Aidoc’s new $150M round is a bet on the hospital AI layer, not one more imaging model

Goldman Sachs is not backing Aidoc because radiology artificial intelligence suddenly became real this week. The newer signal is narrower and more useful: a major late-stage investor is betting hospitals will pay for the software layer that helps them deploy, govern, and monitor many regulated imaging models inside one workflow.

That matters because hospital AI is drifting away from the simple "we have a model" pitch. Aidoc, an Israeli clinical AI company, is pitching aiOS as the operating layer around multiple imaging tools, not just another one-purpose algorithm. For buyers, the hard part is less whether one model works in a demo and more whether a hospital can make a growing pile of models usable without creating a governance mess.

Growth Equity at Goldman Sachs Alternatives led Aidoc's new $150 million Series E, with participation from General Catalyst, SoftBank Investment Advisors, and NVentures, according to PR Newswire. The same release said the round brings Aidoc's total funding to more than $500 million.

That also makes the cleaner press-release frame a little late. Nvidia's venture arm is not a fresh signal that healthcare AI has arrived. Aidoc already announced a separate $150 million financing round in July 2025 led by General Catalyst and Square Peg, with NVentures participating, to build what it called its Clinical AI Reasoning Engine, or CARE. The new information is not that Nvidia believes in medical imaging software. It is that Goldman now seems willing to back the operating layer wrapped around that earlier model story.

Aidoc does have product progress underneath the financing. In January, the company said the FDA cleared what it called healthcare's first comprehensive foundation model AI for acute abdomen and chest conditions. The system combines 11 newly cleared indications with three previously cleared ones into one computed tomography, or CT, triage workflow. Aidoc said the FDA-reviewed pivotal study showed mean sensitivity of 97 percent and mean specificity of 98 percent across the 11 newly cleared indications, which is the kind of performance claim a hospital buyer would need before even caring about the surrounding software layer.

Those numbers are company-reported, and that caveat matters. Most of the accessible sourcing around this round still comes from Aidoc's own materials, not from independent hospital buyers explaining why they chose the platform over rivals. The company's Series E blog post says CARE cut false positives and false negatives tenfold versus earlier iterations and doubled disease coverage within a few months, but those are still Aidoc's claims about Aidoc.

The stronger support for the platform thesis is not the marketing language. It is the way Aidoc keeps trying to make third-party models part of the story. In December, Aidoc said a new partnership with MONAI would let health systems and academic labs connect MONAI-based imaging models into aiOS through one Aidoc application programming interface, or API. Back in July 2025, the company said 69 percent of its customers were already running non-Aidoc models on aiOS.

My read is that this is the strategic claim underneath the financing, not a sourced fact investors have stated publicly. Radiology AI is old enough now that "we have a model" is not much of a moat by itself. A hospital buyer has to care about where those tools sit in workflow, how they are monitored, how many findings they can cover in one pass, and whether internal or third-party models can be added without turning procurement into a permanent integration project.

Aidoc's scale claims are real signals, but they need to stay separated because the company describes different things in different documents. In its Series E blog post, chief executive Elad Walach said Aidoc now operates in nearly 200 health systems and exceeded 60 million AI patient cases annually last year. The PR Newswire release instead said Aidoc supports approximately 60 million patients each year, has been deployed at nearly 2,000 hospitals, and has analyzed more than 110 million patient cases historically. Those figures are not directly comparable. They use different denominators and time windows, which is a reminder that this market is still narrated largely by vendors.

The skeptical view is simple. This may still be a funding story dressed up as inevitability. The available sources do not show hospitals publicly converging on one default operating system for clinical AI, and they do not prove Aidoc has already won that role. They show a company with real regulatory progress, real backers, and a strong incentive to argue that the messy middleware layer of clinical AI is where value is shifting.

That argument may be right. But the new signal is Goldman, not Nvidia, and the object of the bet is not radiology AI suddenly becoming credible. It is the less glamorous idea that once hospitals own a pile of models, someone has to make the pile usable.