Insilico Medicine Signs $2.5 Billion AI Drug Discovery Deal With SK Biopharmaceuticals for CNS Disorders

The most expensive problem in medicine may be the first step. Identifying the right target for a drug — the precise biological mechanism that, if disrupted, would change the course of a disease — takes years of experimental work in traditional pharmaceutical development, and most attempts fail. Insilico Medicine’s bet is that a generative AI platform can compress that first step radically, and on June 22 it signed its biggest single deal yet to prove it.

Insilico Medicine and SK Biopharmaceuticals announced a research and development collaboration worth up to $2.5 billion at the BIO 2026 International Convention, covering the AI-driven discovery and preclinical development of novel drug candidates targeting neuroimmune disorders of the central nervous system. Insilico will contribute its Pharma.AI platform across target validation, generative chemistry, and molecule optimization; SK Biopharmaceuticals will steer late-stage development and commercialization, drawing on the CNS expertise it built through XCOPRI, the epilepsy drug it became the first Korean pharmaceutical company to independently develop and commercialize in the United States.

The deal is the largest by total potential value that Insilico has signed with an Asia-Pacific partner to date, and it extends a deal-making run that has seen the company sign agreements with Eli Lilly ($2.75 billion, March 2026), Servier, Sanofi, Exelixis, and Fosun Pharma, as the commercial bet on AI-native drug discovery moves from early-adopter positioning to something that looks more like a consensus among large pharmaceutical companies.

What Happened

According to the companies joint announcement, Insilico will apply Pharma.AI — the platform’s three core modules covering target discovery and validation, generative molecular design and chemistry, and molecule optimization — to identify and advance new drug candidates for neuroinflammatory, neurodegenerative, and rare neurological diseases. The collaboration targets the neuroimmune intersection of CNS: conditions where the brain’s immune signaling is dysregulated in ways that contribute to inflammation, neurodegeneration, or rare disease progression, a category that has historically had among the lowest clinical success rates in drug development.

Financially, Insilico will receive up to $18 million in upfront and near-term milestone payments — a figure that is a small fraction of the $2.5 billion headline because, as is standard for this type of pharmaceutical partnership, the vast majority of the potential value is contingent on achieving development, regulatory, and commercial milestones that may or may not materialize depending on whether the resulting drug candidates succeed in clinical trials and reach the market. SK Biopharmaceuticals will contribute its clinical development and commercialization capabilities for the programs Insilico’s platform generates. Both companies will be responsible for their own costs associated with their respective contributions.

The deal was confirmed by RTT News on the day of the BIO 2026 announcement, with SK Biopharmaceuticals’ president and chief executive Donghoon Lee describing the collaboration as expanding the company’s growth “beyond epilepsy into new CNS therapeutic areas, building on the deep CNS expertise we have established through the successful development and commercialization of XCOPRI.” Insilico co-CEO Feng Ren framed the deal around the AI platform’s ability to unlock previously intractable therapeutic targets, saying the companies aim to deliver “breakthrough therapies, spanning both traditional small molecules and advanced new modalities.”

The Mechanism: How Pharma.AI Works

Insilico’s platform is organized around the three steps where time is most commonly lost in conventional drug discovery: identifying the right biological target, designing a molecule that will interact with it in the desired way, and optimizing that molecule until it is ready for preclinical testing. Traditional approaches to each step are experimental — testing thousands or tens of thousands of compounds against a target to find one that works, then running further rounds of synthesis and testing to optimize it. Pharma.AI treats each of these as a generative problem instead of a screening problem: using large language models and deep learning trained on biological data to generate candidate molecules with specified properties, rather than searching through pre-existing libraries for the closest match.

The performance claims Insilico publishes for this approach are striking in the context of pharmaceutical norms. The company says it has reached preclinical candidate nomination — the point at which a molecule is selected for rigorous preclinical safety and efficacy testing before entering human trials — in an average of 12 to 18 months per program, compared with 2.5 to 4 years through conventional early-stage discovery. It has done so while synthesizing and testing between 60 and 200 molecules per program, rather than the tens of thousands that standard high-throughput screening involves. Since 2021, Insilico has nominated 31 preclinical candidates; 13 of those 31 have received IND approval or clearance, meaning they have been cleared to enter human clinical trials.

The neuroimmune CNS category this deal targets is particularly difficult by both conventional and AI drug discovery standards, because many of the relevant disease mechanisms involve complex interactions between the brain’s own immune cells, the blood-brain barrier that makes systemic drug delivery difficult, and target biology that remains incompletely understood. That combination — important unmet need, hard biology, historically high failure rates — is both the reason the area has attracted few approved drugs and the reason it is an attractive testing ground for an AI platform that claims to navigate difficult target-and-molecule search spaces faster than conventional methods can.

The Backstory

Insilico’s deal-making pace in 2026 reflects an acceleration in large-pharma appetite for AI drug discovery partnerships that has built gradually since the company first brought an AI-designed molecule into human clinical trials in 2021. A March 2026 agreement with Eli Lilly gave Lilly exclusive global rights to manufacture and commercialize a range of oral therapies developed using Insilico’s AI, in a deal valued at up to $2.75 billion with a $115 million upfront payment — the largest single partnership Insilico had signed before this week, and one of the largest AI drug discovery deals in the industry’s short history. The SK Biopharmaceuticals deal announced today is Insilico’s largest APAC deal and sits in an ongoing string of collaborations that also includes Sanofi (oncology), Servier (oncology, worth up to $888 million in a multi-year deal), and Fosun Pharma (market entry for China). This wave of billion-dollar AI licensing deals in pharma mirrors the broader pattern of large-scale AI investment commitments sweeping multiple industries in 2026, as established companies deploy capital to acquire AI capabilities faster than they could build them internally.

SK Biopharmaceuticals is not new to CNS drug development. Its XCOPRI (cenobamate) epilepsy drug is the benchmark clinical and commercial achievement the company is building on — a genuinely novel mechanism of action approved by the FDA in 2019 and commercialized in the US, South Korea, and Europe, representing the first time a Korean pharmaceutical company had independently taken a new molecule from discovery through to US market approval. That track record in clinical execution for CNS drugs is precisely the capability Insilico lacks as a discovery-stage company, and it is what makes the division of labor in this partnership structurally plausible: Insilico generates candidates faster than the traditional approach; SK Biopharmaceuticals then applies the clinical development rigor those candidates need to have any chance of reaching patients.

Reactions

Insilico’s co-CEO Feng Ren described the collaboration as targeting a category of disease that traditional pharmaceutical approaches have consistently failed to crack: “Neuroimmune disorders represent one of the most underserved and scientifically complex areas in drug discovery. By uniting Insilico’s AI-driven target-to-candidate engine with SK Biopharmaceuticals’ deep CNS mastery, we aim to unlock breakthrough therapies … to address critical patient needs.”

SK Biopharmaceuticals’ CEO Donghoon Lee emphasized the commercial logic as much as the scientific ambition: the deal extends SK Biopharmaceuticals beyond epilepsy, where its XCOPRI revenue base has been concentrated, into a broader set of neurological disease categories without requiring the company to build its own drug-discovery infrastructure from scratch. Rather than hiring computational biology teams and building its own AI platform, SK Biopharmaceuticals is effectively licensing Insilico’s discovery engine for the price of contingent milestones — a lower upfront investment than replicating that capability internally.

The Dispute: $2.5 Billion or $18 Million?

The headline figure of $2.5 billion requires a careful reading that press releases are rarely designed to encourage. Insilico will receive up to $18 million in the near term; the remaining $2.48 billion is contingent on a cascade of development, regulatory, and commercial milestones that, in aggregate, assume the collaboration produces multiple drug candidates that successfully complete preclinical testing, survive multiple phases of clinical trials, gain regulatory approval in major markets, and achieve significant commercial sales. Each of those steps has a materially lower-than-100-percent probability of success — and neuroimmune CNS drugs, which sit in one of the historically highest-failure-rate therapeutic categories in medicine, have lower base-rate success probabilities than most.

The $2.5 billion figure is the deal’s maximum possible value under the most optimistic scenario where all milestones are hit; the expected value, accounting for typical clinical attrition rates in neuroscience programs, is considerably lower. That is not unusual for pharmaceutical partnerships structured this way — the entire milestone-based deal model exists precisely to let partners share risk in proportion to actual outcomes rather than pay upfront for programs that may fail — but it does mean the headline number functions as a statement of maximum ambition rather than a description of what either company will actually pay or receive. The more meaningful near-term figure is the $18 million upfront, and the more meaningful long-term signal is whether Insilico’s platform can generate candidates that advance through the clinic at the rates the company claims.

What Happens Next

The collaboration begins now but its first concrete deliverable — a nominated preclinical candidate from Pharma.AI for a neuroimmune target — is likely 12 to 18 months out based on Insilico’s disclosed average program timelines. That milestone, if reached, would be the first real-world test of whether the platform’s discovery speed holds in the specific biological context of neuroimmune CNS, a target space that comes with additional complexity around blood-brain barrier penetration and disease mechanism validation that not all AI platforms have been tested against at scale. SK Biopharmaceuticals’ commercialization track record means that if Insilico delivers viable candidates, the back half of the pipeline — clinical development and market launch — has more credible execution capacity than many comparable partnering arrangements.

Why It Matters

The Insilico-SK Biopharmaceuticals deal is the latest data point in what is becoming a structural shift in how large pharmaceutical companies approach the early-stage drug discovery problem. The pattern across Insilico’s 2026 deals — with Lilly in oncology and metabolic disease, Servier in oncology, and now SK Biopharmaceuticals in neuroimmune CNS — suggests that the hypothesis being tested is not whether AI can discover drugs generally, but whether it can work in the specific high-difficulty therapeutic areas where conventional drug discovery has historically failed most often. CNS and neuroimmune disorders sit precisely at that frontier: enormous unmet clinical need, complex and not fully understood biology, and decades of late-stage clinical failures that have discouraged traditional investment. The commercial confidence behind deals of this size also reflects a broader enterprise AI trend: studies tracking AI adoption across industries find that the companies leaning hardest into AI investment are pulling away from peers on productivity metrics, a dynamic that is now reaching pharmaceutical R&D as AI-native discovery timelines begin to diverge meaningfully from conventional ones. If Insilico’s platform can produce viable candidates in this category at the speeds it claims, it would be among the most consequential demonstrations of AI’s practical utility in medicine yet seen — and the $2.5 billion headline attached to this deal would begin to look less like a negotiating outcome and more like a reflection of what the upside could actually be worth.

Sources

PRNewswire; RTT News; AI Journal; MobiHealthNews; PharmaTimes.