Mgi Tech Co.,Ltd. (SSE: 688114) held its interim results conference on August 24, providing in-depth insights into AI applications across its three core business lines and the deployment of spatial omics in cutting-edge fields. Management stated that, driven by AI, the company has established a three-tier architecture comprising "data entry, execution foundation, and intelligent closed loop."
The subsidiary Yongsheng Intelligent's ProtoPilot self-evolving multi-agent system has already been serving AI4S life science scenarios, including protein and enzyme design, blood pre-processing, and PBMC separation. Meanwhile, the synergistic value from two acquisitions completed earlier this year is accelerating its realization, with spatial omics continuing to gain traction in frontier areas such as life foundation models, brain science, pathology research, and AI-driven drug discovery. Nanopore sequencing, leveraging its direct RNA sequencing advantage, is accelerating empowerment of related R&D and business expansion.
According to disclosures in MGI Tech's investor relations activity records, management signaled during the meeting that AI has upgraded from a supporting tool to a core driver across all three business lines, highlighting emerging structural upgrade opportunities in the industry. The following are key points disclosed in the minutes:
Q: Over the past few years, the company has been consistently investing in AI applications. In which business segments has AI currently been deployed, and what are the specific empowerment directions or scenarios?
A: In the second half of 2025, the company upgraded its original three business lines into full-length sequencing (SEQALL), intelligent automation (GLI), and multi-omics (OMICS), with AI having transitioned from a supporting tool to the core driver across all three lines. In full-length sequencing, AI has been comprehensively applied to core processes such as sequencing enzyme screening and performance enhancement. The G10-FR integrates generative AI with self-luminescent biochemistry technology, achieving the industry's leading flash-speed sequencing in as fast as 2 hours. The T1+ sequencer, released in 2025, features an integrated bioinformatics computing module supporting automated advanced analysis. The CycloneSEQ-S1 sequencing chip, launched in the first half of 2026, applies the company's latest AI algorithms, delivering a dual leap in sequencing accuracy and chip throughput while supporting long-read and direct methylation sequencing. Additionally, the company, in collaboration with Google, released an AI variant detection model trained on DNBSEQ data, which can reduce false positive rates by up to 73%, achieving world-leading accuracy.
In intelligent automation, the company has built a comprehensive "AI brain plus intelligent limbs" system. The released PrimeGen dry-wet collaborative multi-agent system, with large language models as its core, transforms the entire process from primer design to experimental execution into standardized automated protocols. The αLabStudio platform achieves atomic-level precise control and intelligent scheduling of heterogeneous devices through digital twins. For automated equipment such as the PrepALL liquid handling platform and Smart8 pipetting robots, the company has deeply integrated precision control with AI vision algorithms, helping users enhance quality and efficiency in experimental and library construction processes. The GLI business segment has delivered landmark projects including the Hong Kong Baptist University Smart Laboratory and the Beijing Brain Science and Brain-Inspired Research Institute's fully automated unmanned dark laboratory, with the high-throughput enzyme screening platform achieving screening success rates three times higher than traditional methods.
In the multi-omics domain, AI has been widely applied across single-cell, spatial omics, and imaging diagnostics. The DNBelab C is the world's first fully automated unattended single-cell library preparation platform, driving single-cell sequencing toward scale and standardization. The FluoXpert Vision image analysis platform integrates large models, supporting natural language instruction processing of pathological statistical data to accelerate pathology AI implementation. The handheld ultrasound diagnostic device incorporates an AI-assisted diagnostic system capable of automatic lesion identification.
In the first half of 2026, the company's multi-omics business segment revenue increased 44.91% year-on-year. Beyond growing research demand, another core driver is the increasing reliance of AI large models on multi-omics data interpretation. The rigid demand for high-quality data in model training and iteration is continuously transmitting upstream, driving the scaled deployment of data collection platforms such as single-cell sequencing and spatial omics.
To further accelerate AI deployment, the company established the wholly-owned subsidiary Yongsheng Intelligent in the first half of 2026. In collaboration with the Shanghai Artificial Intelligence Laboratory, it released the ProtoPilot self-evolving multi-agent system, which for the first time achieves a full-chain dry-wet closed loop from natural language experimental requirements to physical execution in wet experiments, truly realizing the concept of "PhysicalAI for life science laboratories." This covers capabilities including solution understanding and design, Protocol/SOP generation, intelligent scheduling and multi-device coordination, automated code generation, visual recognition and quality control, experimental data feedback, and strategy iteration, advancing experimental workflows from single-point automation toward a closed loop integrating solution design, device execution, and result feedback.
This capability serves not only sequencing scenarios but also extends broadly across AI4S life science applications, currently empowering protein and enzyme design, directed evolution, blood pre-processing and PBMC separation, plant gene editing, and mass spectrometry method development. It is also expanding into emerging fields such as sports medicine, precision health, and marine bio-manufacturing, driving life science laboratories from experience-driven approaches toward data-driven, intelligent decision-making, and continuous optimization.
Overall, the company has constructed a three-tier architecture of "data entry, execution foundation, and intelligent closed loop": full-length sequencing and multi-omics serve as the data entry, intelligent automation forms the execution foundation, and Yongsheng Intelligent is the ultimate anchor of the intelligent closed loop. The three tiers work synergistically to generate more data that feeds back into experimental intelligence, which in turn drives automation business growth, forming a positive flywheel. Going forward, Yongsheng Intelligent will also become a key direction in the company's AI strategic layout.
Q: Following the acquisitions of Shenzhen Sanjian Qifa and Hangzhou Xufeng, how is the business integration progressing? Have the synergies contributed tangible results to the financial performance?
A: Shenzhen Sanjian Qifa possesses leading technical capabilities in spatial omics, with its Stereo-seq technology ranking internationally in both resolution and field of view. Its spatial reagent kits are deeply integrated with the company's sequencers, with a revenue driving ratio of approximately 3:1 between reagent kit revenue and sequencing product revenue, demonstrating significant synergy. Since 2026, application demand for spatial omics across multiple frontier fields has continued to emerge, primarily reflected in three aspects. First, in life foundation models, multiple institutions have utilized spatial omics' full-omics capabilities for life foundation model training, including the co-construction of a digital biology spatiotemporal computing innovation research platform with South China University of Technology. Second, in brain science, benchmark cooperation cases have been established with institutions such as the Shanghai Center for Brain Science and Brain-Inspired Intelligence. Third, in pathology research and AI drug discovery, successful cases have been implemented with institutions including the Sixth Affiliated Hospital of Sun Yat-sen University and Central South University's Furong Laboratory.
In the second half of 2026, Shenzhen Sanjian Qifa's R&D team will further intensify new product development. Technologies such as spatial immune repertoire sequencing and chip-based single-cell technology (Stereo-cell) will be launched successively, continuously supporting the company's capability expansion in the multi-omics field. Shenzhen Sanjian Qifa is committed to its performance commitments this year, and the company will continue to advance related business implementation.
Hangzhou Xufeng's nanopore sequencing instruments (CycloneSEQ series) feature long-read, single-molecule sequencing, real-time data acquisition, and native DNA/RNA information detection capabilities. In scientific research, they can be applied to complex genome de novo assembly, structural variant analysis, haplotype phasing, full-length transcriptome, and epigenetics research. In medical and clinical translation, they can be applied to rapid pathogen detection, complex genetic variant analysis, family and embryo genetic analysis, and on-site mobile monitoring scenarios. Its long-read capability can span repetitive sequences, complex structural regions, and multiple genetic variant sites, complementing short-read sequencing platforms in throughput, accuracy, and complex region resolution capabilities.
Since being integrated into MGI Tech, the "short-read plus long-read" synergy has formed successful cases across multiple fields. In scientific research, collaborations have been established with Peking University, Fudan University, and other domestic institutions in population cohorts, animal and plant research, and agricultural studies, covering genomics, transcriptomics, and epigenomics. In clinical applications, CycloneSEQ demonstrates significant advantages in pathogen surveillance, with a cooperation with Shengting Medical landed in the first half of 2026, marking the first domestic OEM partner for CycloneSEQ nanopore sequencing. Additionally, nanopore sequencing possesses natural advantages in direct RNA sequencing, effectively applicable to mRNA sequence quality inspection for vaccine manufacturers. Successful cases already exist in this area, and Hangzhou Xufeng is continuously advancing related R&D and business expansion. In the service provider sector, partnerships have been established with domestic collaborators including Berry Genomics and Aijibeike.
Comments