In these speeches, the key themes of physical AI, hybrid AI, and spatial intelligence are not isolated but rather form a co-evolving ecosystem.
Article author and source: Ba Jiuling, Wu Xiaobo Channel
Las Vegas, the plane has landed. In the first week of the new year, many people are heading to CES, the annual tech extravaganza is about to begin.
The focus these past few days has been on the people and what they said.
January 4th and 5th, the two days before the official start of the exhibition, are CES media days, when many companies choose to launch new products. As in previous years, these two days have almost turned into a "summit of leaders" for the world's leading AI companies: Nvidia founder Jensen Huang, AMD chairman Lisa Su, Intel CEO Liwu Chen, Lenovo CEO Yuanqing Yang, Qualcomm CEO Cristiano Amon, and others have all appeared and delivered speeches.
But these speeches are not just about new product launches; they are more like statements about the company's technological path and industrial layout. Every judgment made will be repeatedly analyzed and pondered by the world.
In response, Zhuang Minghao, a senior tech media personality, said that CES is supposed to be a consumer electronics show, but watching the presentations by Nvidia and AMD, especially by Jensen Huang, it was clear that there was almost no content related to the "consumer market." Instead, the presentations were all about the GPUs needed for data centers in the context of current AI, and grand topics such as the so-called "physical" limits. It seems that our consumer products are no longer that important.
Over the next four days, as the main stage lights dimmed and the exhibition hall doors opened one by one, the official CES exhibition period had just begun. Reports indicate that the over 2.5 million square feet of exhibition space attracted more than 4,000 exhibitors and hundreds of thousands of attendees from around the world. The hottest tech products, including robots, AI PCs, and smart glasses, will all be showcased here.
CES 2026 Opening
Chinese companies are also prominently represented. These include established manufacturers like Lenovo, Hisense, and TCL, as well as robotics companies such as Unitree Robotics, Logic Robotics, and CloudMinds. Chinese manufacturers are also present in niche markets such as floor scrubbers, lawnmowers, stair climbers, and pool cleaners.
As Jensen Huang repeatedly emphasized in his speech, "The AI race has begun, and everyone is striving to reach the next level."
So, how did the leaders of major AI companies interpret industry trends and outline their respective companies' future blueprints during the multiple keynote speeches on the first day? According to Du Yu, Dean of the Unknown Artificial Intelligence Research Institute:
Jensen Huang's core logic is that AI must understand the common sense of the physical world in order to truly interact with the real world. From a business perspective, the ceiling of the real world is much larger than that of the online world.
Huang Renxun wore crocodile skin this time.
Intel emphasizes hybrid AI and edge AI because there are many real-world scenarios that require edge AI, such as healthcare, finance, and industry, emphasizing data privacy, low latency, and zero network outages.
AMD's Lisa Su stated that computing power needs to increase 100 times in the next few years, essentially addressing the "computing power shortage." AMD's strategy is to seize Nvidia's data center market with more cost-effective computing power.
Li Feifei, the "AI godmother" invited by Su Zifeng, once again emphasized the inherent limitation that "large language models are ultimately constrained by language itself." She believes that language is a tool for describing the world, but not the world itself.
From physical AI, hybrid AI, and edge AI to computing power, spatial intelligence, and AI agents… these keynote speeches, featuring both new and old concepts, collectively outlined a panoramic view of the future development of AI. We have compiled fourteen key quotes related to the future development of AI, marking a small timeline for this rapidly evolving AI era, waiting to be validated in the future. We have also invited experts in related fields to share their perspectives on these insights.
Fourteen Golden Quotes
1. "The ChatGPT moment for physics AI is coming soon."
—Huang Renxun
2. "The breakthrough in physical AI has brought AI from the screen into our physical world—and this is timely, as the world is building all sorts of factories for chips, computers, life-saving drugs, and AI. With the global labor shortage worsening, we need automation driven by physical AI and robotics more than ever before."
—Huang Renxun
3. "Today I want to talk about AIpamayo, which is what we've been doing in the field of self-driving cars—we've not only open-sourced the models, but also the data we use to train those models. Because only in this way can you truly trust the source of the models. We open-source all our models, and we help you create derivatives from them."
—Huang Renxun
4. "If you look at the models of the world, OpenAI generates more tokens than any other model, and the second largest group is probably open source models. My guess is that, over time, open source models may become number one."
—Huang Renxun
5. "We believe that as AI capabilities continue to improve, localized computing will only become more and more important. First, the higher the degree of localization, the lower the latency, and therefore the better the performance. Second, the more localized the AI, the safer it is, and the more truly 'it belongs to you.' Third, every AI inference costs money, and centralized cloud computing has cost and bandwidth bottlenecks, while localized computing can significantly reduce overall costs by reducing reliance on transmission and infrastructure. Fourth, for enterprises, the core value of local computing is not in performance, but in bringing data, intelligence, and control back into their own hands."
— Perplexity CEO Srinivas
6. "The era of hybrid artificial intelligence has begun... Local AI securely performs tasks, keeping the data on the machine, while cloud AI is responsible for global reasoning, planning, and multi-agent orchestration."
— Jim, Senior Vice President of Intel's Client Computing Group
7. "You will see hundreds of edge devices in different form factors in key areas such as smart cities, factories, healthcare and various automation systems, and the demand for them is huge and growing."
— Jim, Senior Vice President of Intel's Client Computing Group
8. "The number of AI users has jumped from the initial 1 million to more than 1 billion active users today... We expect the number of active AI users to grow to more than 5 billion, and AI will truly integrate into all aspects of our lives, just like mobile phones and the Internet today."
— Lisa Su
9. "The computing power we have now is far from enough to support everything that AI can do... In order to make AI ubiquitous, we need to increase global computing power by a hundred times in the next few years, or by more than ten times in the next five years."
— Lisa Su
10. "In the future, a country's GDP growth will largely be determined by its available computing power."
—Gregor, President of OpenAI
11. "What excites me is that a new generation of artificial intelligence technologies has emerged, including embodied AI and generative AI, which finally allow us to give machines capabilities that are closer to human levels—spatial intelligence."
—Li Feifei
12. "We are moving from a 'system that passively understands the world' to a 'system that helps us interact with the world'."
—Li Feifei
13. "Most AI assistants today are reactive agents. You open an app, ask a question, and get a response. But when AI is running quickly on the device and always on, it can proactively perform tasks for you."
—Amit, CEO of Luma AI
14. "2026 will be the year of AI agents. AI will be able to help you complete more tasks, and may even be able to complete the entire task end-to-end, rather than just doing some piecemeal work."
—Amit, CEO of Luma AI
Big Head Has Something to Say
Hu Yanping: Distinguished Professor of Shanghai University, scholar of intelligent technology industry and intelligent economy.
Jensen Huang's statement that "the ChatGPT moment for physical AI is coming" could be more specifically pinpointed to 2026, which might represent the GPT3-GPT3.5 phase of physical AI. This means significant progress, but we shouldn't have overly high expectations, especially regarding the "brain development" of robots. However, the arrival of L3 mass production for autonomous driving this year is a relatively certain and optimistic prospect.
Furthermore, both AMD and Nvidia are emphasizing full-stack AI capabilities, from chips to hosts and clusters. Their products cover a wide range of vertical scenarios, from chip computing power and development environments to vertical applications, rather than focusing on isolated breakthroughs. AMD, in particular, has caught up rapidly, achieving a complete AI layout from data centers to personal devices, while also significantly improving product performance.
Fei-Fei Li showcased Marble, World Labs' first commercial world model, which aims to generate persistent, navigable, and consistent 3D worlds to support human creativity rather than replace it, aligning with her past "human-centered" AI development philosophy.
Li Feifei delivered a speech.
Next, at CES 2026, I will focus on five key areas: first, embodied intelligence such as robots; second, smart devices such as smart glasses; third, the L3 autonomous driving industry chain; fourth, training and inference computing architectures and changes in terminal and edge AI computing power; and fifth, the changes brought to the field of smart health by the injection of model capabilities based on sensing algorithms.
The "AI China Chain" in the sense of the AI industry chain has taken shape, and the "AI China Loop" in the sense of AI technology and application has also been closed. While launching a wide variety of innovative products, Chinese companies are expected to provide more solutions for the world.
Du Yu: Dean of the Unknown Artificial Intelligence Research Institute, PhD in Technology Economics from the Chinese Academy of Social Sciences
After listening to several presentations, my impression is that the "computing power arms race" has reached a fever pitch. All three giants are emphasizing computing power, but their approaches are completely different: Nvidia pursues absolute performance, AMD pursues cost-effectiveness, and Intel pursues edge computing. This reminds me of the recent surge in domestically produced GPUs, the "Four Little Dragons," each with its own strengths. This also means that in the next 2-3 years, the chip price war and performance war will be extremely fierce, which will be a major boon for startups and enterprise customers—computing power costs will drop significantly.
China may not catch up in chip manufacturing in the short term, but it's making great strides in AI application hardware. At CES, Chinese companies accounted for more than half of the robotics exhibitors, including Unitree Robotics, Calcium Robotics, and Dreame Robotic Vacuum Cleaner—all Chinese brands. Robotics is just one typical example. This is because the Chinese market is large, has numerous application scenarios, and iterates rapidly. Therefore, my advice to Chinese startups is: differentiated competition and long-term vision.
Zhang Xiaorong: Dean of the Deep Technology Research Institute
In his speech, Jensen Huang's viewpoint was not merely about technological upgrades, but also about a paradigm shift. Previous AI was "keyboard and mouse interaction," current AI is "visual and language interaction," and Huang defines "physical AI" as "action interaction." Through the Cosmos model (learning physical laws through videos) and the Newton engine (real-time physics computation), he attempts to solve the "illusion" problem of AI—to make machines understand that "water is a fluid and glass is brittle." This is to enable AI to work safely in factories and homes, not just for writing poetry and painting.
Both AMD and Intel have avoided direct competition with Nvidia GPUs, each pursuing alternative paths:
Lisa Su showcased AMD's position in the AI industry chain. Her emphasis on the Helios system and the Ryzen AI 400 series for PCs sends a clear signal: the competition for computing power is a protracted war, and AMD is prepared to fight, focusing more on edge computing and cost-effectiveness. AMD's strategy is more like that of a "pragmatic idealist," emphasizing "breaking through computing power bottlenecks," which essentially means lowering the barrier to entry for AI. If computing power costs cannot be reduced, small and medium-sized enterprises will not be able to afford it, and the AI ecosystem will wither.
Lisa Su introduces the AMD Instinct MI455X GPU
Intel's emphasis on "local computing" is actually a way to differentiate itself and bring AI to ordinary users amidst Nvidia's powerful cloud monopoly. Intel is indeed fighting a defensive battle, but it's also a battle they must win—they've identified a pain point: not all AI tasks need to be in the cloud. Privacy concerns (such as home surveillance), latency issues (such as game responsiveness), and cost all necessitate the decentralization of computing power. By deploying NPUs (Neural Processing Units) in PCs and edge devices, Intel is building the "capillaries" of the AI era. It's unhealthy if AI only exists in the hands of cloud giants; Intel is trying to make AI truly "ubiquitous."
The core concepts of several speeches—physical AI, local computing, breaking through bottlenecks, space—outline the complete path of AI moving from the "cloud brain" to the "physical world."
Based on the trends observed at CES 2026, my vision for the future can be summarized in three words: "intelligent agents, embodied intelligence, and technological solutions."
1. AI will emerge from the "dialog box": in the future, it will no longer be just a copilot, but a co-worker. We expect to see AI agents helping us book tickets, operate software, and even have robots clean our rooms.
2. The explosion of hardware forms: 2026 will be a year of explosive growth for humanoid robots and AI-defined cars. We will see more and more AI physical entities in factories and on roads.
3. Decreasing Costs: With the introduction of Nvidia's Rubin architecture and AMD/Intel solutions, AI inference costs will drop significantly. This means we will see more affordable and user-friendly AI applications, rather than just expensive luxuries.
Liu Xingliang: Renowned digital economy scholar and member of the Expert Committee on Information and Communication Economics of the Ministry of Industry and Information Technology
In these speeches, the key themes of physical AI, hybrid AI, and spatial intelligence are not presented in isolation, but rather as a co-evolving ecosystem. AMD and NVIDIA provide near-limitless computing power in the cloud, driving incredibly powerful models. These capabilities are then distributed to devices around us through Intel's hybrid architecture and the multimodal models envisioned by Fei-Fei Li, ultimately generating value in the real world through physical AI and robotics technologies championed by NVIDIA.
In this grand picture, Chinese companies are by no means bystanders, but rather crucial participants, and are expected to play a key role in the following areas:
The main battleground for application innovation: China boasts the world's largest and most diverse range of application scenarios and market demands. In fields such as e-commerce, social media, mobile payments, smart cities, and manufacturing, Chinese companies can combine globally leading AI foundational models and hardware with deep local insights to foster world-class AI application innovation. For example, China has a significant advantage in AI-enabled supply chains, personalized recommendations, and industrial automation.
A key link in the hardware industry chain: China occupies a central position in global electronics manufacturing and supply chains. From server manufacturing and AI terminal devices (PCs, mobile phones, robots) to data center construction, Chinese companies are an indispensable force in transforming advanced AI technologies into physical products and achieving large-scale delivery.
Breakthroughs in Specific Fields: Chinese companies have accumulated substantial technological strength in areas such as AI chip design (e.g., Huawei Ascend, Cambricon), autonomous driving (e.g., Baidu Apollo, XPeng, Huawei Inside model), and robotics. Facing international competition, they are expected to achieve differentiated technological breakthroughs by focusing on specific vertical sectors and leveraging the domestic market to achieve economies of scale.
Active contributors to the open-source ecosystem: A growing number of Chinese tech companies are embracing open source, contributing code, models (such as DeepSeek), and datasets globally. This helps enhance China's influence in the global technology community and allows them to benefit from collaboration.
The future depicted at CES 2026 is clear and exciting. Chinese companies need to leverage their strengths in markets, supply chains, and application innovation, actively integrating into the global technology ecosystem while boldly tackling challenges in core technologies. The future AI world will undoubtedly be a multipolar, collaborative, and highly competitive arena, and Chinese companies are destined to be among the main players.
