The infrastructure bet for the next decade
The most expensive lesson in healthcare technology is being right too early.
A wearable for autism, built in 2014. It read heart signals to detect emotional dysregulation before the person was consciously aware of it. The clinical insight was right. The sensors were primitive, the integration path didn't exist, and no care pathway was ready to act on continuous physiological data. It stayed a prototype.
Sensor networks for elderly Singaporeans, deployed nationally by 2016. The clinical signal was real. The hardware cost over $700 per home and the economics never closed. The programme stopped at five hundred homes.
Neither was wrong. Both were early. The distance between those two words is where most healthcare technology bets are lost.
Being early twice teaches a specific discipline: watch the curves, not the trend pieces. Three curves decide when a healthcare technology becomes buildable. The cost of capturing the data. The standards for moving it. The readiness of the clinical system to act on it.
In 2014, none of the three had arrived. In 2016, one. Today, for the first time in thirteen years of watching, all three are moving at once. Sensing is a commodity — the watch on a patient's wrist outperforms the $700 kit of a decade ago. Integration standards exist and are maturing. And AI has changed the third curve at its root: continuous data no longer needs a human watching it. The first level of analysis can now run all day, every day, with the clinician placed at the decision boundary instead of in front of a monitoring screen.
That is the bet for the next decade. Not the app. Not the dashboard. The foundation where continuous data, the clinical record, and the care decision finally share one architecture — ready at the moment the curves cross, instead of years ahead of them.
Most of health tech is still betting on interfaces. The curves say the next decade belongs to whoever builds the foundation.
Tomorrow: the full argument. What healthcare's infrastructure problem actually is, and what replaces it.