MEGAROBO raises nearly $200 million for AI infrastructure in the physical world

MEGAROBO completed a financing round of nearly $200 million, with participation from Oman Investment Authority and follow-on investment from existing shareholders. Funds will support core R&D, global operations and AI infrastructure for the physical world.
Cumulative financing of nearly $600 million.
36Kr has learned that MEGAROBO recently completed a new financing round of nearly $200 million. Participants included the Oman Investment Authority (OIA), the Sultanate of Oman’s sovereign wealth fund; existing shareholders Asia Investment Capital, Granite Asia and Sinovation Ventures added investment. CMB International, Suzhou Capital Group, Sanjiang China Merchants Group and others also participated, spanning international sovereign capital, major financial investors and local state-owned industrial capital.
For AI for Science to work in practice, one foundational problem remains unavoidable: data. Scientific data that is scalable, high-quality and reusable is still scarce, and most domestic life-science laboratories’ equipment and consumables do not yet support producing data consistently at scale.
MEGAROBO founder and CEO Yuqing Huang has described this as “the conflict between scientists’ growing demand for data and relatively underdeveloped production capacity.” That tension has brought another wave of interest in AI-automated laboratories amid this year’s AI4S investment boom.
As AI moves from the digital into the physical world, data alone is insufficient. AI must enter complex physical systems such as research, laboratories and production lines, connecting models, data, software, hardware and physical execution across thinking, design, execution, observation and continuing feedback-based learning and optimization.

Image credit: MEGAROBO
MEGAROBO was among the earlier Chinese companies to pursue this direction. Over its ten years, it has raised nearly $600 million in total. In June 2025, it filed an application with the Hong Kong Stock Exchange. According to its prospectus, its order backlog stood at RMB 1.5 billion as of June 2025.
Early on, the team wanted to build a general-purpose robot deployable in more than a million life-science laboratories. It soon found that scientists needed an entire solution from experimental design through execution and data processing, rather than a single device.
In 2019, the team therefore shifted toward building a super laboratory. Its core is the MEGAROBO Kunpeng Lab, which the company defines as a “data factory for the life-science industry.” Version 1.0 began operating in 2021 with round-the-clock unattended experiments; it has now reached version 4.0.
According to the company’s public information, the updated Kunpeng Lab’s biggest change is a shift from “automated execution” to “autonomous agent-driven” operation, positioning it as a new AI-native scientific-discovery and validation platform for AI for Science. It has two “brains”: the dry-lab brain automatically reviews literature and designs experimental routes after receiving a goal; the wet-lab brain translates and dispatches instructions. A scheduling agent coordinates robotic-arm transfers between instruments, reorganizing R&D into a loop of “AI design—autonomous experiments—real validation—data production—model learning.”
For example, recent operating data indicates that the platform advanced 31 drug-development pipelines at scale over seven months.
The business now has two main areas: scientific intelligence and industrial intelligence. In scientific intelligence, the company has delivered infrastructure to more than 500 customers, including multinational pharmaceutical firms, CROs, biotech companies and research institutions. It supports early research, target exploration, drug screening, candidate optimization and production quality control. Kunpeng Lab provides research services to over 200 customers.
One frequently cited case concerns an antibody-drug team using Kunpeng Lab to complete 56 dry/wet experimental iterations across 18 targets in four months—described as equivalent to three or four years’ work by dozens of researchers under a traditional approach.
MEGAROBO recently joined Tencent Cloud and Twist Bioscience to launch the global Poreva Open AI protein-design challenge, without restricting model or technical approach. Kunpeng Lab will perform wet-lab validation, aiming to establish an “open evaluation yardstick” for AI4S at the scale of more than 10,000 protein-design sequences.
Industrial intelligence focuses on manufacturing areas including semiconductors, optical interconnects and new displays, providing defect recognition, quality control, yield analysis, anomaly diagnosis and process-parameter optimization. The goal extends beyond “finding defects” toward helping AI understand the complex relationships among materials, processes, equipment and yield.
The company’s business model is also evolving into two delivery approaches. One centers on intelligent infrastructure, offering products and platforms integrating AI, software and hardware. The other is ARO, or AI-Native Research Organization, combining AI capabilities with experimental execution to provide research services. The first supplies infrastructure and technical capabilities; the second converts those capabilities into actual research outcomes, continually producing high-quality real-world feedback to accelerate discovery.
MEGAROBO says it will use this round to increase core-technology R&D and global business investment, strengthening partnerships in China, Europe, North America, the Middle East and Southeast Asia. “Previously, AI’s most important capability was learning from knowledge people had already accumulated. Next, it will increasingly enter the real world and, through experiments, manufacturing and continuing feedback, help people create new knowledge, materials, processes and products. We want to build the infrastructure that lets this happen at scale,” Huang said.
This article originally appeared on the WeChat account “36Kr,” written by 36Kr.
Related topics
After reading
Join the AINRK community
Read how to join the community and connect with people working in AI.




