$20 Billion into AI Drug Discovery, Zero FDA Approvals── The gap between funding and clinical outcomes reveals how partnerships and M&A are actually structured

01$20 billion flowed in over two years, with no FDA approval yet

In 2024, venture investment in AI drug discovery reached 264 rounds totalling roughly $8.9 billion. In 2025, the pace accelerated: 348 rounds raised approximately $11 billion. The two-year total came to about $20 billion. During the same period, not a single drug candidate discovered or designed by AI received approval from the U.S. Food and Drug Administration.

Is this gap abnormal? Drug development takes over a decade. The first wave of large-scale AI drug discovery investment arrived around 2020, and not enough time has elapsed for those programmes to pass through clinical trials. But investors have priced in that delay and continue to deploy capital. The productive question is not "why is there no approval yet" but rather "what structure is the money taking before approval arrives."

02Early-stage clinical data is pulling in the capital

One basis for continued investment is early clinical trial performance. A peer-reviewed study published in 2024 by a BCG research group found that AI-discovered molecules achieved Phase I success rates of 80–90%, well above the 40–65% seen with conventional methods. Phase II success, however, came in at roughly 40%, no different from the historical average.

Phase I primarily tests safety. High success rates at this stage suggest AI excels at predicting toxicity. Phase II tests efficacy, and the unchanged success rate implies AI does not yet reliably forecast complex biological responses in humans. For investors, the practical significance lies in cost efficiency: if failures are caught earlier and more cheaply, the expected value across a portfolio improves.

In July 2026, a compound designed by an AI drug discovery company entered a Phase III trial for idiopathic pulmonary fibrosis. It was the first AI-designed molecule to reach Phase III, making it a clinical milestone that influences investment decisions across the sector.

03Funding is concentrating into fewer, larger rounds

Beyond aggregate totals, individual round sizes have expanded. In April 2024, one AI drug discovery startup announced $1 billion in funding at launch — the largest debut round ever in the field. In March 2025, another AI drug company raised $600 million in its first external financing.

The proliferation of billion-dollar rounds signals that capital is concentrating in a small number of well-positioned companies. According to PitchBook, AI biotechnology investment in 2025 reached its second-highest annual level, but the majority went to the top ten firms. For early-stage startups with less clinical data, the fundraising environment is less accommodating.

04Partnerships run on milestones, and upfront payments are small

When AI drug discovery firms partner with large pharmaceutical companies, headlines cite deal values in the billions. Yet the bulk of these sums consists of milestone payments, and the upfront cash typically amounts to only a few percent of the reported total.

A milestone-based partnership breaks into three layers. First, research milestones, paid when a candidate compound is identified or preclinical work is completed. Second, development milestones, triggered by the start or completion of Phase I, II, or III trials. Third, commercial milestones, activated when post-approval sales cross specified thresholds. In one widely reported deal, the upfront payment was $110 million while development milestones reached up to $1.6 billion and sales milestones up to $3.6 billion. The headline total exceeded $5 billion, but most of that sum depends on clinical and commercial success that has not yet occurred.

This structure means large pharmaceutical companies are placing conditional bets. If clinical results disappoint, payments stop. For the AI company, the economic value of a partnership hinges entirely on clinical and commercial execution.

05M&A is consolidating standalone tool companies into integrated platforms

In 2025, AI drug discovery M&A reached 99 transactions worth $12.3 billion. A defining trend is the merger of AI companies with each other. In 2024, two publicly traded AI drug discovery firms agreed to combine in an all-stock deal valued at approximately $690 million. The stated goal was to build a vertically integrated entity spanning target identification through clinical development.

This consolidation reflects a structural shift. In the early years, many startups applied AI to a single step — molecular design, target identification, or toxicity prediction. But standalone tool companies struggle to generate revenue. Licence fees and SaaS pricing lose attractiveness once pharmaceutical companies build their own AI capabilities in-house. The move toward integration is a strategy to own clinical pipelines and capture the economic upside of successful drugs.

As of mid-2026, the two companies with the largest publicly disclosed clinical pipelines were both integrated players, holding a combined total of more than 40 clinical and preclinical programmes.

06Pharmaceutical accounting changes the visible size of AI investments

Milestone-based partnerships create specific effects in financial reporting. Under IFRS 15's five-step revenue recognition model, variable consideration such as milestone payments is included in the transaction price only to the extent that it is "highly probable" a significant reversal will not occur. Milestones contingent on Phase III success or sales targets are, in most cases, excluded from the transaction price at inception.

This means the "$5 billion partnership" reported in the press largely does not appear in either company's financial statements at signing. For investors reading public filings, estimating the economic value of a partnership requires judging which milestones are close to achievement and assigning probabilities accordingly.

On the AI company side, milestone income arrives at uncertain times and in uncertain amounts, making cash flow forecasting difficult. AI firms that rely on partnerships with small upfront payments must continue raising VC capital or pursue IPOs to fund ongoing development. This dynamic is one reason funding rounds in the sector keep growing.

07The pipeline has surged, but late-stage clinical proof remains scarce

The number of clinical programmes involving AI-discovered or AI-designed drug candidates grew from roughly 24 in late 2023 to over 173 by early 2026 — a sevenfold increase in approximately 18 months. But the vast majority sit in Phase I or earlier. An ASCO/JCO analysis found that as of December 2025, 63 companies had 117 assets in interventional human trials, of which only 8 had completed Phase II and just 1 had reached Phase III.

This distribution confirms that AI drug discovery is still in the stage of generating large volumes of early candidates. The Phase II barrier depends not only on AI prediction accuracy but also on conventional development capabilities: trial design, patient selection, and biomarker strategy. No matter how fast AI designs molecules, the time and cost of clinical trials remain largely unchanged.

08Three distances to measure when reading AI drug discovery investment

Reading investment in AI drug discovery requires measuring three distances. First, the distance between funding and outcomes. $20 billion has been deployed, but FDA approval remains at zero. High Phase I success rates are encouraging; unchanged Phase II rates are sobering. Whether this distance narrows depends on Phase II and III results over the next two to three years.

Second, the distance between headlines and reality. A "$5 billion partnership" is the sum of upfront payments, development milestones, and sales royalties, most of which are contingent on clinical and commercial success. The headline figure and the confirmed economic value at signing differ substantially.

Third, the distance between early pipeline volume and late-stage clinical proof. Over 173 programmes are underway, but only one has reached Phase III. Pipeline counts reflect investor expectations, not approval prospects. McKinsey has estimated that generative AI could create over $50 billion in annual value across pharmaceutical R&D, but realisation depends on success in the clinic.