AI Drug Discovery’s $45.9B Deal Surge Masks a Deeper Productivity Gap

In a single week in November 2026, drug discovery teams will gather in Barcelona on 10–11 November and at the Broad Institute in Cambridge, Massachusetts on 16 November for two of the year’s most closely watched meetings on artificial intelligence in pharmaceutical R&D. But the real signal is not the conference calendar. It is the data behind it: $45.9 billion in headline partnership value in the first half of 2026, an 80–90% Phase I success rate for AI-discovered molecules, and a blunt finding from Deloitte that AI has not yet reduced the cost of bringing a drug to market.

For executives, founders and investors, these numbers define a maturing field. AI and machine learning are no longer experimental add-ons in drug discovery. They are embedded in pipeline strategy, dealmaking and clinical development. Yet the benefits are uneven: AI is demonstrably improving early-stage success, but it has not solved the industry’s deeper productivity problem. Understanding that distinction — and where the next wave of value will come from — is now a strategic imperative.

A crowded calendar signals strategic urgency

The density of high-level AI drug discovery events in late 2026 reflects how central the technology has become. The AI Drug Discovery & Development Summit in Boston from 27–29 October 2026 is expected to draw more than 1,000 attendees, including representatives from all top 50 pharmaceutical companies. That level of participation would have been unthinkable a decade ago, when AI in drug discovery was largely confined to academic labs and a handful of startups.

Two weeks later, the AI/ML for Drug Discovery conference in Barcelona, organised by Cambridge Healthtech Institute, will run on 10–11 November 2026. It will focus on practical applications of machine learning across target identification, lead optimisation and preclinical development. On 16 November, the Broad Institute will host the Machine Learning in Drug Discovery Symposium, co-organised with CDoT. The symposium brings together computational biologists, chemists and clinical researchers to address how machine learning can improve target validation and translational success.

These meetings are not merely networking opportunities. They are where standards for data quality, model validation and regulatory acceptance are being shaped. For companies that have not yet built internal AI capabilities, the message is clear: the field is moving quickly, and late adopters risk being locked out of the most productive partnerships.

Market estimates diverge, but deal momentum is real

Estimates of the AI in drug discovery market vary widely, ranging from $2.9 billion in 2026 according to Grand View Research to $24.51 billion according to Towards Healthcare. The gap reflects different definitions of what counts as AI-driven drug discovery, as well as whether software, services and internal R&D spending are included. Such divergence is common in emerging technology markets, but it should not obscure the underlying activity.

Deal flow tells a clearer story. In the first half of 2026, AI and machine learning R&D partnerships reached $45.9 billion in headline value. However, the structure of those deals is tightening. Upfront commitments fell from $1.7 billion in the first quarter to just $200 million in the second quarter, indicating that pharmaceutical partners are shifting more risk onto AI companies and tying payments to milestone achievement rather than early promises. The shift toward milestone-based economics is a rational response to the gap between headline deal values and clinical proof. It also means that AI drug discovery companies must maintain sufficient capital to reach late-stage milestones, rather than relying on large upfront payments.

Notable collaborations illustrate both the scale and the shifting terms. Insilico Medicine signed deals with Takeda worth up to $600 million and with SK Biopharmaceuticals worth up to $2.5 billion. Eli Lilly extended its collaboration with Insilico in a deal worth up to $2.75 billion. These headline figures are large, but the declining upfront payments suggest that pharma companies are demanding more proof before committing capital.

Clinical data show early de-risking, not yet cost savings

The most important evidence for AI’s impact comes from clinical development. According to a 2024 analysis published in Drug Discovery Today, AI-discovered molecules achieved an 80–90% Phase I success rate, well above historical averages for the industry. Phase II success rates, however, were around 40%, roughly in line with industry norms. This pattern suggests that AI is particularly effective at selecting molecules that are safe and well-tolerated in humans, but it has not yet solved the harder problem of proving efficacy in larger patient populations. The contrast between Phase I and Phase II results is important because Phase II is where efficacy in patients becomes the primary test. AI’s strength in predicting safety and pharmacokinetics does not automatically translate into predicting therapeutic benefit.

A concrete example is rentosertib, an AI-discovered drug from Insilico Medicine. In 2025, rentosertib reported positive Phase IIa results, including a +98.4 mL improvement in lung function compared with placebo. That is a meaningful clinical signal, but it is still early-stage data. The drug must now demonstrate durable benefit in larger trials before it can be considered a commercial success.

Deloitte’s 2026 report on biopharma R&D returns adds a critical caveat. While the forecast internal rate of return for top biopharma companies rose to 7.0% in 2025, the average cost to develop a new asset increased to $2.671 billion. AI has not yet reduced development costs, according to Deloitte. Instead, its current value lies in de-risking early development — improving the odds that a molecule survives Phase I and reaches later stages, where the real capital is spent.

Strategic implications for executives and founders

For leaders in biopharma and AI-driven drug discovery, the 2026 landscape offers a clear set of priorities.

• Partnerships remain essential. The largest deals involve AI companies working alongside established pharma organisations that have the clinical infrastructure, regulatory expertise and capital to advance candidates.
• Data infrastructure is a differentiator. AI models are only as good as the data they are trained on, and companies that invest in high-quality, well-annotated datasets are more likely to attract premium partnerships.
• Expectations around cost reduction should be tempered. AI is not yet a shortcut to cheaper drug development. It is a tool for improving early-stage decision-making and reducing the probability of late-stage failure, which is where most R&D waste occurs.

Executives who frame AI investments in terms of pipeline acceleration and risk reduction, rather than immediate cost savings, are more likely to make sound strategic choices. The next 12 to 18 months will test whether AI-discovered molecules can maintain their early advantage through Phase II and Phase III trials. The conferences in Barcelona, Boston and Cambridge will showcase new data, but the real measure of progress will be in partnership terms, clinical readouts and the gradual shift from headline deal values to milestone-based economics. AI has earned a place in the drug discovery toolkit. The challenge now is to prove it can change the economics of the industry, not just the odds of early success.