GIL 2026 | Frost & Sullivan 2026 New Investment Conference “Advancements and Leaps in AI” Panel Successfully Concluded

Top Ten Technical Trends for the Future of Global AI Development

Ms. Li Qing, Executive Director of Frost & Sullivan China, systematically discussed the top ten core technical trends in global AI development. She noted that the competition in the AI industry is shifting from pure model capabilities to practical applications: companies no longer focus solely on whether models can complete tasks, but also on unit task cost, task completion quality, operational reliability, and stable delivery capabilities in complex business processes.

Focusing on the next phase of technological evolution, Ms. Li Qing analyzed key areas such as AI-native data centers, inference-first hybrid computing architecture, inference computing expansion, domain-specific models and efficient small models, native full-modal real-time interaction, world models and VLA, Agent continuous memory and experience learning, autonomous long-term Agents, AI-native software development paradigms, and trusted AI technology stacks. She stated that as AI moves from content generation to tool invocation and real business execution, industry judgment criteria will shift from "whether it can be achieved" to "whether it can create productivity at a reasonable cost," which will also become an important dividing line in the value differentiation of AI in the future stage.

Mr. Li Yiyong, Founder and CEO of Benyuan Intelligence

Core Infrastructure for AI Industry Development: The Transition and Reconstruction of AI Data Services

Mr. Li Yiyong, Founder and CEO of Benyuan Intelligence, shared the evolution of the AI industry from the perspective of data infrastructure. He pointed out that algorithms, computing power, and data together constitute the core elements of AI development. As large models, world models, embodied models, and agents continue to evolve, the breadth, depth, and accuracy of data supply become important variables affecting model evolution speed. Data services have also expanded from traditional collection and annotation to evaluation, synthesis, management, and full lifecycle model services.

For the next phase, Mr. Li Yiyong suggested that data services need to further cover digital tasks and physical world interaction tasks, continuously accumulate high-quality data such as expert trajectories and embodied operations. As models continuously learn human language, vision, operations, and professional experience, the value of data infrastructure will upgrade from mere data processing to an important industrial foundation supporting continuous AI capability iteration.

Mr. Sun Baigui, Head of Artificial Intelligence at TreeRoot Technology

From Connection to Collaboration: The Breakthrough and Reconstruction of Industrial Intelligent Agents

Mr. Sun Baigui, Head of Artificial Intelligence at TreeRoot Technology, shared insights on the implementation path of industrial intelligent agents. He said that the value of industrial AI does not lie in single-point answering or process automation, but in truly entering the production system through intelligent agents to connect devices, data, and business processes, achieving continuous upgrades from device connection and knowledge understanding to production collaboration and global decision-making.

Combining industrial scenarios, Mr. Sun Baigui pointed out that the implementation of industrial intelligent agents requires considering underlying data, knowledge models, toolchains, business processes, and security systems, and achieving the transition from "understanding" to "execution" through full-stack capabilities. Industrial scenarios have complex processes, high real-time requirements, and high reliability standards, so AI must deeply integrate with existing industrial systems to form a sustainable operation and optimization engineering system, thereby truly reconstructing the manufacturing production paradigm.

Unleashing Data Value, Tencent Cloud Supports the Industrialization of Embodied Intelligence

Mr. Chen Zheng, General Manager of Tencent Cloud Storage Products, analyzed the key bottlenecks in the industrialization of embodied intelligence from a data engineering perspective. He believed that embodied intelligence is moving from laboratory demonstrations to large-scale application before the night, and after algorithms and computing power capabilities converge, high-quality, scalable data supply becomes a crucial factor determining the upper limit of model capabilities. Real machine data, non-embodied collection data, simulation synthetic data, and internet video data each have value, and the core is not to replace each other, but how to form a reasonable data formula and a continuously evolving data flywheel.

Focusing on data collection, processing, management, delivery, and training, Mr. Chen Zheng introduced the data infrastructure approach throughout the lifecycle of embodied intelligence. He said that only by enabling massive heterogeneous data to be "stored effectively, found easily, used well, delivered quickly, and safe to deliver" and continuously returning issues from real machine verification to the next round of data collection and model training can the model iteration efficiency be improved, and embodied intelligence move from technical validation to large-scale industrial application.

From "Special Machines and Special Brains" to "Brain Sharing": Exploring the Distribution of Embodied Intelligence Capabilities

RoboScience机器科学创始人兼CEO田野先生围绕具身智能能力的规模化分发展开分享。他认为，具身智能当前的重要瓶颈之一在于"专机专脑"：一个模型往往只能适配特定本体、任务和场景，复制成本较高，而机器人硬件形态又长期存在差异化。要推动具身智能真正进入各行各业，需要让智能能力能够像云计算和Model as a Service一样被高效分发。

田野先生提出，具身智能基础模型需要具备跨本体、跨任务、跨物体的泛化能力，并通过仿真、视频与真机数据形成持续进化的数据闭环。基于云端部署和API化调用，场景方、机器人厂商与集成商可以按需获得具身操作能力，减少重复建设数据和模型体系的成本。具身智能由此有望从单机能力向可共享、可扩展的基础服务演进。

Frost & Sullivan’s “AI Third Pole” reshapes the industrial economy through decision intelligence

Mr. Wang Xi, Co-Founder of Sanguan Technology, shared his insights on the relationship between decision-making AI and industrial decision-making. He pointed out that artificial intelligence should not be understood merely as large language models; capabilities such as prediction, rules, simulation, reinforcement learning, mathematical programming, and optimization solutions are also essential components of AI. For real industrial challenges, the key is to focus on business objectives and problem-solving, combine various technical approaches, and transform information, predictions, and knowledge into actionable plans.

In scenarios with strong constraints and low tolerance for errors, such as energy and power, industrial manufacturing, and Logistics & Supply Chain, decision-making problems often involve large-scale variables, complex constraints, and real-time changes. Mr. Wang Xi stated that solvers and optimization capabilities play a crucial role in such situations. The core value of decision-making AI lies in creating a complete loop from data interpretation, solution generation, optimization, and execution feedback, enabling AI to move from simply providing suggestions to truly supporting decision-making and actions.

From Model to System: The Era of AI AgentBig Data & AIReconstruct

Mr. Zhu Yibo, Co-Founder and Chief Technology Officer of Step Star, shared the reconfiguration trends of AI Infra in the Agent era. He stated that as AI moves from simulating human expressions to exploring, executing, and performing more complex intelligent tasks, the boundaries between infrastructure and models are continuously merging. In the future, high-quality AI capabilities depend not only on the models themselves but also on whether infrastructure such as training, inference, software-hardware integration, scheduling, and systems engineering can continuously enhance intelligent output per unit computing power.

朱亦博先生指出，多模态、Agent与端云协同等趋势，将进一步推动AI Infra由支撑模型训练的后台系统，升级为直接决定模型体验、推理成本和交付效率的核心能力。模型与基础设施的联合设计，将成为AI企业提升系统效率、释放模型能力并形成规模化应用的重要方向。

Algorithm-based E-commerce: AI and algorithms are reshaping brand development in the next decade

Mr. Li Wenliang, Co-Founder and CEO of Zhixinghehe, analyzed the brand growth logic in the algorithm era from the perspectives of consumer and marketing. He stated that as search, recommendation, large models, and intelligent agents increasingly influence consumers' information acquisition and purchasing decisions, the key targets for brand building are no longer just consumers, but also the understanding that algorithm systems have of businesses, products, and brands.

Mr. Li Wenliang believes that algorithm-based e-commerce represents not only changes in e-commerce channels but may also lead to a reconfiguration of marketing methods, organizational structures, and growth strategies. Companies need to reassess the relationship between content, data, brand signals, and algorithms. Especially for To B businesses that have neglected marketing efforts over time, it is essential to understand how brand assets are recognized and utilized by models in the AI era, so as to achieve sustained growth under new information distribution rules.