GenstorAIge Chairman Sun Dan at World Robot Conference Roundtable: When Domestic Chips Meet Embodied Intelligence, What Will Define the Next Phase of the Robotics Industry?
Photo: Ms. Sun Dan, Chairwoman and CEO of GenstorAIgedelivers a keynote at the 2026 WRC Summit Forum
August 20, the second day of the 2026 World Robot Conference. At the summit forum on "The Evolution of AI Physical Agents: From Digital Brains to High-Risk, Complex Field Operations," GenstorAIge Chairwoman and CEO Sun Dan shared the stage with industry authorities including Professor Liu Yushu of Tsinghua University, Unitree Robotics Co-founder Chen Li, Chinese Society for Electrical Engineering Fellow Pu Tianjiao, and Guangdong Robot Association Executive President Ren Yutong. They engaged in an in-depth discussion on the core proposition of embodied intelligence moving from the lab to real industrial sites.
"The bottleneck in scaling robots from a few pilot units to thousands is not the robots themselves, but a data foundation that connects cloud, edge, and device." This statement by Sun Dan during the roundtable forum became the core footnote of the entire conversation.
From "One Robot" to "Thousands of Mobile Data Centers"
Photo: Ms. Sun Dan, Chairwoman and CEO of GenstorAIge delivers a keynote at the 2026 WRC Summit Forum
Forum moderator Dr. Huang Shuo, R&D Director of Zhongke Kaichuang, posed a question to Ms. Sun Dan that struck at the heart of the industry: Is today's computing power, storage, and data link infrastructure ready to support the scaling of robots from "a few pilot units" to "thousands of units"?
Ms. Sun Dan's answer was very concise: "Not yet."
She pointed out: "A single robot is a walking data factory. The daily volume of 3D point clouds, multi-channel video, and operational data generated by a single embodied intelligence robot far exceeds that of traditional information systems. When scaled to thousands of units, it's equivalent to thousands of mobile small data centers operating simultaneously—and on-site networks are often unreliable."
Ms. Sun Dan emphasized that safety responses cannot afford a round trip to the cloud; inference must be a closed loop locally. The data generated by embodied intelligence robots is continuous, multimodal, and unstructured, with token consumption growing exponentially. Under the strict power consumption constraints of edge-side computing, the only path to large-scale deployment of thousands of embodied intelligence robots is compute-storage convergence, using storage to amplify computing power.
Photo: Ms. Sun Dan, Chairwoman and CEO of GenstorAIge delivers a keynote at the 2026 WRC Summit Forum
Subsequently, she further introduced AI90—integrating high-performance SSDs into the GPU memory architecture to build a three-tier HBM + DRAM + SSD storage system. This allows large model inference to be affordable and runnable on the edge, enabling local closed-loop operation even when disconnected from the network.
When asked about GenstorAIge's position in the embodied intelligence landscape, Ms. Sun Dan's answer was concise and powerful: "GenstorAIge's mission is to pave the way for data to reach intelligence. Our three business segments—storage, connectivity, and solutions—ensure that in embodied scenarios, data can be 'stored reliably, transmitted efficiently, computed affordably, and retained securely.'"
In one sentence, she articulated the irreplaceable value of the data foundation in the era of embodied intelligence.
In Closing
When domestic "chips" meet embodied intelligence, it is a two-way journey between the data foundation and embodied intelligence. Ms. Sun Dan concluded the forum by saying: "The second half of the robotics industry competition is not about whose robot looks more human, but about whose data foundation is more solid."
And what GenstorAIge is doing is paving that invisible track for this second half of the competition.