A running file on healthcare companies and the industry around them.
I am an undergraduate at Michigan studying business administration, mathematical biology and neuroscience, and this is where I keep track of healthcare. Some of it is drug discovery, some is software that sits inside hospitals, some is hardware people put in their skulls, and a lot of it is the money and policy moving underneath all of that. The through line is that I want to understand how a scientific idea turns into something a patient actually receives, and where that process breaks.
There are two halves. Twelve companies written up as full memos, because working through the same structure every time forces me to answer questions I would otherwise skip. And a dated record of things happening across the sector with what I made of each one. The second half is the part I care about most. Anyone can write a confident company summary on a single afternoon. Following a sector for a year and recording where your read held up and where it did not is harder, and it is the only way I know to find out whether I am actually learning anything.
The money left healthcare, AI took a bigger share of what remained, and the survivors are the ones who already have a drug in a human.
Everyone quotes the statistic that AI took 46 percent of healthcare venture dollars last year. Almost nobody puts the other number next to it. Total healthcare investment fell 12 percent, to $46.8 billion. Biopharma was down 19 percent. Diagnostics and tools were down 33 percent. AI did not grow the pie. It ate a larger slice of a shrinking one.
The number that actually explains the year sits further upstream. Healthcare-focused venture firms raised about $7 billion in new funds during 2025. In 2021 that figure was roughly $41 billion. The limited partners left. Everything downstream follows from that: firms deploying older and smaller funds cannot afford many misses, and a fund that cannot afford misses buys things that already work.
You can see the behaviour clearly in the first half of 2026. Around 68 biotech companies raised roughly $9.1 billion, the strongest first half since 2022, which sounds like a recovery. But about two thirds of those rounds went to companies with an asset already in human testing, and autoimmune and oncology together took more than 40 percent of the dollars. This is not capital returning to the sector. It is capital concentrating on the part of the sector where the risk is visible.
The same pattern shows up on the buy side, for a different reason. More than $230 billion of pharmaceutical revenue loses exclusivity by 2030. Keytruda alone was roughly $29.5 billion in 2023 and comes off around 2028. Eliquis and Opdivo land in the same window. So the large companies are compelled buyers on a deadline, the supply of clinically validated assets is fixed, and 2026 dealmaking passed $123 billion by late June. Lilly alone spent close to $21 billion in a year.
Put those together and you get a market with a very specific shape. If you have human data, you have never had more leverage, because both the venture market and the pharma market want the same scarce thing. If you are a platform company whose pitch is that you will eventually generate assets, you are being repriced, and the repricing is not gentle. Turn Bio had good Stanford science, a licensing deal and encouraging regulatory feedback, and its technology was sold at auction in May because it could not raise.
The uncomfortable part for me is that most of what I find interesting sits on the wrong side of that line. I do not think the market is wrong. I think it is being rational in a way that will make a small number of platform companies enormously valuable and starve the rest, and I do not have a reliable method for telling which is which in advance.
Runway is a competitive advantage, and I learned that the expensive way
For most of last year I evaluated preclinical companies on the quality of the science and treated milestones as evidence of health. Turn Bio taught me those are close to independent variables. A licensing deal and positive FDA feedback do not put money in the bank on a schedule that keeps a company alive. In a market where specialist funds raised $7 billion against a $41 billion peak, how long a company can fund itself is the first question, not the last one. NewLimit raising $435 million for the same mechanism five weeks after Turn went to auction is the cleanest illustration that the difference was framing and timing rather than science.
The proof for AI drug discovery is thinner than the funding
Isomorphic has raised $2.7 billion and has not put a molecule into a person. Its first-trial timeline moved from end of 2025 to end of 2026, announced at Davos in January. Insilico, which raises far less and gets far less attention, completed first-in-human dosing of an AI-designed NLRP3 inhibitor in June. What I now look for is not benchmarks, which companies choose for themselves, but three things: a regulator letting a computationally designed molecule into a human, a sophisticated domain partner paying real money for the capability, and a molecule surviving Phase 1. Alnylam paying Inceptive for siRNA design satisfies the second. Nobody has done the third, and Phase 1 is precisely where structure prediction has no advantage.
The durable clinical software is whatever sits next to the decision
Ambient transcription is commoditised, and the pricing shows it. What matters is what a company converts that access into. Abridge is turning it into billing codes and now pharma trial recruitment, with Lilly taking equity. Ambience is turning it into revenue cycle and documentation integrity. OpenEvidence skipped documentation entirely and built the reference layer physicians use during the decision, growing revenue from roughly $7.9 million to about $150 million annualised at 90 percent gross margins. Same insight, three angles: the recording is worthless and the position next to the clinician is extremely valuable.
Speed of development is now a country-level advantage
Roughly 38 percent of large pharma licensing deals now originate with Chinese partners, and average upfronts have gone from around $52 million in 2022 to well past $150 million. I do not think this is mainly about price, and I do not think it is mainly about novel biology either. The advantage is in development rather than discovery. A company that enrols a trial in months rather than years reaches human data on a known mechanism before a Western company reaches human data on a better one. That is worse for US early-stage biotech than genuine Chinese innovation would be, because you cannot out-innovate an execution advantage.
Whether the platform companies I find most interesting are early or simply wrong. Every argument I make about compressed discovery timelines and virtual cell models depends on evidence that has not arrived, and the market has spent two years telling me it does not want to wait. I could construct a version of this thesis where the 2021 vintage of AI biology companies is remembered the way the 2015 vintage of digital therapeutics is. I do not believe that, but I cannot rule it out from the data I have, and anyone reading this should know that the most confident-sounding parts of my thinking rest on the least evidence.
Why the memos look like memos
Every company below runs through the same sections: problem, product, why now, market, traction, competition, team, risks against mitigants, open diligence questions, and what would change my mind. I did not invent that structure and that is the point. It is roughly what a real investment memo covers, and using it stops me writing three paragraphs about the science and forgetting to ask who pays.
Which numbers I trust
The ones that are hard to manufacture. Revenue growth with a margin attached. Patients dosed. Regulatory decisions. Whether a sophisticated partner paid real money. I discount registered users, market size projections, and anything a company can improve by changing a definition. Where a figure is company-reported and unaudited, the memo says so.
Where I am weakest
I read papers, I do not run experiments, and I have never worked inside a hospital or a pharma company. So I am better on business models, financing and regulatory paths than on whether a specific molecule will work. On the science I lean on what independent parties do rather than on my own read of the data, which is why the diligence sections are as long as they are.
Nothing here is investment advice, and I am not managing money. These are companies I would want to own if I were, written up in memo form as a way of forcing myself to think clearly.