AI Drug Discovery: Who Sells the Picks and Shovels?

Faster discovery may not remove the bottleneck. It may move it.

AI has demonstrated meaningful productivity gains in the earliest stages of drug discovery. But the evidence weakens sharply as we move toward clinical outcomes.

This distinction matters for investors. The conventional picks-and-shovels thesis assumes that faster AI-enabled discovery will create a flood of viable drug candidates and push bottlenecks downstream. Current evidence does not support that as the baseline case.

A Better, Not More outcome is at least as plausible: AI may improve selection so that more digital hypotheses lead to fewer, but better, physical development programs.

The relevant investment variable is therefore not gross downstream demand, but persistent net scarcity: demand that remains after AI-enabled efficiency and capacity expansion are taken into account.

Demand expansion is only half of the equation.

AI may increase downstream demand while simultaneously expanding the capacity to meet that demand through automation, self-driving laboratories, patient-trial matching and other efficiency gains.

A new bottleneck emerges only where demand grows faster than available capacity.

Figure 4 — Net Bottleneck Migration.

The evidence is uneven across the downstream stack.

A bottleneck is not an investment thesis.

Structural exposure must pass through a series of economic gates:

Even real scarcity may fail to create shareholder value if the economics are passed through, the advantage is replicable, or expectations already price in a larger fundamental uplift.

Figure 6 — The Bottleneck-to-Investment Test.

Persistent scarcity matters more than gross demand.

AI has already demonstrated that it can improve selected computational and early experimental processes in drug discovery. There is not yet sufficient evidence that these gains translate into higher productivity across the full drug-development system or into more successful medicines.

Candidate Flood should therefore not be treated as a proven baseline. Better, Not More is at least as plausible.

The downstream investment opportunity depends on persistent net scarcity, and even that is not enough. Scarcity must transmit economically to a company, the company must retain the value, the advantage must be defensible, and valuation still matters.

That is where the picks and shovels may be, if they exist at all.

Read the complete 13-page report.

The full publication includes the evidence ladder for AI productivity, the historical drug-development funnel, Candidate Flood vs. Better, Not More scenarios, bottleneck analysis, company transmission case studies, the investment framework, falsification criteria, and 19 references and source notes.

Suggested citation.