What Scientific AI Companies Sell Next: Experimental Data, Labs, and Drug Discovery Assets

Putting scientific AI predictions to work in drug discovery requires experimental validation. Who pays for that work, and what are they paying for? Two Ginkgo partnerships and a new company co-founded by Schrödinger, announced in September, illustrate different ways to earn revenue that model-access fees alone do not capture.

Sourcing Experimental Data Externally

On September 16, 2026, Ginkgo announced an agreement with TuneLab, Eli Lilly’s AI platform for drug discovery. Ginkgo says it will provide participating companies with experimental services to assess small-molecule ADME—absorption, distribution, metabolism, and excretion—and antibody developability, returning results in a format suited to machine learning. The service is designed to test AI predictions experimentally. Ginkgo announcement, syndicated ↗

As a business, this means earning fees for experimentally validating predictions, separately from charging for AI access. More customers running repeated experiments could mean more orders, but reagents, equipment, and labor also carry costs. Alongside the number of model users, repeat orders and the cost per experiment are worth tracking. Pricing and the size of this agreement have not been disclosed, but those measures can help us assess whether wider use would also bring higher profits.

Bringing an Autonomous Lab In-House

The day before, on September 15, Ginkgo announced that it had been selected to build an autonomous lab at Novo’s new research site in Waltham, in the United States. The lab will combine RAC robotic equipment with Catalyst integration software. According to the announcement, the relationship is expanding from external research support to supplying equipment and software for Novo’s in-house lab. This was an announcement of a selection, not a report that the lab had begun operating. Ginkgo announcement, reproduced without editing ↗

Here, Ginkgo is providing the equipment and software a customer needs to keep running experiments, rather than taking on individual experiments as a service. The business looks different depending on whether revenue comes only from equipment sales or also includes ongoing maintenance and software fees. In general, installation numbers matter for equipment sales, while renewals matter for ongoing services; the breakdown of payments in this agreement is unknown. For the buyer, comparing the arrangement with outsourcing requires considering not only the initial cost, but also maintenance expenses and whether the equipment will be used enough.

Seeking Future Returns From Drug Discovery Assets

Schrödinger’s September 9 announcement illustrates another approach. The company contributed two early-stage small-molecule drug discovery programs to Tectora, which it co-founded with an investment firm. In return, Schrödinger received equity and rights to future milestone and royalty payments. Tectora’s $55 million financing is not Schrödinger revenue. The aim is to advance the programs to the clinical-candidate stage. Schrödinger’s official announcement ↗

Unlike charging for software use, this business can generate future revenue depending on the outcome of drug discovery. The equity stake and payment terms have not been disclosed. Rather than counting these rights toward future sales as if they were a service-order backlog, we want to track the stage of development alongside the funding required.

Distinguishing the Three Businesses

All the measures in the table are items for future review, not reported results disclosed in these announcements.

These distinctions also help researchers and business teams choose partners. For outsourced experiments, compare turnaround times and the terms for repeat experiments; for a lab installation, maintenance costs and operating arrangements; for joint drug discovery, how outcomes and rights are shared. Beyond the headline of an “AI partnership,” what does the contract deliver, what are you paying for, and how much of the burden remains yours? Examining those differences can turn expectations about the technology into concrete decisions.