Insilico Medicine’s Alex Zhavoronkov bets China and AI can deliver the drug industry’s next breakthrough

Insilico Medicine began Phase 3 trials in China for rentosertib, an idiopathic pulmonary fibrosis treatment it says is the first AI-discovered drug to reach large-scale clinical testing; the first patient was dosed on Sept. 10. The startup reported $35.5 million in first-half 2026 net profit, supported by business-development deals including partnerships worth up to $2.75 billion with Eli Lilly, $2.5 billion with SK Biopharmaceuticals and $600 million with Takeda. A Nature Biotechnology study found preliminary, small biological-age reversal signals in blood samples from 42 rentosertib patients, though executives cautioned the effect may not persist. Insilico is expanding its China-for-China licensing strategy and commercializing AI capabilities through its molecular-model fine-tuning platform.

Analysis

LLY gains inexpensive option value from externalized discovery productivity, but the near-term P&L effect is likely immaterial relative to its pipeline and obesity franchise. The more investable read-through is for AI-discovery comparables RXRX, SDGR and EXAI: a late-stage asset provides credibility to the platform model, yet does not validate the claimed speed/cost advantage until randomized efficacy, safety and commercial-scale manufacturing are demonstrated. Expect a sentiment-led rerating over days to weeks, followed by differentiation based on cash runway, partner-funded R&D and milestone-revenue quality.

A China-only pivotal program has asymmetric implications for Chinese innovative-biotech players such as ZLAB, BGNE and HCM. Faster enrollment and lower development costs can increase asset IRRs and create a licensing-arbitrage model, but it may also intensify competition for trial sites, investigators and reimbursement budgets; local distributors and established commercial platforms capture more value than discovery vendors. The claimed profitability should be discounted until investors can separate upfront/milestone recognition from recurring software, research-service or royalty economics; nonrecurring deal revenue supports valuation less than durable platform gross margin.

Over 1-3 months, clinical-trial initiation and partnership announcements can sustain AI-biotech beta, especially if large pharma continues to outsource early discovery. Over 6-18 months, the decisive catalyst is whether pivotal data show a clinically meaningful fibrosis benefit with acceptable tolerability, not biomarker or biological-age signals. A weak efficacy signal, safety imbalance, delayed enrollment, or lack of ex-China regulatory alignment would reverse the narrative and likely compress the entire AI-drug-discovery cohort, whose valuations still embed platform optionality rather than proven commercial output.

Contrarian view: the market may overpay for the "AI-designed" label while underpricing China’s role as a lower-cost development and licensing venue. If the asset succeeds only in China, economics may accrue disproportionately to local licensees and manufacturers rather than to global platform peers. Conversely, independently replicated efficacy could force large pharma to raise external-discovery budgets, making current public AI-discovery valuations less compelling than selective pharma partners with balance sheets and commercialization capability.

Key Decisions for Investors

• Maintain/establish a small long LLY position over the next 1-3 months versus a diversified pharma basket; treat external AI discovery as upside optionality, not an earnings driver. Risk/reward is modestly favorable if partnership cadence continues, but exit the incremental thesis if LLY signals lower external-R&D spend or reports material program discontinuations.
• Use any 10-15% AI-biotech sympathy rally to construct a relative-value basket: long SDGR, which has partner support and software exposure, versus short RXRX or EXAI in equal dollar size only after reviewing cash runway and milestone concentration. Target 10-15% spread return over 3-6 months; stop out on major proprietary clinical validation or a large non-dilutive partnership for the short leg.
• Watch ZLAB and BGNE rather than chase immediately: initiate exposure only if Chinese innovative-drug reimbursement policy, licensing receipts, and trial-start data confirm that development speed converts into commercial economics. A 6-18 month long thesis is falsified by reimbursement-price pressure, weakening cross-border licensing, or evidence of trial-capacity bottlenecks.
• Set an event alert for pivotal efficacy, discontinuation, or material safety updates from the fibrosis program. Positive hard clinical endpoints would justify adding AI-discovery beta; biomarker-only updates should not. Negative data would support reducing RXRX/SDGR/EXAI exposure and favoring profitable large-cap pharma.