XtalPi Holdings Announces Full Year 2025 Annual Results

HONG KONG, March 25, 2026 /PRNewswire/ --

Financial Highlights:

Key Business Milestones:

R&D Progress: Deep Integration of AI, Robotics, and Multi -Agent – How the Flywheel Effect is Reshaping the R&D Paradigm

1.  Multi-Agent Matrix: Acting as Intelligent "Project Managers" to Drive Fully Autonomous Workflows

Serving as the central orchestration hub of the development flywheel, a multi-agent system autonomously break down research objectives, coordinate multiple AI models, and manage closed-loop robotic experiments. Leveraging the Group's proprietary AI models, robotics laboratories, and extensive data assets, the Group is building an end-to-end intelligent agent ecosystem that has been deployed across both internal workflows and external application scenarios.

1)     The Group is building an integrated agent system covering the entire R&D workflow, incorporating Agentic Genius, X-Buddy, PatSight, MolAgent, the Vast Agent virtual compound library, synthesis robots, and a robotics orchestration platform, enabling autonomous decision-making across the full process from procurement and experiment execution to reaction optimization, data management and validation. Currently, the Compnay's AI agents independently orchestrate tens of thousands of compound synthesis experiments per week, significantly enhancing throughput and data accumulation efficiency, and establishing a closed-loop R&D system covering small molecule drug design, high-throughput automated synthesis, and cloosed-loop data feedback.

2)     During the Reporting Period, the Group partnered with a leading global pharmaceutical company to deploy an autonomous reaction condition screening system integrating a multi-agent system, robotics laboratories, and chemistry AI models to establish a closed-loop workflow spanning AI-driven condition recommendation, autonomous experimental parameter design, high-throughput robotic execution, and reaction optimization. Trained on hundreds of thousands of experimental data points, the system supports development objectives expressed in natural language, with robotic solutions precisely performing operations such as reagent dispensing, temperature control, and analytical testing, significantly improving data consistency and experimental reliability, while providing a scalable and replicable solution for intelligent R&D transformation.

2.  AI Models: Reinforcing the "Expert Brain" and Expanding into Novel Drugs

By the end of 2025, the Group had developed over 200 industry-specific AI models, covering the entire workflow from target discovery to preclinical candidate screening. During the Reporting Period, the Group's AI models achieved systematic results in deepening strategic collaborations, upgrading core models, and expanding into novel drugs.

1)     The Group continued to deepen industry partnerships and upgrade our AI-driven drug discovery models, building a data closed loop from molecular design to biological testing. By integrating molecular dynamics simulation data as a core input, we have embeded dynamic physical-based principles into our AI models, equipping them to interpret the nature of protein-protein interactions and providing strong support for tackling 'undruggable' targets. In molecular activity prediction, we further advanced our strategic collaboration with Pfizer, launching a next-generation molecular simulation platform in 2025. By combining the precision of physics-based algorithms with the speed and scalability of AI, this platform significantly improves simulation efficiency, accuracy, and chemical space coverage to accelerate small molecule drug discovery.

2)     The Group's self-developed generative AI platform for protein therapeutics, XenProT™, introduced the XMPNN inverse folding design algorithm, achieving world-leading performance in evaluations on public benchmark datasets by leverageing high-quality data and proprietary domain knowledge, and demonstrating its value across multiple challenging pipeline projects, with its bispecific antibody design capabilities recognized by leading global pharmaceutical companies. Meanwhile, XtalFold® launched its Ultra mode, boosting the accuracy of antigen–antibody complex structure prediction by approximately 10 percentage points, and was selected as one of the "Top 10 AI Innovation Technologies (Products)" at WAIC.

3)     Building AI Platforms for Novel Drugs – Establishing Leadership in Emerging Frontiers.

3.  Robotic Laboratory (Physical AI): Building "Precision Hands" to Redefine the Experimental Paradigm

During the Reporting Period, the Group continued to drive technological breakthroughs across its robotic laboratories, achieving several key milestones. "NeoDispenser" design overcomes key challenges in robotic laboratory technology, and has gained strong receognition from leading global customers. Key technical features include:

1)     Advanced Visual Perception: Features cutting-edge visual perception algorithms that analyze powder characteristics in real time, enabling precise identification and handling of different powder types.

2)     Intelligent Database: Integrates a comprehensive powder attribute database with strong learning capabilities that continuously improves powder handling strategies.

3)     Powder Analysis Foundation Model: Powered by a specialized AI model trained and fine-tuned on diverse powder datasets to accurately predict powder properties and behavior.

4)     Seamless Robot Integration: Seamlessly connects with powder aliquoting and other robotic systems, adjusting parameters and control strategies in real-time based on visual feedback.

4.  High-Quality Data Analysis and Accumulation: Continuous "Fuel" for the AI Flywheel

The Group has established an industry-leading biomedical data asset system and an open development data infrastructure platform, providing robust data support for the continuous operation of the AI flywheel.

1)     Multimodal Data Mining Infrastructure: The Group's multimodal Layout model achieves 95.3% accuracy in document recognition. A self-developed domain-specific 32B LLM surpasses the performance of GPT-4o, delivering both "data localization" and cost efficiency advantages. The OCSR optical chemical structure recognition model has reached industry-leading state-of-the-art performance.

2)     High-Quality Chemical Reaction Data Platform: The Group has developed a data-mining agent for the fine-grained extraction of reaction-step-level data with >95% recognition accuracy, having structured hundreds of thousands of patents to build a massive high-quality reaction dataset that underpins AI models and robotic laboratry experiment training.

3)     High-Quality Large Molecule Data Platform: The Group has developed an antibody data- mining agent, achieving >99% extraction accuracy for both conventional and VHH antibodies, and 98% accuracy in affinity recognition. The Group have also built large-scale antibody patent databases with amino acid sequences, forming a robust biological data asset moat.

During the Reporting Period, the Group's data-driven R&D innovation capabilities received national recognition. In 2025, leveraging the "AI + Robotics Empowering Drug R&D Data Element Construction and Application" project, the Group stood out from over 22,000 entries in the national finals of the "Data Elements ×" competition hosted by the National Data Bureau, winning the First Prize.

5.  Consumer Health, New Materials and Agriculture: Strategic Expansion and the Flywheel's Value Extension

The Group's "AI + Robotics+Multi-Agent" flywheel approach has been fully validated for drug discovery. Building on this foundation, its core capabilities are now rapidly expanding to areas such as consumer health and new materials, evolving from capability replication to strategic expansion.

1)     Expansion into Consumer Health. Both innovative topical molecules for hair growth and hair retention, developed in-house by the Group,  have completed INCI registration, and the combined product Groland formulation has completed U.S. FDA cosmetic facility registration and product listing. The product currently ranks first on Tmall Global's new arrivals list within its category and has been featured by several leading fashion media outlets.

2)     Driving a Paradigm Shift in Photovoltaics Development. In January 2026, the Group signed a strategic cooperation agreement with a subsidiary of JinkoSolar (688223.SH) on AI and automation-driven high-throughput R&D for tandem solar cell, jointly establishing the world's first fully closed-loop intelligent manufactureing line for tandem solar cells, intergrating AI-driven decision-making, robotics execution, and data feedback to accelerate deployment of next-generation photovoltaic technologies, including perovskite-based tandem cells. Leveraging the strengths of both parties, perovskite tandem cells are expected to achieve large-scaled mass production within approximately the next three years, with the Group entitled to share in the commercialization returns.

3)     AI4S in Agriculture. In collaboration with a partner, the Group have validated the successul transformation of thousands of mu (hundreds of hectares) of land, setting a new global benchmark for desert restoration.

Commercialization Progress: Accelerating the Momentum of the AI + Robotics + Multi-Agent Platform, Powering a Flywheel of Scalable Growth

1.  Drug Discovery Solutions: AI + Robotics +Multi-Agent Empowering New Drug Development, Platform Value Continues to Grow

During the Reporting Period, revenue from the Drug Discovery Solutions surged by 418.9%, rising from RMB 103.7 million for the year ended December 31, 2024, to RMB 537.9 million for the year ended December 31, 2025. This growth was driven by multiple factors, including: 1) the rapid expansion of the Group's antibody business; 2) progression of service-enabled and incubated enterprise innovation pipelines into the validation stage, with multiple projects achieving key delivery milestones.

1)     Key Commercial Milestones: From Pipeline Services to Platform Partnerships, Opening a New Chapter in Global Value Co-Creation 

The Group's "AI + Robotics + Multi-Agent" platform has been validated at industrial scale and continues to evolve. Building on this, the Group has secured multiple high-value partnerships across small molecules, biologics, and novel drugs, with a client base spanning the world's leading pharmaceutical companies and pioneering biotechs. The Group's business model has also successfully evolved from standalone project services to a strategic model of "platform licensing + co-development + milestone revenues", laying a solid foundation for sustainable growth.

2)     AI Platform Enabling Pipeline Progress: Broad Disease Coverage, Key Technical Milestones Validated

During the Reporting Period, multiple innovative drug pipelines of the Group's empowered partners and incubated companies achieved important progress across a range of disease areas, while the Group remains positioned to capture their long-term commercial value and development upside.

2. AI4S Intelligent Solutions[3]: Scaling from Solutions to Platform- AI + Robotics + Multi- Agent Driving R&D Efficiency Across Domains

During the Reporting Period, revenue from AI4S Intelligent Solutions grew at a rapid pace, rising 62.6% from RMB 162.8 million for the year ended December 31, 2024 to RMB 264.7 million for the year ended December 31, 2025, driven primarily by new strategic initiatives within the business.

The AI4S intelligent solutions business has matured into a fully integrated intelligent R&D system serving biopharmaceuticals and new materials and beyond. By integrating robotics hardware with intelligent algorithms, it enhances R&D efficiency and establishes a next-generation AI4S smart platform centered on AI models, robotic laboratories, and smart agents. The synergy between the two business lines forms a closed-loop ecosystem of "solution deployment, core capability iteration, and platform service delivery." The Group have officially renamed the Intelligent Robotics Solutions business as "AI4S Intelligent Solutions."

1)     AI4S Intellignet Robotic Labs: Winning Landmark Clients Across Industries

XtalPi R&D Solutions has been renamed AI4S Intelligent Robotic Labs, which delivers locally deployable intelligent R&D systems to customers. Built around robotic laboratories and AI-driven platforms, these solutions replace manual experimentation to enhance R&D efficiency, improve experimental success rates, and enable the accumulation of high-quality, high-value proprietary data assets. During the Reporting Period, the AI4S Intelligent Robotic Labs business delivered strong performance, mainly driven by the continued progression of the Group's globalization strategy, with breakthrough commercial progress across both biopharmaceuticals and new materials:

2)     AI4S Intelligent Services: End-to-Engd Smart Platform Accelerates Commercialization

The Group's AI4S Intelligent Services business is built around a core framework of AI models, robotic laboratories, and multi agent, delivering a full-process intelligent R&D service capabilities that spans key stages including raw material supply forecsting, synthesizability prediction to high-throughput screening, separation, and separation and purification. Leveraging the VAST Agent virtual compound library and synthesis robots, the platform also enables rapid delivery and value-added services. In 2025, the Group successfully expanded its customer base to multiple large pharmaceutical enterprises. Its core techology models achieved reaction success rates exceeding 85% in real-world projects, while chemical services recorded an overall service repeat purchase rate of over 75%. Technologies such as spectral analysis and high-throughput condition recommendation have been deployed across multiple client projects and received strong recognition. Looking ahead, AI4S Intelligent Services is well positioned to espport R&D scenarios across pharmaceuticals, chemistry, chemical engineering, mew materials, new energy and more, with a total addressable market potentially worth trillions of US dollars.

3． Continued Integration of Cross-Border M&A Assets

In March 2026, the Mayor of the Liverpool City Region Steve Rotherham warmly welcomed senior leaders from XtalPi to the region. Since the acquisition of Liverpool ChiroChem (LCC) in 2025 and the subsequent formation of LCC Technologies Ltd., the Group has continued to deepen local engagement, working closely with the city government and innovation partners to explore advantages in scientific talent, laboratory infrastructure, and development sites. Following the acquisition, the Group has integrated LCC's PACE™ automated chiral chiral-chemistry platform with its AI-driven and robotics-empowered R&D system, continuously advancing technological upgrades, expanding automated synthesis capacity, and supporting data-driven drug discovery.

Prospects and Outlook

The Group's "AI + robotics + multi agent" R&D flywheel has achieved both technical validation and commercial deployment in biopharmaceuticals and new materials, and delivering its first full-year profit, marking a new milestone in its platform's value. Moving ahead, the Group is well positioned to strengthenthe flywheel effect through continuous iteration, consolidating its AI4S leadership via an intelligent closed-loop system built around over 200 end-to-end AI models, high-throughput robotic laboratories, and scalable AI agents, while expanding into consumer health, photovoltaic materials, advanced chemicals sectors and beyond.

About XtalPi

XtalPi Holdings Limited (XtalPi, 2228.HK) was founded in 2015 by three physicists from the Massachusetts Institute of Technology (MIT). It is an innovative R&D platform powered by quantum physics, artificial intelligence, and robotics. By integrating first-principles calculations, AI algorithms, high-performance cloud computing, and standardized automation systems, XtalPi provides digital and intelligent R&D solutions for companies in the pharmaceutical, materials science, agricultural technology, energy, new chemicals, and cosmetics industries.

SOURCE XtalPi Inc.