What 2025 Will Teach Us About 2026 in Radiology

2025 Did Not Quietly Evolve

RSNA 2025 was not a conference. It was a stock market compressed into a convention centre. And the ticker tape was obvious: the industry is re-pricing itself around who owns the substrate, who owns the workflow, and who owns the distribution rails. Everyone else is negotiating their future in someone else’s ecosystem. 

If 2024 was the year of endless pilots and polite optimism, 2025 was the year radiology started behaving like an actual market again. One with winners, losers, and a ruthless appetite for consolidation even when nobody signs a merger agreement. 

Here is what 2025 taught us about what 2026 will bring. 

Lesson One: Foundation Models Are Not Features, They Are Gravity 

The loudest signal all year was the rise of foundational models as the new centre of gravity. The reason is not clinical. It is economic. 

Point solutions are a tax on workflow. Every new tool adds procurement drag, integration risk, training friction, and support overhead. Hospitals do not have the time or margin for that anymore. Radiologists do not have the patience either. So the market is pulling toward fewer integrations and broader capability per integration. 

That is why the booths that mattered were the ones selling a new substrate, not a clever algorithm. Mecha Health showing up with enterprise scale multimodal foundations. Hoppr pitching connective tissue across modalities and workflows like radiology is about to become a platform economy. a2z gaining attention for generalisation across service lines. Aidoc sprinting toward multi finding detection and leaning on the one thing most companies still under price, the size and quality of the data set. 

2026 prediction: the foundational players will stop marketing themselves as AI companies and start behaving like infrastructure vendors. Pricing will shift toward platform contracts, enterprise deals, and usage based economics tied to throughput and care pathways. The pitch will be continuity, not capability. 

Lesson Two: Workflow is Now the Scarce Resource

Radiology spent years pretending staff was the bottleneck. Staff is a bottleneck. But 2025 made something clearer. Workflow is the bottleneck you can actually monetise. 

Backlogs are now economically dangerous. Missed SLAs, delayed diagnosis, staffing overtime, and downstream cost ballooning have turned workflow from operational headache into CFO level risk. 

That is why orchestration became the battlefield. Deepc landed like an air traffic controller for departments drowning in complexity. Sectra’s worklist logic is tightening into spine status where compatibility becomes a requirement, not a feature.  

2026 prediction: hospitals will buy fewer tools but pay more for the ones that reduce chaos. Orchestration will be evaluated like revenue cycle systems. Not as innovation. As a survival layer. Vendors that can prove backlog reduction, productivity lift, and measurable throughput improvement will win budgets even in tight spending environments. 

Lesson Three: Marketplaces are Collapsing Under Their Own Economics

2025 also delivered a brutal obituary for the AI marketplace fantasy. Bayer walking away from Calantic and stepping back from platform play was not a one off. It was an admission that the middle layer markup model does not survive in a broke hospital environment. 

The market  rejected platforms because it hates paying for layers that do not change outcomes. 

Hospitals do not want an extra toll booth. They want less friction. And most marketplaces ended up selling more complexity in a prettier wrapper. 

But then Harrison.ai dropped something that feels like the counterpunch. Zero markup, open architecture, ROI first. A platform that openly refuses the traditional take rate. That is a direct attack on the incumbent platform economics, and it forces a question the industry can no longer avoid. 

If the platform does not take a cut, what is the platform actually for? 

2026 prediction: the platform category splits in two. Markup marketplaces continue to die quietly. Infrastructure gateways that reduce integration burden and procurement friction survive, but only if they prove they are cost neutral and time saving. Open wins only if it is operationally cleaner than closed. 

Lesson Four: OEMs are Not Background Noise, They are the Distribution Cartel

Every year startups try to pretend the modality giants are just hardware. 2025 reminded everyone that hardware is still the political capital of radiology. 

GE and Siemens were visibly on guard at RSNA because they understand the threat. Foundational AI is an attempt to own the intelligence layer that sits above the machine. If that layer becomes the primary buying driver, OEMs risk becoming pipes. 

So they are doing what incumbents do when threatened. They bundle. They lock in. They expand vertically. 

You saw it in GE leaning into spectral CT and AI reconstruction. You saw it in Siemens pushing AI tied to system efficiency and cost. You saw it in Siemens signing value partnerships that bake proprietary AI into multi year infrastructure contracts. No pilots. No debate. Just procurement gravity. 

And then you saw the most important move of the year. GE HealthCare taking Intelerad for $2.3B in cash. That was not a PACS acquisition. That was a declaration of vertical integration. Hardware plus enterprise imaging software plus workflow plus image exchange plus recurring SaaS margins. It is a full stack bid to own distribution. 

2026 prediction: the OEMs keep buying software assets that control routing, storage, exchange, and workflow. The winners will be the ones who turn their installed base into a recurring revenue engine without triggering provider rebellion. 

Lesson Five: Capital is Back, but it is Looking for Margin

The most misunderstood shift of 2025 is that funding returned, but it returned with a different appetite. 

Investors are no longer chasing AI novelty. They are chasing predictable profit pools. You could see it in the logic behind the narratives that actually resonated. 

Foundational AI sells a platform story. Investors love lock in. Workflow automation sells measurable ROI. CFOs love cost avoidance.  Hardware renewal plus AI upgrades hide recurring margin inside capex cycles. Public markets love visibility. 

Interoperability improvements lower integration risk. Everyone loves shorter sales cycles. 

Even the private equity cycle signaled the same thing. The Hologic take private storyline was not about innovation. It was about reliable EBITDA being mis priced in public markets. The play is operational leverage, portfolio rationalisation, and cash flow without quarterly theatre. 

2026 prediction: deal flow tilts toward boring winners. Workflow, imaging networks, and recurring software attached to installed bases. Pure point solution AI without reimbursement logic will get stranded. 

The Mistakes 2025 Exposed

First mistake: confusing regulatory clearance with adoption. The market is drowning in cleared tools while usage stays stubbornly low. Clearance gets you permission. It does not get you budget. 

Second mistake: selling AI as a feature instead of selling economics. Hospitals are broke. Burnout is high. If your product does not save time, reduce misses, or protect reimbursement, it becomes a nice to have that dies in committee. 

Third mistake: believing neutrality survives scale. If Blackford gets swallowed by an OEM, neutrality becomes conditional. Distribution layers tend to become gatekeepers. That is how power works. 

What 2026 Looks Like When You Connect the Dots

Radiology is heading into consolidation without the mergers. The hierarchy is tightening. Foundational model companies want to own the substrate.
Workflow engines want to control the arteries. PACS and enterprise imaging platforms want to be the distribution rails. OEMs want to make sure they are not demoted to box suppliers. Private capital wants recurring revenue and defendable margin. 

The fight in 2026 will be about compressing the distance between foundation and workflow. Whoever makes intelligence native to the reading room without adding clicks becomes the new centre of gravity. 

RSNA 2025 showed us the market is moving fast. Not toward more products. Toward fewer layers and more control. Radiology is not entering an innovation cycle. It is entering an ownership cycle. 

And in 2026, the companies that win will not be the ones with the best demo. They will be the ones that make hospitals more profitable while pretending they are just helping clinicians. 

Disclaimer. The views and opinions expressed in this article are solely those of the author and do not necessarily reflect the official policy or position of Test Labs Limited. The content provided is for informational purposes only and is not intended to constitute legal or professional advice. Test Labs assumes no responsibility for any errors or omissions in the content of this article, nor for any actions taken in reliance thereon.