At the 2026 Lean Digital Innovation Conference, Ding Yu, Vice President of Guangdong Yuantu Future Technology Co., Ltd., emphasized that computing power serves as the cornerstone for intelligent upgrades in the manufacturing sector. He stressed that companies adopting large language models must prioritize both computing power investment and the exploitation of data assets.
The conference opened on the 6th at the National Exhibition and Convention Center (Tianjin), with Yuantu participating as an ecosystem partner, contributing to keynote speeches and the exposition segment. Ding noted that the company's presence aimed to learn from manufacturing enterprises' practical experiences while showcasing the role of computing infrastructure in digital transformation.
Ding pointed out that while manufacturing has long utilized automation and early-stage intelligent technologies—particularly machine vision-based AI deployed for years—traditional AI is designed for specific scenarios. In contrast, large language models offer stronger adaptability and new possibilities for the industry. However, he cautioned that manufacturers hold vast amounts of proprietary internal data that constitute core assets and cannot be easily shared externally.
To effectively harness large language models, Ding explained that companies must first transform raw data into usable data assets and deploy internal computing power to run privatized models. This process involves multiple stages, including computing infrastructure, data services, enterprise AI platforms, and scenario-specific applications, with computing power underpinning every step.
Addressing common obstacles in manufacturing transformation—such as difficulties in computing deployment, high costs, and bottlenecks in traditional cluster architectures like memory walls, power walls, and communication walls—Ding proposed a two-pronged approach. For professional computing service providers, super-node technology can enhance overall efficiency and lower unit costs in the near term. For manufacturers lacking specialized technical teams, collaboration with partners is essential to achieve end-to-end optimization—from underlying infrastructure upgrades to integrated solutions—preventing scenarios where investments in platforms yield disappointing performance.
Ding introduced Yuantu's full-stack capabilities under four pillars: storage, computing, networking, and management. This framework covers data storage, high-performance computing, network interconnection, and unified management platforms, while also helping enterprises establish AI application platforms that enable mainstream models to run smoothly on their own hardware.
Positioning Yuantu itself as a typical industrial manufacturer, Ding revealed that the company already leverages AI-assisted tools in development work to boost efficiency and incorporates AI models for decision support in complex scenarios. The company is also actively exploring embodied intelligence applications on production lines to tackle automation challenges arising from multi-variety, flexible manufacturing.
Looking ahead, Ding predicts that over the next two to three years, as model intelligence and computing power continue to advance, enterprises will run increasingly sophisticated models internally to support production scheduling, operational planning, and even executive-level strategic decisions. Embodied intelligence is expected to generate tangible value in production lines, logistics, and service sectors. Ding stated that Yuantu plans to build on its computing platform deliveries, integrating its own manufacturing experience to provide customers with application support layers, while co-creating an ecosystem with robotics, model, and application partners to deliver embodied intelligence solutions tailored to specific use cases.