SHENZHEN, China, Aug. 19, 2026 /PRNewswire/ —
Financial Highlights:
- In the first half of 2026, the Group recorded revenue of RMB393.6 million, compared with RMB517.1 million in the corresponding period of last year. The change was primarily attributable to the high base in the corresponding period of last year, which reflected the recognition of an upfront payment of US$51.0 million from a major pipeline licensing project. The collaboration progressed well during the Reporting Period, and the Group received the second payment of US$19.0 million. Excluding this impact, revenue increased by 73.8% year on year.
- Revenue from AI4S Intelligent Solutions amounted to RMB193.5 million, representing a year-on-year increase of 136.4%, with both AI4S Intelligent Robotic Laboratories (Physical AI) and AI4S Intelligent Services maintaining rapid growth.
- In the first half of 2026, the Group recorded a net loss of RMB224.9 million and an adjusted net loss of RMB105.5 million, primarily due to the year-on-year decline in revenue and a 66.0% year-on-year increase in R&D expenses. The increase in R&D expenses was mainly attributable to the Group’s continued investment in its autonomous laboratories, agent systems, pipeline programs under development and multimodal technology platforms.
- As of June 30, 2026, the Group had a total cash balance of RMB8,671.2 million, comprising cash and cash equivalents, bank deposits, the current portion of financial assets at fair value through profit or loss, and restricted cash.
Business Highlights:
As the Group’s technology platforms continued to be deployed and commercially validated across drug discovery and AI4S scenarios, the efficiency of pipeline discovery and advancement improved significantly, driving a number of important business breakthroughs:
- Leading the development of AI4S infrastructure: The industry’s only full-stack AI4S infrastructure system integrating robotic laboratories, Scientific Agents and autonomous synthesis, and among the first to achieve commercial deployment at scale. During the Reporting Period, AI4S Intelligent Solutions scaled rapidly, with revenue increasing by 136.4% year on year.
- The most comprehensive AI drug R&D platform: Established four core technology platforms spanning small molecules, large molecules, peptides and oligonucleotides. Across each field, the Group has accumulated high-quality, standardized proprietary data and used it to train industry-leading generative and predictive models.
- Efficient development of a differentiated and diversified pipeline portfolio: Among the Group’s partnered and proprietary pipelines, three have entered the clinical stage, more than 10 have received IND approval or reached the IND preparation stage, and nearly 10 have reached the preclinical candidate (PCC) stage. By 2027, more than 10 pipelines are expected to be in the clinical stage, more than 10 to have received IND approval or reached the IND preparation stage, and approximately 20 to have reached the PCC stage.
- “Real-world-ready” AI addressing traditional drug R&D bottlenecks: AI has continued to demonstrate its ability to solve complex R&D challenges across multiple drug modalities. These include achieving novel mechanisms of action, optimized therapeutic windows, and breakthroughs in safety and druggability in antibody programs; optimizing target protein degradation activity to picomolar concentrations within one quarter in certain molecular glue programs; identifying hit compounds within two months of target selection in an oral cyclic peptide program; and generating strong non-human primate efficacy data approximately seven months after initiating an oligonucleotide program for IgA nephropathy.
- Strong recognition and broad-based collaboration with leading customers: Continued to deepen collaborations with leading global customers across platform services, model licensing, pipeline transactions and AI4S infrastructure deployment. The Group also entered into a strategic AI drug discovery collaboration with a renowned international biopharmaceutical company with a total potential value exceeding US$400 million, as well as multiple collaborations with domestic innovative pharmaceutical companies.
- Launched a self-evolving AI retrosynthesis system: Reduced the chemical hallucination rate to 4.6%, only one-sixth that of leading large models in the industry. Its top-1 recommended route accuracy reached 74.3%, 2.2 to 3.5 times that of existing specialized and general-purpose models, materially enhancing the reliability and R&D translation efficiency of AI-assisted synthetic route design.
- The world’s first comprehensive open scientific research platform with a closed Physical AI loop: Officially launched the XtalPi Science platform and the Genius Agent suite of Scientific Agents, standardizing and platformizing the Group’s internal scientific research capabilities and building underlying AI4S infrastructure that global industry partners and research institutions can access on demand.
- Building leading biological simulation capabilities: Expanded biological modeling and validation capabilities through investments, incubation and other approaches in virtual cells, human organ-on-a-chip models and organoids, extending R&D capabilities from molecular design to mechanism studies, translational prediction and efficacy validation at the cellular and tissue levels.
Business Overview
As AI rapidly extends from the digital world into scientific R&D and the physical world, AI4S is advancing AI from isolated tools that assist research toward “autonomous scientific discovery” spanning reasoning through validation. The Group was among the first to classify autonomous scientific discovery into five levels: L1 Tools, L2 Co-Pilots, L3 Agents, L4 Domain-Specific Autonomous and L5 General Autonomous AI4S. Drawing on extensive real-world project experience and accumulated capabilities, the Group has achieved end-to-end L4 autonomous discovery across multiple R&D scenarios.
Through unified agent-based orchestration of scientific models and automated experimental facilities, the Group has established an R&D closed loop encompassing task planning, experimental validation and iterative feedback. Agentic HTE can autonomously match experimental conditions, generate protocols, orchestrate automation, analyze results and plan subsequent experiments. Agentic Synthesis connects the full workflow from raw-material verification and project creation through experimental-condition generation and automated execution. Drawing on more than a decade of AI4S R&D and industry experience, the Group has built integrated scientific research infrastructure connecting digital R&D with real-world physical experimentation, which continues to be validated, iterated and upgraded through real-world projects.
Technology Engine: An Industrial-Grade AI R&D Paradigm Connecting the Digital and Physical Worlds
Across the full workflow of complex scientific research tasks, and through sustained execution of internal R&D and external service projects, the Group has progressively established a four-layer core technology architecture comprising Genius Agent, Scientific AI, Physical AI and the Data Moat. This industrial-grade, integrated scientific research infrastructure has been continuously operated, validated and iterated within real-world industrial R&D workflows.
- Genius Agent (Intelligent Hub): As the core orchestration hub and R&D matrix, Genius Agent establishes a multi-agent system combining global planning with scenario-specific execution. It centrally orchestrates scientific models, specialized tools, R&D workflows and data resources, while using project context to continuously advance long-horizon R&D.
- Scientific AI: The Group continues to develop AI capabilities spanning different molecular modalities and specialized tasks, including small molecules, large molecules, peptides and oligonucleotide therapeutics. In an evaluation involving 350 real-world industrial molecules, SureRoute, its self-evolving AI retrosynthesis system, reduced the chemical hallucination rate to 4.6%, only one-sixth that of leading large models in the industry, and achieved a top-1 recommended route accuracy of 74.3%.
- Physical AI: Centered on the Group’s proprietary Intelligent Robotic Laboratories, Physical AI translates experimental plans generated by scientific models and agents into standardized, automated and traceable experimental workflows. The Group has deployed more than 300 automated workstations worldwide, covering over 20 types of R&D scenarios.
- Data Moat (High-Quality Data Foundation): Integrates public scientific data, proprietary R&D data and real-world experimental data generated by Physical AI to create traceable data assets encompassing experimental results, process parameters and failed experiments. The system has supported more than 100 drug and advanced materials discovery projects, generating over 50,000 reaction-yield data points and 300,000 process data points each month. It has accumulated more than 500,000 real-world experimental records, approximately 80% of which are negative results from failed experiments that are relatively scarce in published literature.
Business Model: A Deeply Integrated Business Portfolio
Built on its R&D system connecting the digital and physical worlds, the Group has established a dual-engine business model centered on Drug Discovery Solutions and AI4S Intelligent Solutions (AI4S Infrastructure), combining recurring cash flow with the potential for asset value realization.
- Drug Discovery Solutions are built around the Group’s AI-driven drug discovery capabilities and encompass platform-based collaboration services and the out-licensing of proprietary assets. Platform-based collaboration projects provide R&D services by leveraging the Group’s AI drug discovery capabilities, while proprietary pipeline assets may generate upfront licensing payments, milestone payments and potential royalties through out-licensing, co-development and other arrangements.
- AI4S Intelligent Solutions provide customers with AI4S infrastructure comprising AI4S Intelligent Robotic Laboratories (Physical AI) and AI4S Intelligent Services. The Intelligent Robotic Laboratories support standardized, high-throughput experimental execution and systematically generate high-quality, traceable experimental data. Leveraging innovative molecular building blocks, the VAST Virtual Compound Library and a high-throughput autonomous synthesis platform, AI4S Intelligent Services enable chemical-space expansion and rapid validation from molecular design through physical synthesis.
The two businesses operate in deep synergy to create a self-evolving closed loop. Drug Discovery Solutions continuously generate high-quality data, demand for experimental validation and momentum for technological iteration, driving capability upgrades in AI4S Intelligent Solutions. In turn, AI4S Intelligent Solutions provide efficient, reusable R&D infrastructure for drug discovery. Together, they form a self-evolving closed loop of “scenario-driven development, experimental validation, data feedback and capability evolution.” The underlying capabilities have also been extended to advanced materials, consumer health and other fields.
Outlook
The AI4S industry, particularly AI-driven drug discovery (AIDD), is in a period of rapid growth. The rapid expansion of AI-driven drug discovery is driving significant demand for novel molecule synthesis and R&D data. Leveraging its AI-native experimentation system and leading intelligent laboratories, the Group is capturing incremental orders from leading pharmaceutical companies. Over the medium term, the Group can generate stable revenue through platform-based technology services while monetizing high-quality proprietary pipeline assets through pipeline transactions, creating dual growth drivers from platform services and asset monetization. Over the medium to long term, as proprietary pipelines advance toward regulatory filings and clinical development, the Group is expected to establish a dual-engine growth model combining R&D services with proprietary drug development. Over the long term, its end-to-end intelligent R&D system will continue to improve efficiency, reduce costs and increase success rates, while strengthening its proprietary data and algorithmic advantages.
Business Progress
Drug Discovery Solutions: Platform Technologies Accelerating Pipeline Asset Development and Commercialization
During the Reporting Period, revenue from Drug Discovery Solutions amounted to approximately RMB200.1 million, compared with RMB435.2 million in the corresponding period of last year. The change was primarily attributable to the high base in the corresponding period of last year, when an upfront payment of US$51.0 million from a major pipeline licensing project was recognized. The collaboration progressed well during the Reporting Period, and the Group received the second payment of US$19.0 million. The Group continued to advance its multimodal drug technology platforms and develop proprietary innovative drug pipelines with substantial clinical value and commercial potential.
Drug R&D Platforms and Pipeline Progress: Broad-Based Advances across Therapeutic Modalities, Accelerating Clinical Translation
The Group has established a systematic technology portfolio covering key drug modalities, including small molecules, large molecules, oligonucleotides and peptides, spanning target understanding, molecular design, function prediction, candidate optimization and experimental validation. AI has been deeply integrated into R&D scenarios including the discovery of molecular glue candidates with high degradation activity, remediation of protein aggregation and immunogenicity optimization for large molecules, and oligonucleotide drug design and personalized modification recommendations. Among the Group’s proprietary and partnered pipelines, three have entered the clinical stage, more than 10 have received IND approval or reached the IND preparation stage, and nearly 10 have reached the PCC stage. By 2027, more than 10 pipelines are expected to be in the clinical stage, more than 10 to have received IND approval or reached the IND preparation stage, and approximately 20 to have reached the PCC stage. The pipelines span oncology, autoimmune diseases, metabolic and chronic diseases, neurological disorders and consumer health, representing potential markets worth hundreds of billions of U.S. dollars.
The small-molecule R&D platform encompasses AI-powered computational prediction, physics-constrained modeling, synthetic route planning and closed-loop experimental validation. The XGlue™ platform for molecular glue discovery has built a virtual compound library containing millions of compounds and a physical scaffold library containing tens of thousands of scaffolds, while its experimental operations can synthesize and validate hundreds to more than 1,000 compounds each week. The Group has established multiple molecular glue programs for autoimmune diseases and identified hits against multiple targets. For certain programs, target protein degradation activity was optimized to picomolar concentrations in approximately one quarter. The Group plans to advance the relevant projects into the preclinical stage in 2027. If their druggability and differentiated advantages are subsequently validated, these assets may realize value through co-development or out-licensing.
The Group’s proprietary TRK/RET dual-target small-molecule candidate demonstrated low-nanomolar inhibition of both targets at the protein level, together with strong selectivity and gut-restricted properties, resulting in an excellent safety window. Intended for gut pain-related indications including irritable bowel syndrome and inflammatory bowel disease, it is the world’s first candidate targeting this dual-target combination to be filed for clinical development (First-in-Class). The program has completed submission of materials for a U.S. pre-IND meeting, and the Group expects to submit IND applications in both the United States and China in the second half of 2026.
- Multiple Partnered Pipelines Continued to Advance toward Clinical Translation:
SIGX1094, a dual FAK/SRC inhibitor discovered in collaboration with Signet Therapeutics, has received IND clearance in both China and the United States and is undergoing a Phase I clinical trial at Peking University Cancer Hospital. Preliminary results have shown a favorable safety profile and antitumor signals. It has also received orphan drug designation and fast track designation from the U.S. FDA. The CDE has accepted an application for a Phase II/III clinical trial of SIGX1094 in combination with Innovent Biologics’ approved KRAS-G12C inhibitor fulzerasib tablets for KRAS-G12C-mutated non-small cell lung cancer. SIGX2649, a pan-TEAD inhibitor being developed with Signet Therapeutics, has received IND clearance in both China and the United States. A Phase I clinical trial is planned to begin as early as the second half of 2026.
RTX-117, an eIF2B small-molecule activator enabled for ReviR Therapeutics, has received clinical trial approvals in China and the United States for Charcot–Marie–Tooth disease and clinical trial approval in China for vanishing white matter disease. A Phase I trial in healthy adults is underway, and a Phase II trial is expected to begin in the first half of 2027.
The world’s first oral small-molecule inhibitor of LDH (lactate dehydrogenase), being developed in collaboration with Meta Pharmaceuticals, has entered the IND application stage, with IBD as its lead indication.
PEP08, a next-generation PRMT5 inhibitor being developed in collaboration with PharmaEngine, has begun enrolling patients with solid tumors. The parties have also initiated a second AI drug discovery program targeting a novel synthetic lethal target.
A high-value, tumor-agnostic asset being developed with DoveTree has entered the IND stage. Preliminary biological activity validation demonstrated a clear target intervention effect and an excellent selectivity window. The parties will further deepen their collaboration on the agreed difficult-to-drug targets, advance R&D efforts including those involving molecular glues, and accelerate the clinical translation of drug candidates.
The Group’s large-molecule platform, Ailux, is a globally leading AI-native antibody drug development platform and has established collaborations with multinational pharmaceutical companies (MNCs) across its models, platforms and assets. Ailux’s core strength lies in the deep integration of models, data and wet-lab experimentation.
Models: Ailux is powered by three core engines—the XtalFold® structural modeling platform, the XenProT® generative AI platform and the Xentient® discriminative AI platform—covering the full large-molecule R&D workflow from structure prediction and molecular generation to function assessment and candidate optimization. The platform has been validated across more than 100 internal and external projects.
Data: The proprietary AtlaX™ data foundation builds data resources through proprietary wet-lab systems, high-throughput data generation and the LuxSight™ patent-mining agent. Across key data types including antibody affinity, antigen–antibody complex structures, antigen–antibody pairing and native heavy- and light-chain sequences, AtlaX™ offers a scale advantage ranging from several-fold to tens of times that of public datasets.
Wet Lab: The platform uses proprietary experimental workflows to generate functional labels for polyreactivity, stability, immunogenicity and other properties that are difficult to capture in public datasets, creating a differentiated data moat for complex antibody drug development.
The Group appointed Dr. Maria G. Belvisi as Chief Scientific Officer. Dr. Belvisi brings experience across multinational pharmaceutical companies and international academia and has more than 30 years of leadership experience in drug R&D and academia. She spent approximately 10 years at AstraZeneca, where she served as Senior Vice President of Respiratory and Immunology in BioPharmaceuticals R&D.
The Group is advancing three large-molecule programs for autoimmune diseases, all of which are expected to enter Phase I clinical trials in 2027: ALX001, a bispecific antibody targeting TL1A and IL-23p19 for inflammatory bowel disease; ALX002, a T-cell engager targeting CD19 and BCMA for B-cell-mediated autoimmune diseases including systemic lupus erythematosus and rheumatoid arthritis; and ALX005, a long-acting FcRn-blocking antibody for pathogenic IgG antibody-driven autoimmune diseases including myasthenia gravis and immune thrombocytopenia.
PepiX™ integrates precision AI design, automated synthesis and high-throughput wet-lab screening to create an efficient dry- and wet-lab closed loop. The platform has established a proprietary database containing more than 5,000 unnatural amino acids, and its core HELM-DIFF model is used to generate, screen and optimize complex peptide molecules. Tensotide™, a peptide developed using PepiX™, has achieved self-affirmed GRAS status in the United States and may be used in food and dietary supplement products in the United States. The brain-delivery program has entered in vivo animal testing and optimization and is expected to reach PCC in the first half of 2027. An oral cyclic peptide program for autoimmune indications has entered the hit-to-lead stage and is expected to achieve PCC by mid-2027.
Kodexia™, the Group’s siRNA drug development platform, integrates first-principles-driven biological mechanism modeling, generative AI and high-throughput automated experimentation. It has accumulated tens of thousands of wet-lab data points and established an siRNA chemical modification database. Based on publicly comparable metrics, the platform has more than doubled R&D efficiency relative to conventional methods and improved molecular property prediction accuracy by approximately 266%. Across multiple pipelines, more than 50% of molecules from the first design round demonstrated better in vivo activity than positive controls. The platform has built six siRNA programs spanning IgA nephropathy, metabolic diseases and central nervous system disorders; more than half have completed in vivo efficacy evaluations, and the most advanced program has reached the PCC stage. The lead IgA nephropathy program generated non-human primate efficacy data approximately seven months after initiation and demonstrated better potency and durability than a clinical-stage reference molecule against the same target. The Group plans to commercialize the relevant programs through joint R&D, asset co-development, out-licensing and other models.
- Biological Simulation Platforms:
Through investments, incubation and other approaches, the Group has expanded into virtual cells, human organ-on-a-chip models and organoids.
Virtual Cells: OCOO-T, an AI virtual cell model developed by XtalPi-incubated company INFevo, achieved state-of-the-art performance across three perturbation benchmarks covering chemical compounds, genes and cytokines. INFevo also launched The Popper Project (TPP), a scientific discovery engine. In the first half of 2026, INFevo completed an angel financing round raising tens of millions of RMB, with participation from Shunwei Capital, Sequoia China and Green Pine Capital Partners.
Organ-on-a-Chip: XtalPi-incubated company Xellar Biosystems completed delivery of Sanofi’s iDEA-TECH project and collected the final payment in full. Its AI-powered toxicity prediction system jointly developed with Pfizer also achieved a key delivery milestone. As one of the first companies to participate in the CDE’s multicenter collaborative validation of new approach methodologies (NAMs), Xellar Biosystems was selected for the “Pioneer Program” and is also advancing model qualification under the U.S. FDA’s ISTAND program. Xellar Biosystems completed a RMB400 million Series A financing in the first half of 2026 and is currently advancing its Series B financing.
Organoids: The organoid-plus-AI platform developed in collaboration with Signet Therapeutics has established more than 15 gene-edited tumor organoid models. It has also combined AI with normal organoids of the heart, kidney and liver to establish drug toxicity prediction and evaluation models. Its drug cardiotoxicity prediction model based on cardiac organoids achieved an accuracy rate of 81.25%, compared with 43.75% for conventional methods.
Key Commercial Progress: Multiple Major Collaborations, Further Diversifying Monetization Models
- The Group entered into a strategic AI drug discovery collaboration with a renowned international biopharmaceutical company that has a broad pipeline and multiple commercialized products. The collaboration has a total potential value exceeding US$400 million. The parties will jointly develop a potentially best-in-class innovative oral small-molecule drug against a GPCR target. The partner will pay an upfront payment and fund all early-stage R&D expenses. The Group will also be eligible to receive preclinical, clinical and commercial milestone payments, as well as royalties on future sales.
- The Group entered into a key collaboration with Visen Pharmaceuticals, integrating XtalPi’s AI-driven robotic drug R&D platform with Visen Pharmaceuticals’ expertise in endocrinology to focus on indications with high clinical value and innovative targets in endocrinology and metabolic diseases, and to jointly advance the early discovery and clinical translation of innovative therapies.
- The Group signed a strategic collaboration agreement with Sunshine Lake Pharma. The parties intend to establish a joint venture to jointly develop an AI-driven robotic laboratory and a foundation model for preclinical drug development, and to collaborate on underlying technologies, innovative drug pipelines and commercialization. Sunshine Lake Pharma is expected to invest several hundred million RMB. The two companies aim to build an industry-leading AI drug discovery engine, bring the technology to international markets and establish a diversified monetization model centered on “pipeline co-creation and shared success through technology.”
- The Group received the second payment of US$19.0 million stipulated under the definitive agreement with DoveTree. The parties will continue R&D activities, including work on molecular glues, against the agreed hard-to-drug targets.
- In November 2025, the Group and Gan & Lee Pharmaceuticals entered into a global strategic collaboration and platform licensing agreement for the R&D of innovative AI-designed peptide drugs. The project is progressing. The jointly established “AI-Driven Intelligent Peptide Delivery Laboratory” — Beijing Key Laboratory of Artificial Intelligence for Peptide Drug Design and Delivery Systems — was officially recognized as a Beijing Key Laboratory and inaugurated.
AI4S Intelligent Solutions: Accelerating Platform Value Realization and Delivering Breakthrough Revenue Growth
During the Reporting Period, AI4S Intelligent Solutions generated revenue of RMB193.5 million, representing a year-on-year increase of 136.4%, with both AI4S Intelligent Robotic Laboratories (Physical AI) and AI4S Intelligent Services maintaining rapid growth.
AI4S Intelligent Robotic Laboratories (Physical AI): Embedded into Molecular R&D Workflows to Drive Scalable Growth
During the Reporting Period, the Group’s AI4S Intelligent Robotic Laboratory business made progress in both overseas and domestic markets. Overseas, the compound management system for Eli Lilly was contracted and delivered, while the first HTE system completed factory acceptance testing (FAT) and user training. The intelligent autonomous drug synthesis and process R&D system developed with JW Pharmaceutical was fully delivered in April. In China, the Group secured multiple projects valued at tens of millions of RMB, while its intelligent synthesis workstation, a semi-standardized product, was replicated and deployed across 13 customers. These capabilities have also expanded into advanced materials fields including perovskites, lithium batteries and molecular sieves.
During the Reporting Period, the Group’s AI4S Intelligent Robotic Laboratory business successfully delivered multiple flagship projects:
- The Group signed a compound storage and management system project worth tens of millions of RMB with Eli Lilly and completed delivery at its Shanghai R&D center. The system covers compound storage and retrieval, micro-powder dispensing, and sample preparation and output in DMSO solutions, integrating sample management and preparation on a unified automated platform.
- The Group signed an HTE (high-throughput experimentation) platform collaboration project worth tens of millions of RMB with Eli Lilly. The Group is providing a modular high-throughput experimentation platform centered on a condition-screening glovebox, supporting automated dispensing, reactions, dilution and filtration under anhydrous and oxygen-free conditions and connecting with XtalPi’s Agentic AI algorithms.
- The high-throughput automated synthesis workstation, AI-powered reaction condition optimization system and intelligent analytics platform provided by the Group to JW Pharmaceutical were fully delivered in April. The platform supports JW Pharmaceutical’s R&D requirements in automated drug candidate screening, synthesis and process optimization.
- Domestic and Advanced Materials Projects:
The Group signed a mesoporous materials intelligent high-throughput preparation project worth millions of RMB with Wusong Materials Laboratory of Fudan University; an end-to-end automated perovskite solar cell project worth millions of RMB with a leading university; and high-throughput automated electrolyte preparation and testing platform projects worth millions of RMB with Peking University, the Dalian Institute of Chemical Physics and other institutions.
AI4S Intelligent Services: A Key Advance in End-to-End Autonomous AI Decision-Making, Driving Rapid Order Growth
The Group supports chemical-space expansion through innovative molecular building blocks and the VAST Virtual Compound Library and has established a design–make–test–analyze (DMTA) closed loop through its high-throughput autonomous synthesis platform. The value of new orders signed in the first half of 2026 grew rapidly, including orders for more than 20,000 molecules through the VAST Virtual Compound Library.
During the Reporting Period, the Group developed and commercialized two solutions—Agentic Synthesis and Agentic HTE—enabling autonomous AI decision-making across the entire workflow.
- Agentic Synthesis: Centered on an Agentic System connecting Scientific AI with Physical AI, the autonomous synthesis platform creates a ten-step closed loop from target molecule to final-product delivery. The SureRXN™ synthesizability prediction and condition recommendation module achieved an experimental success rate of over 90%, reducing the average number of experiments to 1.19. The high-pressure separation algorithm achieved an automation rate of 76%, while increasing the first-delivery success rate from 83% to 94%. The LCMS spectral analysis algorithm achieved an overall prediction accuracy of 95%, rising to 98% within the high-confidence range. The NMR spectral interpretation algorithm automatically analyzed more than 70% of spectra across four projects. At the Agentic System level, seven dedicated agents work together across the full lifecycle from project initiation to product shipment.
- Agentic HTE: This end-to-end high-throughput experimentation solution is centered on an Agentic System that connects Scientific AI with Physical AI. Through intent understanding, skill orchestration, long-running task management and human–AI collaboration, it shortens the conventional HTE iteration cycle from three to four weeks to approximately six days.
While strengthening its end-to-end autonomous AI decision-making capabilities, the Group is also expanding its capabilities in upstream chemical-space design and synthesis. In June 2025, the Group completed the acquisition of LCC. Its PACE (Parallel Automated Chiral Engine) platform integrates AI software with automation technologies to virtually screen target molecules from a chiral chemical library comprising hundreds of millions of molecules, followed by automated synthesis and physical testing. The platform has been validated through internal drug discovery projects. LCC is currently in late-stage discussions with major pharmaceutical companies, biotechnology companies and leading research institutions regarding collaborations centered on PACE, with the aim of creating intellectual property and assets against partners’ high-priority targets.
Advanced Materials and Consumer Health: AI4S Capabilities Broadening Applications and Achieving Key Milestones
Building on the systematic validation in drug R&D of the Group’s autonomous AI R&D closed loop comprising Scientific AI, Physical AI and the Agentic System, the Group is extending its core capabilities into advanced materials, consumer health and other scenarios.
Advanced Materials: Intelligent R&D Platform Empowering Materials Innovation and Driving Breakthrough Progress in Perovskite Tandem Cell R&D
The Group has established an advanced materials R&D team. During the Reporting Period, the Group entered into a strategic collaboration agreement with a subsidiary of JinkoSolar to advance AI- and automation-driven high-throughput R&D for tandem solar cells. The parties have established a joint venture to build the world’s first fully closed-loop intelligent manufacturing line for tandem solar cells, integrating “AI-driven decision-making, robotic execution and data feedback.” The project is progressing. The Group has also established an AI- and automated laboratory-driven perovskite formulation R&D platform. Small-area modules achieved a laboratory-tested efficiency of 27.0% and a third-party-certified efficiency of 26.51%, while large-area modules achieved a laboratory-tested efficiency of 23.0% and a third-party-certified efficiency of 22.74%. The high-throughput automated production line for tandem cells is designed for a daily throughput of no fewer than 1,000 cells. Compared with conventional manual R&D, the optimization cycle for each iteration has been reduced from several months to several hours, while the overall R&D cycle has been shortened from four to six years to one to six months.
Consumer Health: Groland Expands Omnichannel Reach and Advances Commercialization
During the Reporting Period, Groland, a combination formulation incorporating two proprietary topical molecules developed by XtalPi to address hair growth and retention, completed market validation and brand development and began building a marketing network spanning domestic and international markets and online and offline channels. The brand operates official stores on Tmall, JD.com and Douyin. Groland’s AquaKine Scalp Serum ranked No. 1 on Tmall’s “New Anti-Hair Loss Scalp Oil Products” chart, while its Tmall store ranked among the top three emerging personal care stores by gross merchandise value during the 618 Shopping Festival. The brand has established a product portfolio spanning pre-shampoo treatments, shampoos, conditioners, leave-in treatments, hair growth serums and red-light hair growth brushes. The two proprietary molecules have completed regulatory filings as new cosmetic ingredients in China, while the regulatory filings and launch of the general-trade versions are expected to be completed by the end of 2026.
Strategic Upgrade: Evolving XtalPi Science into an Open Scientific AI Infrastructure Platform
In July 2026, the Group launched XtalPi Science, its Scientific AI platform, together with the Genius Agent suite of Scientific Agents. XtalPi Science is the world’s first comprehensive AI4S platform integrating large language models (LLMs), Scientific Agents and large-scale automated robotic experimentation. It further standardizes and platformizes the Group’s integrated scientific research infrastructure, validated through real-world projects, transforming it into platform capabilities that can be accessed through a unified interface, orchestrated on demand and continuously expanded. In doing so, XtalPi Science is progressively building a Global Scientific Utility for industry partners and research institutions worldwide, providing a unified entry point for cross-institutional R&D collaboration and the scaled application of Scientific AI.
As the core orchestration hub and R&D matrix, Genius Agent can autonomously understand complex research objectives, break down and advance long-horizon interdisciplinary tasks, and centrally orchestrate domain-specific models, specialized tools, R&D workflows and physical execution infrastructure. The platform generates scientific hypotheses and performs specialized predictions in the digital world, followed by experimental validation in the physical world through Physical AI, completing a closed loop spanning “digital hypothesis generation, specialized prediction, physical validation and data feedback.”
The Group and 26 partners jointly launched the Open Ecosystem Alliance for Scientific AI, whose members span multiple segments of the scientific innovation value chain. XtalPi Science also plans to introduce Science Token as a unified access and metering mechanism for scientific research resources. Taking into account customer needs and different R&D scenarios, the Group will explore diverse platform service and collaboration models as it continues to evolve into an open Scientific AI infrastructure platform.
About XtalPi
XtalPi Holdings Limited (“XtalPi,” HKEX: 2228) was founded in 2015 by physicists from the Massachusetts Institute of Technology (MIT). The company is a technology platform focused on quantum physics-based and AI-driven innovation in drug and materials discovery. By integrating quantum physics, artificial intelligence, cloud computing, and large-scale automation, XtalPi provides research and development solutions and services to global pharmaceutical, materials science, consumer products, energy, and advanced chemicals industries. XtalPi leverages AI Agents, proprietary modeling, and advanced robotics to accelerate scientific discovery through an autonomous paradigm designed to solve the most challenging molecular discovery problems. XtalPi’s team currently spans Shenzhen, Shanghai, and Beijing in China, Boston in the United States and Liverpool in the United Kingdom.
