It’s hard to believe that we’re now doing 2025 recaps already! I was just writing the 2024 recap for this blog. And, in my last article of 2024 for Fast Company, I had made some predictions for what would happen with health AI in 2025. Well, it’s good to go back and revisit those articles to see if they make sense in retrospect and how much of the predictions actually happened. One of the things I’ve learned about predictions is that they’re not necessarily right or wrong (sometimes they are!!) but it’s the pace with which things happen. I’ve previously mentioned the story of being on a health technology panel in 2014 with a Silicon Valley VC investor and billionaire. At that time, he predicted that in 5 years we would not need radiologists. Well, here we are 10 years later and we’ve never had a greater shortage of radiologists. So, what happened? The fact that computer vision can be trained to read a scan for a specific type of abnormality does not mean that it can now be turned loose to autonomously read radiology scans and humans are no longer needed. In this example alone, it would have been easy at the time to consider the confluence of other issues that could prevent that prediction from coming true (I raised some of these issues when it was my turn to talk on that panel!)
If you train an algorithm to read an X-ray for cancer lesions, it’s not trained to read pneumonia, pneumothorax, and many other lesions that a radiologist may identify when reviewing a scan. You still need to secure reimbursement for your AI solution (most haven’t yet, 10 years later!) and do real-world prospective trials to show improved patient outcomes. Most of that hasn’t happened yet but may someday, so when you make predictions, discussing the timelines and the drivers and barriers that can accelerate or slow down the timelines is important. Right now, there’s optimism and euphoria around what AI can and will do in the years to come. It’s important to understand that achieving that promise may take longer than any of us want.
In my recap of 2024, I called out Gen AI’s promise to start automating certain things such as note taking, coding, and some of the RCM activities. In my predictions for 2025, I focused on some of the same use cases gaining momentum, in addition to other use cases like home testing and passive monitoring. The momentum for workflow and RCM use cases definitely accelerated into 2025 and they were the top two funded areas in health AI for the year. One of the major themes that I touched on in my 2024 recap was the push by organizations like CHAI to create certification process for health AI tools and accredit labs around the country that would do this work. This would make it easier for buyers as they would know that the solutions they were considering had been tested by an independent third party. In 2025, a new administration has upended this process to a degree. Some of the funding eyed for the Assurance Labs that would start doing the certification process was slashed. There was also some back and forth between the new bosses at HHS and CHAI leadership that signaled differences in their visions. It looks like the certification planning through the Assurance Labs may not be on track as previously envisioned but CHAI is partnering with certain companies, such as beekeeperAI in partnership with Mount Sinai and Morehouse, that will test the models on a nationally representative datasets and issue their scorecard. My guess is that this is a scaled down version of that vision where these partners perform that function but there’s not necessarily an organized network of them around the country, at least not yet. This may be due to funding issues and lack of support from the leadership at HHS.
One of my 2024 observations was the mention of a couple of IPOs, Tempus and Waystar, in the health AI space. I was being rather generous since neither of these are pure health AI plays but rather companies that are using AI to power their existing business models. While there’s nothing wrong with that, one of the drivers of growth in this space would be large IPOs for native health AI companies that create a lot of value for their investors. In 2025, we actually had a couple of IPOs of companies whose product and business model is based on AI capabilities. Hinge Health (NASDAQ: HNGE) provides remote physical therapy based on computer vision using the smartphone camera and feedback to the patients. They’ve actually done in the public markets so far as their share price which debuted at around $37 is nearly $50 now. Heartflow (NASDAQ: HTFL) went public in August and their debut share price of around $29 is up to $32 now. They provide an AI-enabled assessment of patient’s cardiac risk by evaluating their cardiac CT scans. While other companies like Caris Life sciences that use AI for precision medicine and decision support also made public debuts, you could say that they’re more similar to Tempus that AI is a component of what they do and not what the business is based on. As such, even though 2025 started out as a promising year for public market and exists overall, the policy headwinds and resulting uncertainty has kept the floodgates from opening. There’s certainly a long line of companies in digital health that need to exit via M&A or by going public to return capital to their investors.
As for funding of health AI in 2025, the headline numbers have been strong and show a slight growth from 2024. This is good news, especially given the challenging exit markets in the last few years. However, total funding amounts are hiding some challenging trends. The most important trend in venture funding this past year, across all VC but in also in digital health, is that capital is being increasingly concentrated in less companies. This is true of late stage and early stage investments with the middle getting much thinner. Investors are putting money into late stage companies hoping to get to exit faster to distribute capital back to their Limited Partners (LP.) For early stage companies, they’re requiring more robust proof of concept, milestones, and track record. While Abridge and Open Evidence attracted $500M-plus over the course of the year in multiple rounds to scale their solutions, the number of companies getting funded has continued to drop. This winner-take-all trend can be expected to continue as building and scaling health AI solutions is hard and when investors see signs of success and traction, they will fight to get a piece of the action. This can create high valuations for those companies, great news for those founders in the short-term but if they don’t achieve their milestones in a timely manner, it could be bad news down the road. We’ve seen this movie with a lot of the companies that got big funding in 2020-2021 zero interest rate era and there was no happy ending in that movie
.
So, there you have it. The momentum for adoption of AI in healthcare continued with workflow and RCM solutions being the headline use cases that got funded and saw adoption. Other areas that I predicted would gain more traction such as home testing and passive monitoring made progress but no major breakthroughs that could parallel ambient scribe or RCM solutions. Capital is becoming more concentrated with companies that are showing early traction and exit markets showed signs of life but have not yet fully opened up. All of this sets us up for a cautious outlook for 2026. Recent legislation that cut Medicaid, established site neutrality payment, and put in motion 340B rollback can mean more challenging economics for the providers and result in smaller technology budgets. In this environment, showing that your health AI solution will help with these challenging economic outlook could help you break out of the pack. To a successful 2026!
