There's an unexamined assumption in how people talk about drug discovery companies: that "real" drug discovery requires owning your laboratory infrastructure. Benches, instruments, reagents, a purchasing department, a facilities manager. The implicit argument is that you can't do serious biology without owning the space where it happens.
We don't agree. CompoundAI has never owned a synthesis laboratory or a biology lab. We work through a network of four principal CRO partnerships and several specialized service providers, and we think this is the right structure for a company at our stage — not a compromise we're enduring until we can build our own facilities, but a deliberate strategic choice.
The Capital Argument
Standing up a functional medicinal chemistry laboratory that can support a single discovery program requires, conservatively, $3-5 million in capital equipment — HPLC-MS systems, preparative chromatography, NMR (or access to a shared instrument), analytical instrumentation, fume hoods, and the safety infrastructure those require. That's before you consider leasehold improvements, facilities operating costs, or the salary cost of the permanent scientific staff needed to keep the lab productive.
That same capital, deployed through CRO partnerships, buys roughly 600-800 compound syntheses with full analytical characterization, or 150,000 ADMET assay data points, or the GLP toxicology package for an entire IND program. The productivity comparison isn't close. For an early-stage company, the question isn't whether to have lab access — you obviously need lab access — it's whether the way to have that access is ownership versus relationship.
The Expertise Argument
Building deep internal capability in synthesis, biology, formulation, analytical chemistry, toxicology, and pharmacokinetics simultaneously is not realistic for a team of forty people. Each of those disciplines has its own specialized expertise that takes years to develop. A world-class DMPK scientist and a world-class synthetic chemist are both full-time senior roles that may not be fully utilized across programs if they're permanent headcount.
Our CRO partners are specialists. The synthesis CRO we've worked with for three years has fifteen chemists whose entire practice is small molecule synthesis for drug discovery programs. Their process chemistry team has solved variants of the difficult transformations we need on dozens of programs before ours. They've seen the protecting group strategy that will fail on our substrate class, because they've seen it fail before. That institutional knowledge inside a CRO is genuinely valuable, and it's not something we'd accumulate quickly even with senior internal hires.
The Flexibility Argument
Drug programs are lumpy. You need high synthesis throughput during lead optimization and almost none during target validation. You need extensive biology bandwidth during hit confirmation and essentially no synthesis during GLP tox. The fixed costs of internal lab infrastructure are constant regardless of where the program is. CRO partnerships scale with program demand.
When CAI-001 was in its heaviest lead optimization phase — roughly eight months in 2024 — we were running approximately 80 to 100 new compounds per month through synthesis and primary assay. When the program moved into IND-enabling studies, synthesis dropped to near zero and the primary spend shifted entirely to the GLP CRO and the clinical site setup. An internal lab staffed to handle peak synthesis demand would have been significantly underutilized for two-thirds of the program.
What We Do Keep Internal
There are things that don't work well in a purely outsourced model, and we've built internal capability in those specific areas. Scientific project management — the people who actually run the CRO relationships, review data as it's generated, identify problems before they become expensive, and push back on the CRO when something looks wrong — is entirely internal. CROs execute what you specify; they don't catch the scientific flaw in your own assay design. That oversight has to be internal.
We also have internal computational capability that is deeply integrated with the experimental work. The predictive models that guide compound prioritization, the SAR analysis that interprets incoming assay data, the ADMET scoring that screens virtual libraries — all of that is built and maintained in-house. It's the core of what differentiates our programs from conventional discovery, and it stays internal for that reason.
The question is never really "CRO or internal." It's "which capabilities create competitive differentiation and which are commodity services." The pipetting doesn't differentiate us. The molecular intelligence does.