1. The single-app era is ending
It seems almost every health tech founder has built a clever tool that solves one problem for one type of doctor or patient. Investors have now woken up to the fact these won’t survive on their own. They're too expensive to sell, too easy to copy, and most importantly hospitals (and patients) are sick of juggling multiple different logins. The prediction is a few big platforms swallowing the small tools, something I have been saying for a while. If you're building a point solution, your real exit plan is probably "get bought by the platform," not "become the platform."
2. Nobody can use the data because nobody owns the plumbing
At least three completely separate panels, ironically none of them talking to each other, landed on the same complaint: hospitals and pharma companies aren't short of data, they're short of the boring infrastructure that lets different systems actually talk to each other. This is currently a high priority versus building cool new solutions. Until someone builds that plumbing, every AI tool sitting on top of it is delivering only a fraction of its potential value
3. "Real world evidence" needs a rebrand
Pharma has historically treated real world evidence (what actually happens to patients once they leave a tightly controlled clinical trial) as the less respectable, messier cousin of "proper" trial data. One panellist flipped this cleanly: the clinical trial is the unusual, artificial environment. Real life is the control group we forgot to study. Decades of treatment decisions, approvals and pricing were built on data from people who don't look much like most patients. The upside of fixing that is significant. Evidence collected across genuinely representative populations can identify which patients a treatment actually works best in, not just whether it works on average. That same evidence can now support regulatory approvals and label expansions, opening pathways that didn't previously exist. The trial was never the whole picture. Regulators have known that for some time. What's changed is that the evidence collected outside trials is finally good enough to act on.
4. Data isn't the bottleneck any more. Making sense of it is
A great example from last week was the humble glucose monitor. The sensors have existed for years. What changed adoption wasn't more data, it was someone inventing a single simple measure ("time in range") that a doctor could glance at and act on, and that insurers could understand well enough to pay for. The lesson for anyone building in health: nobody's waiting for more data. They're waiting for someone to make the data make sense.
5. Growth is back in fashion, but founders are getting squeezed
For a couple of years, investors wanted profitability above all else. Now at HLTH several said speed and growth matter again, because AI has genuinely compressed how fast you can build and iterate. The problem is it hasn't touched how long a hospital procurement committee takes, or how slowly clinician behaviour shifts, or how much a pharma compliance review can stretch out a deal. Founders are being asked to show growth metrics that belong to a different industry. You can now build in weeks but good luck getting a hospital to sign in that timeframe.
6. Europe is sitting on some rare earth treasures it can't dig up
Pretty much every European speaker said some version of the same thing: we have richer, more joined-up patient data than the US, spread across an array of public health systems. The problem is getting it shared, and not just between countries. A Swiss provider admitted hospitals there can't see each other's records despite a tiny population. A GP in the same country has no real incentive to use any of this anyway, he's only paid for seeing the patient in person. And patients themselves are often the ones holding it back: plenty are happy for their hospital to use their data, far fewer want it shared any wider than that. But on the cutting edge of the AI models themselves, one investor was blunt: Europe is months behind the US and China and isn't catching up. Different race, different rules. Europe's edge isn't going to be the smartest model, it's going to be the best data underneath it, if it can ever get itself organised.
7. Hospitals are the AI story nobody's pitching
The real question in European healthcare isn't "what AI tool should this hospital buy." It's "should this hospital exist at all." Germany is quietly working out it has roughly a thousand too many, and funding cuts are forcing the issue from January. Once you close the ones that shouldn't be open, what does genuinely good AI-supported care look like in those that remain? Nobody on stage was pitching that. It's not as cool as a chatbot demo. It's also the only question that actually matters at scale.
8. Digital therapeutics are quietly trying again
A few years ago "digital therapeutics" (apps that are prescribed and reimbursed like a drug) were the hot category. Then several high-profile ones went bust and the whole sector went quiet. Takeda's now showing a working example: a digital tool for adult ADHD patients who don't want medication, properly reimbursed in Germany, not competing with their own drug but filling a genuine gap next to it. Investors are still nervous (a few admitted as much), but it's a useful reminder that it was less about the wrong idea than the wrong timing
9. Patients are becoming the data source nobody planned for
Wearables, what people buy at the pharmacy, what they ask ChatGPT at 2am, all of it sits completely outside the official medical record, and nobody has worked out how to use it properly. One speaker's example: a spike in over-the-counter cold and flu medicine sales is already used as an early warning sign for outbreaks. That's been true for years using basic supermarket data. Now imagine that same idea applied to everything from wearables to symptom-checker apps. The patient, not the hospital, is becoming the richest source of everyday health data, and the system hasn't caught up.
10. Big pharma's AI maturity varies wildly, even inside the same company
Sanofi described rolling out AI tools to nearly the entire workforce, not just the scientists, down to expense reports being approved automatically. Impressive. But a pharma-wide research group made an honest admission: most of the industry is still using AI to do existing jobs a bit faster, not to fundamentally rethink how anything gets done. Worth remembering next time a pharma company's press release uses the word "transformation."
And my most important trend, it’s #2, getting the infrastructure and plumbing right. Quite a lot of people approached me at HLTH to ask me what the next big thing is I sensed a few were not expecting me to talk about plumbing. Unglamorous as it is, its currently more important than every clever app and genius piece of tech sitting on top of it.
If you want to acquire a valuable shiny tool then a new plumbing wrench is a great start.
