YC Biotech Startups Bet on Data Infrastructure Over Drug Labs

What happened

In March, Forbes published a list of the 21 most promising companies in Y Combinator's Winter 2026 batch. The list included startups building infrastructure for AI agents, cybersecurity tools, hardware, biotech and even lunar habitats.3 Each company in the batch received $500,000 and spent three months in San Francisco with the accelerator's partners before pitching investors on Demo Day.3 That Demo Day took place on March 24. YC described it as its 45th and said it drew an invite-only audience of about 700 investors and media.10

The list's main biotech pick was not a drug developer. Forbes grouped Strand AI with Shofo AI, which works on video data, and Axion, which works on satellite imagery. Its argument was that the batch's best companies believe producing or organizing the right data can be worth more than building the model that uses it.3 Seven months later, that idea explains more about YC's biotech activity than any one company on the list. It also shows what kind of biotech an accelerator built around software-sized checks can actually fund.

The Strand AI bet

Strand AI was founded in 2025 by Yue Dai and Oded Falik. YC's directory lists it as a two-person San Francisco company.22 Its pitch starts with a large and well-known problem. The company says $60 billion to $100 billion a year goes into clinical trials that never produce approved therapies, and that nine in ten drug trials fail.22 In Strand's view, many trials fail because sponsors can't tell which patients have the biological pathway a drug targets. Patients who respond get mixed in with patients who don't, and the treatment effect disappears in the results.22

Strand's answer is to predict the data that was never collected. On its YC launch page, the company says it trained a foundation model that predicts spatial proteomics from routine H&E pathology staining. It also says the model beats the current state of the art at a fraction of the cost.21 Both founders previously worked at Enable Medicine, building a spatial biology platform at petabyte scale. Dai also worked at Pathos, a Tempus AI initiative, where she trained foundation models on what the company calls the largest multimodal patient dataset in existence.21

Public records add little on funding. A research database lists Strand's only known capital as YC's standard $500,000, structured as $125,000 for 7% plus $375,000 on an uncapped MFN SAFE. It reports no disclosed valuation.24 Another startup tracker describes Strand's total raised as unknown beyond YC.28 One analysis was enthusiastic about Strand but still flagged the main risk. It argued that the company's value depends on whether the FDA will accept AI-imputed data in trial design. If regulators push back, Strand could end up as a research tool rather than a trial tool.29 That reading is the analyst's own, but it identifies the variable that will most affect Strand's valuation at its next round.

A pattern across the batch

Strand was not the only company in the batch selling biological data. Origin, another W26 company, produces patient-derived cancer datasets combining spatial transcriptomics, H&E slides and clinical labels, and licenses them to pharmaceutical and AI companies. It runs its own lab and sources samples from biobank and hospital networks.2 10x Science, also from W26, is building AI models for protein characterization in drug development.2 In the following batches, BioStack Platforms said plainly that "the bottleneck is not models" but access to high-quality biological data.34 Infera is building what it calls a compiler that turns plain-English experiment descriptions into instrument-ready lab runs.2

This is what YC biotech looks like in 2026: mostly picks-and-shovels companies that sell to pharma rather than compete with it. The economics explain why. A data or software company can show customers and revenue on a few hundred thousand dollars. A therapeutics company usually needs years of lab work before it has anything investors can value. One W26 batch database found that one in eight companies was building something physical, with biotech among the categories. It also found three times as many companies reaching $1 million in annualized revenue as in the W25 batch. Revenue-first biotech fits that culture. Wet-lab drug discovery mostly does not.

The funding picture

The numbers support that view. The same W26 database said the best-funded health startup in the batch was Beacon Health, which raised $5.4 million led by Accel, and called it the largest healthcare raise in the batch by a wide margin.5 That ceiling is low next to the costs of drug development. Across YC's portfolio, one analysis of 471 disclosed rounds from January 2025 to September 2026 put the median YC seed round at $4.6 million, with the middle half between $3.5 million and $6.6 million.32

The larger biotech checks are going to older YC companies. ParcelBio, a W24 company developing mRNA medicines, raised a seed round in the $10 million to $15 million range in March 2026.31 Science, an earlier YC company, raised a Series C of more than $100 million the same month. Trackers describe it as a clinical-stage neuroengineering company and list it among the largest YC-backed rounds of the period.1432 In practice, therapeutics and medical-device companies tend to raise big rounds after they have spent years earning credibility, not right after Demo Day.

Where the reporting diverges

Coverage disagrees on how hot the market for YC companies is. After the Winter Demo Day, one outlet said average asks had risen: highlighted companies were raising $5 million to $8 million seed rounds at much higher valuations, compared with the $2 million rounds that used to be typical.7 It also said the healthcare companies investors noticed were practical businesses, such as medical billing automation and clinical-trial patient matching, rather than moonshots.7 By the Summer 2026 Demo Day in September, TechCrunch's survey of venture investors found the opposite on price. The consensus was that valuations were far more grounded than in recent cohorts, even though the batch leaned much further toward deep tech.4

The two accounts can both be true. My reading is that investors are still willing to pay for traction but have stopped paying premiums for ambition alone. That is bad news for early-stage biotech, which usually has scientific promise long before it has revenue.

The science-heavy exceptions

Later batches do include some riskier biology. The Summer 2026 cohort featured TareBio, which says its platform can build personalized or population-level vaccines in weeks by reading the immune cells already fighting a disease. The company says it has produced a vaccine that cured cancer in mice.234 Gamgee is building personalized mRNA cancer vaccines for dogs and says it is running clinical trials in Australia.2 WonderTx is working on oral pills to replace injectable drugs like insulin and Ozempic.2 One write-up noted that WonderTx's founder, Abraham Heifets, co-founded Atomwise, which raised more than $220 million.19 Parasma, which is training human brain cells for computing, made TechCrunch's list of the most talked-about companies at Demo Day.4

These companies are still a small share of their batches. One count of the Summer 2026 cohort found 21 healthcare companies out of 234, or about 8.9%, far behind B2B at 51.7%.6 In the Spring 2026 batch, the programmable cancer-drug startup FinalDose described itself as the only deep-tech biotech company.35 The biggest health check in that batch went to a services company: Clara, an AI primary-care practice from the team that built Circle Medical, raised a $12 million pre-seed from YC and A.Capital.35

Why it matters

YC is not giving up on biology. Its directory lists 138 biotech companies, ranging from Benchling to the current batches.2 But the Forbes list and the batches since then point to a narrower definition of what biotech at an accelerator means. Founders with credible data assets, pharma customers and a software-style path to revenue can raise money at the pace YC's model expects. Companies doing lab-heavy therapeutics work still need a different kind of investor and much more patience.

The next checkpoint is YC's Fall 2026 Demo Day, scheduled for December 2.1 That batch already includes Halmos Labs, which is automating biotech R&D work from wet-lab experiment design to cancer vaccine manufacturing.2 For Strand and the other W26 data companies, the important evidence will come later: whether pharma customers sign contracts, and whether regulators accept predicted biology as a substitute for measured biology. Those two answers will decide whether the data-layer approach becomes the main entry point for biotech or just a smaller, safer version of it.