At the 2026 International Consumer Electronics Show (CES) during the Lenovo Tech World event, Yang Yuanqing, Chairman and CEO of Lenovo Group, forecasted in his keynote address that the next surge in computing power will stem from the rapid expansion of AI inference capabilities.

Yang posits that the global computing power market has undergone three distinct waves: enterprise informatization, the ascent of cloud computing, and the era of large-scale model training. "Now, we stand on the brink of the fourth wave—AI inference," he remarked.

During the event, Lenovo and NVIDIA unveiled the Lenovo AI Cloud Superfactory Initiative. Jensen Huang, CEO of NVIDIA, emphasized that this initiative aims to streamline the AI deployment process, fostering true industrialization of AI infrastructure.

From the advent of enterprise electronic office systems to today's AI-driven intelligent revolution, each leap in computing power infrastructure has fundamentally reshaped business and societal landscapes. Yang Yuanqing succinctly chronicled this evolution in four distinct phases.

The inaugural wave commenced with enterprise informatization and digital transformation. Traditional computing technology formed the backbone of this era, with enterprises deploying servers and databases to digitize paper-based processes, enabling rudimentary data management and process automation.

The second wave was propelled by cloud services and applications. With the emergence of cloud platforms like Amazon AWS and Microsoft Azure, enterprises could access computing resources flexibly, bypassing the need for proprietary data centers. This era witnessed the meteoric rise of cloud computing, fueling the popularity of mobile internet and SaaS models.

The third wave was ignited by the training of large language models. To train models boasting hundreds of billions or even trillions of parameters, such as GPT and Gemini, tech behemoths constructed vast computing clusters. This phase concentrated AI training in the cloud, driving a colossal demand for high-performance AI chips like NVIDIA H100 and AMD MI300X.

Now, the fourth wave—the explosion of AI inference—has dawned. Intelligence is cascading back from the cloud to local and edge devices, democratizing AI capabilities across PCs, mobile phones, servers, and even IoT devices.

Lenovo's proposed Hybrid AI framework seeks to amalgamate personal, enterprise, and public intelligence, crafting an AI deployment architecture tailored to diverse needs through core technologies like intelligent model orchestration.

Why has AI inference assumed such paramount importance? Inference entails AI models making real-time judgments and generating outputs from input data. As the application scenarios of generative AI expand, from real-time voice interactions to complex decision-making analyses, the demand for inference has skyrocketed.

Deploying inference capabilities near the data source enables millisecond-level response times and heightened data security. Yang Yuanqing underscored that this is becoming a veritable competitive edge for enterprises.

Lisa Su, Chairman and CEO of AMD, echoed a similar sentiment at the same event: global enterprises are contemplating how to bring AI closer to their data.

Market analysis corroborates this trend. According to a CSC Securities research report, by 2026, AI penetration rates in mobile phones and PCs are projected to reach 45% and 62%, respectively. The edge AI market is anticipated to surge from 321.9 billion yuan in 2025 to 1.22 trillion yuan in 2029.

Beneath this market explosion lies a paradigm shift in the operational logic of AI models. The traditional centralized cloud processing model grapples with challenges such as network latency, bandwidth costs, and privacy risks when handling massive, high-frequency inference requests.

However, the expansion of inference computing power is not without its apprehensions. Independent research teams caution that the computing power demand for inference models is escalating at an alarming rate. The inference computing power of OpenAI's o3 model already dwarfs that of o1 by a factor of ten.

Confronting the wave of AI inference, Lenovo has embraced an ecological strategy deeply entwined with chip giants, actively constructing a diversified AI hardware ecosystem.

Intel and Lenovo jointly announced the Aura Edition AI PC and a FIFA co-branded gaming computer. In the wearable device arena, Lenovo also collaborated with Qualcomm to co-develop AI-native wearables. Yang Yuanqing predicted that the market size in this domain could surpass one billion units.

Concurrently, Lenovo and AMD unveiled their collaboration on enterprise AI deployment. They will introduce the AI inference server ThinkSystem SR675i, powered by AMD EPYC processors. Specifically designed for deploying AI models locally and at the edge, this server aids enterprises in enhancing inference efficiency and curtailing operational costs.

References:

https://stock.10jqka.com.cn/usstock/20260107/c673812324.shtml

https://finance.sina.com.cn/stock/t/2026-01-07/doc-inhfmycs5208883.shtml