Technology is the single largest driver of rising healthcare costs, responsible for an estimated 38% to 65% of new healthcare spending. A technical review panel advising the Centers for Medicare and Medicaid Services put it more simply: about half of all real growth in health expenditures traces back to medical technology, particularly when new tools for diagnosis and treatment emerge where none existed before. That includes everything from imaging machines and surgical robots to new drugs and genetic tests.
The paradox is real. Technology makes medicine better, but it also makes it more expensive, often in ways that aren’t immediately obvious. Here’s how that happens.
New Treatments Replace Nothing, Not Something
When people think of technology reducing costs, they picture a newer, cheaper option replacing an older, expensive one. In healthcare, the opposite is more common. Many new technologies don’t replace existing treatments. They create entirely new categories of care for conditions that previously had no intervention at all. A patient who once would have been told “there’s nothing we can do” now has a treatment option, and that treatment costs money that wasn’t being spent before.
This pattern repeats across specialties. Conditions that were managed with watchful waiting now have dedicated drugs, devices, or procedures. Each one adds a new line item to the healthcare budget without subtracting an old one.
Drug Development Costs Get Passed to Patients
Bringing a new drug to market is extraordinarily expensive, and those costs shape the prices patients and insurers pay. A 2024 analysis in JAMA Network Open found the average cost to develop a single new drug was about $173 million. Factor in the cost of all the failed candidates that never made it to market, and the figure jumps to $516 million. Add the cost of capital (the money tied up for years during development), and the total climbs to roughly $879 million per successful drug.
Some therapeutic areas are far more expensive. Cancer drugs cost an average of $1.2 billion to develop when failures and capital costs are included. Ophthalmology drugs cost about the same. At the lower end, anti-infective drugs average around $379 million. These development costs don’t disappear. They’re baked into the price of every prescription, spread across the patients who use them. For drugs that treat rare diseases or serve small patient populations, the per-patient cost can be staggering because there are fewer people to share the bill.
More Imaging Machines Mean More Imaging
Diagnostic imaging illustrates one of the most counterintuitive ways technology raises costs: by existing. When hospitals and clinics install more CT and MRI scanners, utilization goes up. A study published in Health Affairs tracked imaging use over a decade in a large integrated health system and found that the annual per-patient cost of radiology more than doubled during that period.
Several forces drive this. More machines mean shorter wait times, which lowers the threshold for ordering a scan. Improvements in image quality make scans useful for conditions where they previously weren’t. Patients increasingly expect imaging as part of a thorough workup. And favorable reimbursement structures give providers little financial reason to say no. The result is a self-reinforcing cycle: wider availability leads to more use, which leads to more spending, which funds even more machines.
Robotic Surgery Costs More Per Procedure
Robotic surgical systems represent a significant upgrade in precision and visualization, but they come with consistently higher price tags than conventional approaches. Across multiple studies comparing robotic, laparoscopic, and open surgery for the same procedures, robotic surgery costs more every time.
For prostate removal, one economic evaluation found total per-patient costs of $19,360 for robotic surgery versus $14,735 for laparoscopic, a difference of nearly $4,625. Another U.S. study reported average total hospital costs of $10,269 for robotic prostatectomy compared to $8,557 for laparoscopic. For hysterectomy, the gap can be even wider: one analysis found robotic surgery cost $50,758 per patient versus $41,436 for laparoscopic, a difference of more than $9,000.
The higher costs come from the robots themselves (which cost millions to purchase and maintain), longer operating room setup times, and disposable instruments that must be replaced with each case. In some procedures, robotic surgery produces better outcomes or faster recovery. In others, the clinical results are comparable to laparoscopic surgery, meaning the extra cost buys better technology without necessarily buying better health.
Capital-Intensive Facilities Raise the Baseline
Some medical technologies require massive upfront investments just to exist. Proton beam therapy, an advanced form of radiation treatment, is a striking example. Building a proton beam facility costs between $25 million and $200 million depending on size and configuration. Even a compact single-room system, installed inside an existing radiation facility, involves roughly $5 million in construction costs plus $25 million for the device itself.
Once a facility like this is built, it needs to treat enough patients to justify its cost. That creates pressure to expand the range of conditions treated with proton therapy, even when conventional radiation might produce similar results for certain cancers. The same dynamic plays out with other capital-intensive technologies: once the investment is made, the incentive is to use it as much as possible, which drives up total spending.
The Induced Demand Effect
Healthcare has a unique economic feature that most other industries don’t: the person recommending you buy a service is often the same person selling it. Physicians act as both advisors and providers, and they have far more knowledge about medical necessity than their patients do. This creates what economists call supplier-induced demand.
When new technology becomes available, physicians can shift recommendations toward using it. This isn’t necessarily cynical. A doctor who has a new diagnostic tool genuinely believes it could help. But the net effect is that the availability of technology changes prescribing and ordering patterns in ways that increase utilization. Empirical research consistently shows a statistically significant correlation between the concentration of physicians (and their available tools) and per-capita healthcare spending in a given area.
The mechanism works at scale. When a new blood test, imaging modality, or screening tool becomes available, clinical guidelines gradually expand to recommend it for broader populations. What starts as a targeted technology for high-risk patients eventually becomes routine for everyone, multiplying its cost impact across millions of people.
Why Technology Rarely Lowers Total Spending
In most industries, technology drives prices down over time. Computers, televisions, and smartphones all follow this pattern. Healthcare doesn’t, for several interconnected reasons.
First, healthcare technology often adds capability rather than efficiency. A new drug that extends life by six months is invaluable, but it doesn’t make the previous treatment cheaper. Second, patients and providers both prefer newer options even when older ones work well, creating demand for the most advanced (and expensive) version of care. Third, the complexity of healthcare pricing means that savings from one technology rarely translate into lower prices elsewhere. A faster surgical recovery might reduce hospital stay costs, but the surgery itself costs more, and the patient now lives longer and consumes more healthcare in subsequent years.
Finally, there’s the “halfway technology” problem. Many innovations manage chronic diseases rather than curing them. Insulin pumps, continuous glucose monitors, implantable defibrillators, and dialysis machines all keep people alive and functioning, but they require ongoing spending for years or decades. A true cure would be a one-time cost. Management technologies are recurring ones, and they accumulate across an aging population.