Pfizer, one of the world’s largest drugmakers, has partnered with the AI company Edison Scientific to roll out Edison’s Kosmos system across parts of its drug development pipeline, the investor news site Simply Wall St reported on October 311. Kosmos is meant to help with three jobs inside Pfizer’s research and development work: screening possible drug compounds, identifying targets (the molecules in the body a drug is designed to act on) and designing clinical trials1.
This is a business announcement, not a study. It describes how Pfizer plans to use an AI tool. It does not show that the tool finds better medicines, or finds them faster.
What happened
According to the report, Pfizer is bringing Kosmos into its research operations as part of core drug development work rather than as a side project1. In the same update, Pfizer also highlighted its work on next-generation AAV delivery for gene therapy, both inside the company and through outside collaborations1. AAV is a harmless virus that is emptied out and used as a delivery truck to carry new genetic instructions into cells.
The report did not give financial terms for the Edison deal and did not name which Pfizer drug programs Kosmos will work on first1.
What Kosmos is
Edison describes Kosmos as a “collaborative AI scientist” that works with research teams on key decisions, uses their private data and aims to help bring new products to market faster2. It is what the field calls an agentic system: software designed to carry out a series of research steps on its own, rather than answering one question at a time for a human researcher3.
In practice, Kosmos runs several AI agents side by side3. One analyzes data by writing and running computer code over and over. Another searches and reads full-length scientific papers in large numbers3. The two share what Edison calls a structured world model, an internal picture of the problem that is updated as the work goes on3.
Edison says that in a single research run, Kosmos can process about 1,500 full-length papers and run about 42,000 lines of code, then deliver a structured report that shows where each finding came from3.
What it does not show yet
The most useful number for judging the tool also comes from the company. Edison reports that about 80% of the statements in Kosmos reports are confirmed as accurate by human scientists3. Put another way, roughly one statement in five does not hold up, so scientists still need to check the system’s work3. The figure was not described as coming from an independent test.
Kosmos builds on earlier systems from FutureHouse, the nonprofit lab Edison came out of3. One of them, called Robin, picked out ripasudil, an eye-drop drug approved in Japan for glaucoma, as a possible treatment for dry age-related macular degeneration, a common cause of vision loss in older people3. That idea has only been explored in the lab and has not moved into human trials3. It is an example of what these tools can suggest, not proof that their suggestions work in patients.
For Pfizer specifically, there is no public data yet on what Kosmos has found, how often its suggestions are used, or whether it changes how quickly drugs move forward. Simply Wall St noted that the update does not answer how fast these AI tools and delivery platforms will feed into drugs in late-stage testing1.
Who is behind it
Edison Scientific was spun out of FutureHouse in 2025 and is backed by $70 million from Spark Capital, Triatomic Capital and other investors2. Its founder and chief executive is Sam Rodriques2.
There is an earlier business link between Pfizer and Edison’s circle. In June, Edison announced a partnership with Population Health Partners, an investor that builds biotech companies, to use Kosmos to create and run new drug companies2. Population Health Partners is the team behind Metsera, an obesity drug company that Pfizer acquired in a deal worth up to $10 billion2.
The main coverage of the Pfizer deal comes from Simply Wall St, a site written for stock investors1. It frames the move in business terms: Pfizer’s investment story leans on $9.7 billion in planned cost savings and wider use of AI across research, manufacturing and sales, at a time when the company expects lower COVID revenue1. In that framing, Kosmos is a tool for making research cheaper and more efficient, which is a separate question from whether it leads to better medicines.
How it fits the field
Pfizer is not the first drugmaker to sign on. In May, Incyte, a company focused on blood disorders, cancer and immune diseases, agreed to use Kosmos across its drug discovery and development work3. Coverage at the time called that one of the first uses of an agentic AI scientist inside a mid-size to large drug company3. Incyte said it would start with finding and checking drug targets and with translational biology (the step that connects lab findings to patients), and expand later3. Financial terms were not disclosed and no specific Incyte programs were named3.
The Population Health Partners deal goes further, planning to use Kosmos through every stage: spotting opportunities, analyzing clinical and lab data, and helping draft scientific reports and documents for regulators2. Its first project involves a drug already being tested in people, and later work is planned in heart disease, inflammatory lung disease and gut diseases2.
Explaining the thinking behind these deals, Rodriques said:
“Our goal is to use AI to dramatically expand what a biotech team can accomplish. If we get this right, we won’t just build better companies, we’ll build more of them, which means more shots on goal for patients waiting on new treatments.”2
For readers interested in aging and long-term health, the appeal is speed and scale: if AI can sort through research faster, more ideas might get tested. Population Health Partners, for example, says it focuses on innovations that could improve healthspan, the years lived in good health, across large populations2. Whether that promise holds will depend on results in patients, which none of these deals has produced so far.
What to watch next
- Named programs. Whether Pfizer says which drug projects Kosmos is working on. Simply Wall St pointed to future research updates that link Kosmos to concrete late-stage milestones, such as new major trial starts or filings with regulators for drugs the AI helped select1.
- Independent checks. Whether outside scientists test Kosmos’s accuracy, beyond the roughly 80% figure Edison reports3.
- Trial results. Whether any drug idea shaped by Kosmos, at Pfizer, Incyte or a Population Health Partners company, reaches human trials and how it performs there.
- Gene therapy delivery. Further details on Pfizer’s next-generation AAV work, which it highlighted alongside the AI rollout1.
Biohack will follow this story as Pfizer and Edison share more. For now, it is a deal that shows where a large drugmaker is placing its bets, not evidence about any new treatment.
