The Evidence Stack Is the Product

The Evidence Stack Is the Product

Caris Life Sciences is presenting 32 studies at the 2026 American Society of Clinical Oncology Annual Meeting in Chicago this week. Every headline will read like a data dump. Almost none of them will tell you what's actually happening.

Caris is not presenting 32 studies to validate drugs. It is presenting 32 studies to validate the methodology that decides which drugs patients get. Each poster, each abstract, each real-world evidence analysis is a proof point for the company's AI-driven evidence stack — a compounding archive of molecular profiles matched to clinical outcomes that now spans 740,000 patients and has accumulated more than 13 quadrillion datapoints. The evidence stack is the product. Everything else is commentary.

That is a meaningful distinction, because the companies paying attention to this ASCO dump are not oncologists — they are biopharma biomarker teams, clinical trial designers, and precision medicine investors. For them, the question is not whether Caris generated a lot of data. The question is whether the data is building something or just accumulating.

The numbers from this year's financial filings make the strategy legible in a way it wasn't a year ago. Caris reported full-year 2025 revenue of $812 million, up 97 percent from $412 million in 2024, according to PR Newswire financial results. Gross margin expanded to 66 percent from 43 percent, a 2,300 basis point jump in a single year. The company posted positive Adjusted EBITDA of $137.7 million. Net loss shrank from $281.9 million to $68.1 million. In the fourth quarter alone, revenue hit $292.9 million, up 125 percent year-over-year.

This is not a diagnostics company anymore. It is a company with enough revenue, enough margin, and enough data to run a decade-long evidence campaign in years. And ASCO 2026 is the loudest single proof of concept for that campaign.

The mechanism is CarisDEAN, the company's AI platform that now powers more than 200 AI tools and underlies every layer of the company's multi-omic profiling — whole exome sequencing, whole transcriptome sequencing, and the matched clinical outcome database that gives those sequences their clinical meaning, per Caris's AI research page. The Caris Precision Oncology Alliance, a network of more than 60 academic and clinical institutions, provides the collaborative substrate. The Flatiron Health partnership, signed in January 2024, added longitudinal electronic health record data to the genomic and transcriptomic picture.

The regulatory validation came in November 2024, when the FDA granted premarket approval to MI Cancer Seek as a companion diagnostic, as Fierce Biotech reported. The approval letter — PMA P240010, publicly available from the FDA's device database — is more interesting than the press release that announced it. MI Cancer Seek is the first and only simultaneous whole exome and whole transcriptome sequencing assay with FDA-approved companion diagnostic indications. It detects single nucleotide variants and insertions/deletions across 228 genes, microsatellite instability, tumor mutational burden, and one copy number amplification in breast cancer. The approval covers one pan-cancer indication and five tumor-specific companion diagnostic indications tied to specific therapies from Novartis, Amgen, Pfizer, Merck, GSK, and Eli Lilly.

But the approval letter also contains a limitation that the announcement did not lead with: genomic findings other than those listed in the approved indications table are not prescriptive or conclusive for labeled use of any specific therapeutic product. That is a precise, FDA-vetted way of saying that the broad molecular profiling Caris sells is not the same as an approved drug selection. The evidence stack generates hypotheses and associations; the FDA's approval covers only the specific validated links.

That gap — between what the platform can profile and what the FDA has approved to direct specific therapies — is where the 32 ASCO studies live. Some will be retrospective biomarker analyses with no prospective clinical validation, according to the ASCO 2026 announcement. Some will be real-world evidence studies drawing on the 740,000-patient matched database. The press release does not sort them by study type, and the abstracts will not be available until the conference begins on May 29. Reading that distinction is the beat.

The historical parallel is Genomic Health and Oncotype DX, which spent more than a decade building a layered evidence architecture — retrospective validation, prospective clinical studies, health economics data — before the test became standard of care in early-stage breast cancer. The Oncotype DX story is instructive because it shows what the evidence-stack strategy actually requires: time, institutional collaboration, regulatory discipline, and the financial staying power to keep running studies without abandoning the approach when results are equivocal. Caris appears to be running the same playbook with a broader platform and at a faster pace, accelerated by AI.

The faster pace is the part that should interest the biopharma teams watching this closely. Foundation Medicine, Guardant Health, and Tempus have comparable or larger commercial footprints in precision medicine profiling. What Caris has that is harder to replicate is the depth of matched outcome data at scale — 740,000 matched profiles with clinical outcomes linked to molecular profiles — combined with an FDA-approved device and a published evidence base that crossed 1,000 studies in December 2024.

If the 32 studies hold up at ASCO, the question for pharma stops being is this validated and becomes what can we build on top of this. That is the inflection point where a diagnostics company becomes an infrastructure dependency. The conference begins May 29 in Chicago. The data will be public. The distinction between evidence and marketing will be in the abstracts.