According to Business Insider, healthcare VCs predict 2026 will see a major investment shift toward AI transparency and data quality, a surge in private equity M&A for AI assets, and a continued quiet period for IPOs. This follows a hot 2025 where AI scribe startups like Abridge and Ambience Healthcare landed huge funding rounds, reaching valuations of $5.3 billion and $1.25 billion respectively, and faced new competition from Epic’s own AI tools. Investors like Sapphire Ventures’ Cathy Gao now call opaque “black box” AI models “uninvestable” for healthcare, forecasting the next unicorns will be “glass box” platforms. While a few companies like Virta Health may eye 2026 for an IPO, VCs aren’t optimistic about a public market deluge, instead pointing to PE firms like New Mountain Capital, Bain Capital, and KKR as likely buyers seeking AI acquisitions for liquidity.
The transparency mandate
Here’s the thing: building an AI that fills out a form is now table stakes. The real money, according to these VCs, is moving to the boring, crucial infrastructure that explains why the AI did what it did. Cathy Gao’s “glass box” comment is pretty damning for a whole generation of startups. It signals a maturity in the market. Everyone got excited about the AI magic trick, but now the enterprise buyers—hospitals, insurers—are demanding to see the wires and mirrors. They need a digital paper trail for regulators, for liability, and for basic trust. Liam Donohue from .406 Ventures basically confirms this is where the smart money is going now. It’s a classic tech cycle: the wild west of innovation gets followed by the sheriffs of governance and infrastructure.
The private equity liquidity play
So if the IPO window is barely cracked open, where do VCs get their returns? Their big hope is private equity. Michael Greeley’s point is key: big buyout funds are “going down market to buy AI assets to wrap around their legacy platform companies.” That’s a huge deal. It means PE sees AI not as a standalone business, but as a feature to bolt onto their existing, often slower-growing, portfolio companies to modernize them. For a VC-backed AI startup, getting acquired by a PE firm like New Mountain Capital (which did this several times in 2025) might be a much cleaner, faster exit than the grueling IPO process. This creates a whole new exit path that didn’t really exist at scale a couple years ago.
The surprising comeback kid
Now, the most interesting prediction might be Alyssa Jaffee’s: “tech-enabled services is coming back, baby.” That’s a big shift. For years, VCs ran from capital-intensive models that actually employed doctors or ran clinics. The margins were tough and scaling was hard. But Jaffee argues the survivors of the 2021 boom and subsequent drought have built “meatier” businesses with durable customer relationships. The successful IPOs of Hinge Health and Omada Health—both tech-enabled service providers—prove the model can work at scale. And Wendy Xiao’s note about employers testing competing services, like for GLP-1 management, on a outcomes-based payment model is huge. It turns the cost argument on its head. If a startup is only paid a portion of what it saves, the employer’s risk plummets. That could open the floodgates for adoption.
The zero-sum AI arms race
Let’s not overlook Todd Cozzens’ nuclear analogy about the payer-provider battle. It’s a fantastic point. For the last year or two, AI investment heavily favored providers (hospitals, clinics) with tools for documentation and billing to capture more revenue from insurers. Cozzens is saying the pendulum is swinging. Insurers are now partnering with the likes of Palantir and Anthropic and will deploy their own AI to scrutinize those bills and manage care. It’s an arms race. And in any arms race, there’s a ton of money spent on defense and offense. The winners might not be the frontline soldiers, but the companies selling the high-quality intelligence (data platforms) and the weapon systems (transparent AI models). It sets up 2026 to be a year of fierce competition, not just in the market, but in the very architecture of healthcare payments.
