Digital Health M&A in 2026: The Median Exit Takes 9.5 Years and a $7.7M Last Round

Of 282 digital health exits recorded across 2025 and the first half of 2026, 268 were acquisitions [1].

Whatever you intend to build, that is the distribution you are building into. The interesting question is not whether you get acquired. It is what the buyers in this cycle are paying for, and the H1 2026 record answers it fairly clearly.

They are not buying users. They are buying elapsed time.

The market you are actually in

Three numbers set the frame, and the third is the one founders should sit with.

Funding held steady while the number of funded companies collapsed. Galen Growth reports global digital health funding broadly flat at $22.6B in H1 2026 against $21.4B a year earlier, with deal count falling from 975 to 608 [2]. Average deal size has gone from $18.4M in H1 2023 to $48.9M in H1 2026. The number of distinct ventures receiving funding fell from 1,459 in H1 2022 to 570 in H1 2026, a 61% decline, while total capital fell only 21% [2]. That is not a sector in retreat. It is a sector rationing.

The exit is an acquisition, almost without exception. Of 282 exits recorded across 2025 and H1 2026, 268 were acquisitions, or 95%. M&A has been at least 91% of digital health exit volume every year since 2022 and reached 99% in 2026 year to date. IPO volume peaked at 57 listings in 2021 and has recorded exactly one in 2026 YTD, Generate Biomedicines at $400M, against seven in each of H1 2024 and H1 2025 [1] [2].

And the path there is now nearly a decade long. Median time from founding to exit has moved from roughly seven years in 2022 to 9.5 years in 2026, with corporate buyers typically acquiring companies two to three years later than venture buyers [1].

For scale on the transactions themselves, digital health M&A specifically ran to 82 deals worth $5.15B in H1 2026, against 118 deals worth $3.62B in H1 2025: fewer transactions, $1.5B more disclosed value [2].

What actually changed hands

Each of those is a purchase of something with years inside it.

Kaia gave Sword an operating position in Germany and a pulmonary line alongside its musculoskeletal business [4]. Eucalyptus gave Hims live operations across Australia, the UK, Germany, Japan and Canada [3]. Neither is a technology purchase. Both are purchases of the thing our moat piece called distribution you already own, which is precisely one of the advantages that post argued capital cannot shorten.

Worth noting that this is not only our reading. Galen Growth categorises the largest disclosed transactions of the period by strategic rationale, and the labels are its own [1]:

Capability, vertical integration, channel. Not one of the five is described as a growth-multiple acquisition, and Galen’s own summary of the table is that the largest deals are “concentrated in capability buys, vertical integration, and channel expansion, not generic growth-multiple acquisitions” [1].

The clearest case is the one that shows what a capability actually costs.

What $340M actually bought

Noctrix Health made the NTX100 Tonic Motor Activation system, a prescription pair of leg-worn devices that stimulate the peroneal nerves to suppress restless legs syndrome symptoms during sleep.

Its regulatory history is the asset:

• May 2020: FDA Breakthrough Device Designation [5]
• April 2023: FDA grants De Novo authorization, DEN220059, for moderate to severe medication-refractory RLS [5] [6]
• Supported by RESTFUL, a multicenter randomized sham-controlled trial in 133 patients, meeting all efficacy outcomes [5]

ResMed announced the acquisition on 30 April 2026 and completed it on 1 June [7].

A De Novo is not a 510(k). There was no predicate device to point at, which is the harder path: it creates a new classification. Between breakthrough designation and authorization sits three years, and behind the authorization sits a sham-controlled randomized trial that had to be designed, run and read out.

ResMed has more capital than Noctrix ever raised. It could not have bought those six years with any of it. That is the entire logic of the deal, and it is the clearest illustration in the H1 record of why the six advantages we wrote about are the ones that hold: not because they are clever, but because the clock is the barrier and money does not move clocks.

The shape of the deals changed mid-year

One more figure worth sitting with. Between Q1 and Q2 2026, the average upfront consideration collapsed from $2.4B to $639M, while the median upfront rose from $240M to $650M [3].

Those move in opposite directions for a reason. Q1 contained a few enormous transactions that dragged the mean up. Q2 had none of them, but the typical deal was substantially larger.

So the mega-deal stopped and the mid-market got serious. For a company in the $200M to $700M range, which is where most venture-backed digital health outcomes land, the second half of that sentence is the relevant one: that band got more liquid, not less.

Your acquirer is probably not your competitor

There is a second pattern in the same deal record, and it matters more for positioning than for valuation.

Look at who the buyers are. ResMed is a sleep device manufacturer. Hims & Hers is a direct-to-consumer telehealth business. Raintree Systems, which acquired Spike Technologies in July 2026, is an electronic health record and practice management vendor for rehabilitation and physical therapy [8]. Only Sword buying Kaia is a straightforward same-category consolidation.

That last one is worth pausing on, because it is the closest to home for anyone reading this. Spike sold a wearables and health data API. Raintree’s announcement describes acquiring agentic voice AI for revenue cycle management and never mentions the wearables API at all [8]. The capability the buyer wanted and the product a set of developers had built their integrations on were not the same thing.

The rest are companies from adjacent industries reaching sideways into a capability they do not have. A device maker bought a neurostimulation indication. An EHR vendor bought voice AI. Neither target was competing with its buyer before the deal.

The practical consequence is that the people most likely to acquire you are probably not on your competitive landscape slide. They are companies with distribution or a regulated position in an adjacent market who are missing exactly one thing, and the question that matters is whether you are legible to them as that one thing. A company that describes itself entirely in terms of how it differs from its direct competitors is optimising for an audience that is not doing the buying.

There is a sharper version of this risk, and it has a live example. Roche’s proposed acquisition of PathAI, valued at up to $1.05B, would take a company embedded across ten pharmaceutical partnerships, seven of which sit outside the Roche Group, into the ownership of a direct competitor of those seven [1]. Galen’s framing is worth repeating exactly: an exit “can reshape governance, data access, vendor neutrality, and competitive dependencies for customers and partners.” Change-of-control clauses address the contract. They do not answer whether a previously neutral vendor is still neutral.

That is the infrastructure-layer version of the risk, and it does not require anyone to act in bad faith. An acquisition reprioritises a roadmap even when nothing is announced about the product, and a data API that was the whole company to its customers can be a line item to its acquirer. What is publicly known about each vendor, and what your contract actually guarantees, is on the comparison pages rather than here, because the answer differs by vendor and changes as the record does.

The strongest objection to this post

“Buyers acquire capabilities they lack” is close to a tautology. All M&A is someone buying something they do not have. If that is all this says, it says nothing.

So here is the falsifiable version. The claim is not that buyers acquire capability. It is that in this cycle the capabilities clearing are ones with irreducible elapsed time in them, and the price appears to track time-to-replicate more than revenue. Noctrix had a De Novo and a completed RCT. Kaia and Eucalyptus had live regulated operations in named countries. None of the three headline deals was priced off a growth multiple in the way a 2021 deal would have been.

What would falsify it: a run of acquisitions priced on ARR multiples for products an incumbent could have built in a year, or a reopened IPO market that makes the trade sale optional again. Neither is visible in the H1 data, but a half-year is a half-year, and three verified deals are three deals. Treat this as a read on a thin sample rather than a law.

I have also deliberately left out one transaction that appeared in my research and did not survive verification on a second pass. If a deal is not confirmable from the acquirer or a primary filing, it does not belong in a table like the one above.

What to do with this

Assume the trade sale. With 268 of 282 exits being acquisitions, planning for anything else is planning for the exception. That does not mean building to sell. It means knowing which specific thing you would be bought for, and whether it is getting harder or easier for someone else to reproduce.

Build the thing with a clock in it. A clearance, a validated result, an operating footprint, a longitudinal dataset that had to accumulate in real time. These are the assets whose price is set by how long the buyer would otherwise wait.

Be legible as one capability, not as a small incumbent. Every deal above is describable in a sentence: an RLS authorization, a European MSK and pulmonary footprint, five-country telehealth operations. A company that requires a paragraph is harder to price.

And treat evidence as a commercial asset, not a cost. The IPO window is described as open only to the evidence-led, the acquisitions that cleared carried trials and authorizations, and the same theme runs through everything we have looked at this week, from what validation actually means to why a cleared feature can still miss most cases. The companies that ran the study are the ones with something to sell.

The short version

268 of 282 digital health exits across 2025 and H1 2026 were acquisitions, 95% of the total, rising to 99% in 2026 year to date against a single IPO. Funding held at $22.6B in H1 2026 while the number of funded companies fell 61% from its 2022 level, so the same money reaches far fewer teams. Digital health M&A itself was 82 deals worth $5.15B.

What cleared were positions with time inside them. ResMed paid $340M for a De Novo authorization granted in 2023 on a sham-controlled trial, six years after the breakthrough designation that started it. Galen’s own labels for the largest deals of the period are capability buy, vertical integration and channel expansion, not growth multiples.

And the path is long. Median time from founding to exit is now 9.5 years, more than 60% of exiting companies had reached Series B or beyond, and the median last round before acquisition was $7.7M. The profile that reaches liquidity is a company that lasted, not one that scaled fastest.

If you are building here, the useful question is not what your product does. It is how long someone with more money than you would need to do it too.