From "groping in the dark" to "searching for a path with a flashlight".
Sanofi Sweeps Up Chinese AI Again.On January 5, Sanofi announced a new strategic partnership with Earendil Labs. According to the disclosure, Earendil Labs' artificial intelligence drug discovery platform will be applied to multiple autoimmune and inflammatory disease projects, and the bispecific antibody candidates generated from the collaboration will be led by Sanofi for subsequent development and global commercialization.In 2023, after announcing its ALL in AI strategy, Sanofi has frequently engaged in BD deals with global AI + pharmaceutical companies. The reason this deal has drawn industry attention is that just eight months ago, Sanofi entered into a BD collaboration with Earendil Labs, which included an upfront payment of $125 million and a total value exceeding $1.8 billion. Additionally, Sanofi secured the global exclusive rights to two bispecific antibody candidate drugs, directly targeting its core area — one of the most crowded R&D fields globally at present — autoimmune and inflammation.This time, Sanofi has directly advanced the collaboration from "a single pipeline" to "an entire AI platform."Almost in the same time window, Insilico Medicine, a popular player in China's AI pharmaceutical industry, successfully went public by the end of 2025 after several IPO setbacks, setting the largest IPO fundraising record that year. At the beginning of 2026, it announced an $880 million collaboration with Servier, gaining significant attention.Looking further back, multinational pharmaceutical companies such as Eli Lilly, Merck, and Novartis have also frequently disclosed their AI-focused R&D strategies in the past two years.Has AI already transformed from an add-on in pharmaceutical R&D to a foundational capability that must be established? The field that was once most easily questioned as "storytelling," has it truly succeeded in taking off this time?Why Sanofi Is Doubling Down AgainMore than $1.8 billion and more than $2.5 billion: In less than a year, Sanofi has aggressively acquired Chinese AI assets.In April 2025, Sanofi entered into a licensing agreement with Earendil Labs, securing the global exclusive rights to two bispecific antibody drugs, HXN-1002 (α4β7/TL1A) and HXN-1003 (TL1A/IL-23), for an upfront payment of $125 million. Both drug candidates are focused on autoimmune and inflammatory diseases, with TL1A considered one of the most promising yet highly competitive targets in the immunology-inflammation field in recent years.Eight months later, Sanofi has once again chosen Earendil Labs, and this time the deal is no longer limited to a single pipeline.Instead, it directly incorporates Earendil Labs' AI drug discovery platform into the early stages of multiple Sanofi R&D projects.Specifically, Earendil Labs is an affiliate of Huashen Zhiyao, which focuses on AI + high-throughput antibody optimization. Huashen Zhiyao was incubated by the Artificial Intelligence Industry Research Institute of Tsinghua University and has developed various innovative algorithms for protein drug design and modeling. Its AI antibody design platform, Helixon Design, can simulate molecular interactions on a large scale, efficiently design multi-target antibodies, and significantly shorten the R&D cycle.From a technical perspective, the collaboration between Earendil Labs and Sanofi leverages a proprietary platform combining generative AI and experimental validation to efficiently and precisely discover and optimize biologics with "first-in-class" or "best-in-class" potential. Sanofi contributes its global clinical development network and commercialization capabilities in immunology, enabling both parties to jointly accelerate the progression of candidate drugs from the lab to patients.The number of new molecular entities in the preclinical stage is vast, but projects that can truly advance to the clinical stage remain few and far between, especially in areas like antibodies, bispecific antibodies, or programmable protein design, where the combinatorial space of amino acid sequences expands exponentially. In a sense, this differs from the traditional drug discovery approach, which resembles searching for a needle in a haystack.AI can precisely generate drug molecules with potential bioactivity for different diseases, overwhelmingly transforming protein structure prediction, molecular design, and physicochemical property optimization, paving a new path for drug development.In simpler terms, AI has transformed the process from blindly stumbling in the dark to searching with a flashlight.Some analysts believe that Sanofi's willingness to repeatedly choose the same AI platform in a short period of time, and to integrate the AI drug discovery platform into the early stages of multiple R&D projects, indicates that Sanofi is at least satisfied with Earendil Labs' AI R&D platform at this stage.Wanlian Securities Research Report Believes That Pharmaceutical Companies' Willingness to Pay High Upfront Fees Indicates Increased Credibility of AI-Generated Molecules."Insanely" betting on AI is not just Sanofi, in the past month, Eli Lilly has also made consecutive moves:From the introduction of an AI-driven bispecific antibody engineering platform, to a new round of AI drug development cooperation around innovative targets, to the successive implementation of multiple Biotech projects utilizing artificial intelligence technology—all within a mere 10 days, over 3 billion yuan was spent.If we broaden the perspective to the entire industry, the decade-long, multi-billion-dollar marathon of new drug development in the pharmaceutical sector is being redefined by artificial intelligence.When the R&D cycle of traditional pharmaceutical companies still lingers at an average of 10 years, AI technology has compressed the early R&D processes of some giants to just 18 months. In the past few years, MNCs such as Eli Lilly, Merck, Novartis, and Pfizer have almost all repeatedly emphasized the strategic importance of AI in their R&D systems on various occasions. This shift is not a sudden enthusiasm for a particular track but more like a release of the pressure brought by uncertainties under the traditional R&D model.On the one hand, whether in oncology, autoimmune diseases, or metabolism, mainstream fields have long entered a highly crowded phase. While there seem to be more and more targets, fewer can truly deliver differentiated efficacy in clinical settings or address yet-uncured disease areas. On the other hand, the cost of clinical failures continues to rise, compounded by performance pressures, making "trial and error" increasingly unaffordable.In this context, the expectation for AI is not to "invent a new drug," but rather a more pragmatic goal — improving efficiency.This is also why multinational pharmaceutical companies' attitude towards AI has gradually shifted from "peripheral pipeline cooperation" to "embedding AI into the R&D system."What's Missing for AI + Pharma to Take Off?Entering 2025, generative artificial intelligence is becoming one of the powerful accelerators driving the development of the life sciences field, significantly compressing the R&D cycle that traditionally required decades and billions of dollars. However, if we turn back the clock a few years, the AI pharmaceuticals sector was once the track most easily questioned for "telling stories."Tracing back to the early days, the main role of AI was still in data analysis and drug screening.But truly AI-led discoveries of new targets are still few, with more projects concentrating on old targets and well-established indications.At the same time, the long R&D cycle and unclear profit path also make the market remain cautious about the commercial value of AI drug development.But now, it has deeply integrated into the entire process of target discovery, molecular design, clinical trial optimization, and even precision treatment.Xiangcai Securities Research Report Points Out: From the pharmaceutical process perspective, AI technology has already found suitable application scenarios in multiple stages and demonstrated tremendous potential.Globally, biopharmaceutical companies using AI as a research foundation are emerging, with algorithms replacing test tubes and models predicting molecular fate, redefining the speed and boundaries of drug discovery.Focusing on China, by the end of 2025, the number of global AI pharmaceutical companies has exceeded 350.Among them, there are at least 101 in China, covering several key therapeutic areas such as oncology, infectious diseases, and immunomodulation.Many AI + pharmaceutical companies in China emphasize the "platformization + industrialization" approach, from target discovery and candidate design to clinical progress, striving to shorten the cycle from algorithm validation to commercial implementation. In the face of increasingly fierce homogenization competition in the innovative drug R&D field, the numerous fundamental innovations brought by AI technology in the drug development process have demonstrated tremendous value.An investor proposed that there are currently two business models in AI-driven drug discovery: one is like Insilico Medicine, akin to "gamblers betting on hidden jade stones," and the other is like XtalPi, acting as "suppliers of mining tools."As popular players in China's AI pharmaceuticals sector, the two companies are taking different paths. XtalPi, which went public earlier, acts as an "infrastructure builder." Its quantum physics algorithms can complete crystal form screening within four weeks, serving 16 of the top 20 global pharmaceutical companies. Its financial report for the first half of 2025 is also quite impressive: achieving a revenue of 517 million yuan in the first half of the year, representing year-over-year growth of over 400%, and realizing its first-ever half-year profit, with an adjusted net profit reaching 142 million yuan. The business model is becoming increasingly clear.After several twists and turns, Insilico Medicine, which just landed on the Hong Kong Stock Exchange on December 30, 2025, has been described as an "explorer." Its Pharma.AI platform compresses the drug discovery cycle from 4 years to 18 months, with an average of 80 molecules synthesized to lock in a candidate compound.The enthusiasm and prospects of the industry for AI pharmaceuticals are also reflected in the secondary market, with Insilico Medicine's public offering receiving 1427 times oversubscription, setting a record for the year. In addition to the market's pursuit of AI pharmaceuticals, on the other hand, it also confirms that Insilico Medicine's business model of "software subscription + pipeline licensing + joint development" is being recognized.The road to Insilico Medicine's repeated attempts at an IPO and eventual successful listing precisely reflects the changing perception of AI drug development in the capital market.Early skepticism mainly focused on three aspects: whether the revenue was sustainable, whether the pipeline was genuine, and whether the value of AI could be quantified. To some extent, these questions were not aimed solely at Insilico Medicine but were common challenges for the entire sector. The turning point came after Insilico Medicine’s pipeline entered clinical trials. As AI-generated drug candidates began to be tested in the real world, the market's evaluation system also started to shift. Investors no longer asked, "What can AI do?" but instead began to focus on "What has AI already achieved?"On January 5, Insilico Medicine announced a multi-year R&D collaboration with Servier, with a total amount of $888 million. Insilico Medicine is eligible to receive upfront and near-term R&D milestone payments of up to $32 million and will lead the discovery and development of potential drug candidates meeting predefined criteria using its AI technology platform; Servier will share the R&D costs and lead the subsequent clinical validation and commercialization process.But beneath the hype, AI + pharmaceuticals still has boundaries that cannot be crossed.On the one hand, the advantages of AI in target discovery and disease modeling applications, as well as target-based and phenotype-based drug discovery, have been validated; but on the other hand, and more crucially, when drug development enters the clinical validation stage, AI still seems to be in the process of "proving itself." As mentioned in agreements such as Sanofi's with Earendil Labs and Servier's collaboration with Insilico Medicine:Clinical trials are conducted by Pharma companies.A CEO of a traditional pharmaceutical company once asked, "Can AI help the industry skip animal testing in drug development, or even bypass human clinical trials altogether?" The answer is obviously no, at least not in the short term.Large pharmaceutical companies are exploring the use of AI to design better clinical trials, such as assisting with patient recruitment and data analysis, rather than replacing traditional clinical trials.In June last year, Merck launched a new generative artificial intelligence platform aimed at streamlining the clinical study report (CSR) drafting process, reducing the initial draft writing time from two to three weeks to three to four days; Sanofi Ventures made a strategic investment in QuantHealth, a company driving clinical trial simulations with artificial intelligence, bringing the total financing to 30 million US dollars.AI-Driven Clinical Trials: Leveraging AI technology to enhance the design, execution, monitoring, and analysis of clinical research with the aim of improving efficiency, reducing costs, enhancing patient safety, and generating deeper mechanistic insights. It can be said that AI can improve probabilities but cannot eliminate uncertainties; it can accelerate decision-making but cannot bypass the inherent complexity of biology. Animal experiments, human trials, and real-world data—these steps cannot yet be replaced by algorithms.Second Review | Li Fangchen
Third Review | Li Jingzhi
