Pharma AI Investment Hits a Critical Bottleneck in 2026

Early 2026 saw an unprecedented wave of cross‑border licensing agreements between Chinese AI drug‑design firms and global pharmaceutical giants. Out‑licensing deals reached US$75 billion in the first five months, up from virtually zero before 2020, and the total number of agreements—169—represented an 87 percent jump from the previous year. One headline deal involved Metis TechBio receiving a US$20 million upfront payment and the prospect of up to US$1.6 billion in milestone payments for its next‑generation cancer candidate MTS‑128.

These figures illustrate how quickly capital has flowed into the front end of drug discovery. Artificial intelligence excels at the chemistry side of the process: it can design promising molecules in weeks instead of years and at a fraction of the traditional cost. Companies such as Insilico Medicine report moving from target selection to a drug candidate ready for first‑in‑human safety tests in approximately 18 months for about US$2.6 million, compared with the historical four‑to‑six‑year, multi‑hundred‑million‑dollar timeline.

However, the financial reality of bringing a drug to market remains dominated by later‑stage expenses. The total cost of development still averages around US$2.6 billion, with Phase 1 trials costing roughly US$4 million, Phase 2 about US$13 million, and Phase 3 often exceeding US$20 million. While AI‑generated molecules perform well in early safety studies—showing an 80‑90 percent success rate in Phase 1—their efficacy in patients drops back to the industry norm of around 40 percent in Phase 2. The current data suggest that AI’s primary value so far is “making the cheap end cheaper,” without substantially altering the costly, failure‑prone later phases.

Public policy has not yet caught up with this shift. The United States’ Orphan Drug Act, designed to offset the high cost of late‑stage development for rare diseases, still targets the same expensive clinical milestones that AI has not reduced. Moreover, recent U.S. security measures—adding major Chinese tech firms to a military blacklist—add a layer of regulatory scrutiny that could temper the pace of international collaborations.

Investors and regulators therefore face a strategic dilemma. The flood of early‑stage funding has produced a pipeline rich in AI‑designed candidates, but without parallel investment in tools that can de‑risk efficacy, biomarker selection, and trial design, many of these candidates will falter in Phase 2 or Phase 3. The next wave of value creation will likely come from applying advanced analytics to the efficacy frontier—identifying the right biological targets, optimizing trial populations, and shortening the time to proof of concept.

In summary, 2026’s AI boom successfully transformed molecule discovery into a fast, low‑cost process. Yet the true test of whether AI can improve patient outcomes lies in its ability to tackle the expensive, failure‑prone stages that dominate drug development budgets. Aligning capital, public subsidies, and regulatory frameworks with these later‑stage challenges will be essential for the sector to realize the full promise of artificial intelligence in medicine.