During COMPUTEX 2026, Lip-Bu Tan, who has been Intel CEO for 14 months, stated that he is the first CEO in Intel's history to deliver a keynote speech fluent in Mandarin (and to communicate directly with the local supply chain in Taiwan). Furthermore, addressing external criticisms of Intel, Tan clearly stated that "execution" remains the top priority on his list.
Chen Liwu emphasized that Intel's soul is that of an "engineering company." From his first day in office, he personally reviewed all engineering reports, striving to bring the focus back to the execution and delivery of core technologies.
In this keynote speech, Intel stated that it has launched the Core Ultra Series 3 processor for various edge computing scenarios, the Arc G3 series processor for handheld devices, and even the Xeon 6+ for servers, fully announcing that Intel's 18A process technology has entered the stage of large-scale mass production and commercial application.
Intel's 18A process goes all out: Core Ultra Series 3 and Arc G3 target the terminal and handheld markets.
Regarding PC computing applications, Alex Katouzian, who previously worked at Qualcomm for a long time and then moved to Intel to serve as the general manager of the User Computing Group and the Physical AI Division, pointed out that the Core Ultra Series 3 processor, which is built entirely with Intel's 18A process technology, has accumulated more than 70 design products to enter the market in just a few months since its official launch in April this year, with the overall product lineup reaching nearly 400 designs.
In response to the explosive growth of the handheld gaming market in recent years, Intel announced the launch of a device also based on the 18A process and the Panther Lake architecture.Arc G3 series processorsCompared to the previous generation of products built on the Lunar Lake architecture, the Arc G3 offers up to 40% better performance at the same power consumption, and its power consumption is halved when operating at the same performance.
In addition, Alex Katouzian explained that Intel not only has technological advantages in semiconductor manufacturing processes, but also has a comprehensive layout in system and software design, as well as ecosystem scale, encompassing...Multiple types of robotsThe fact that physical AI has also made significant progress means that Intel's 18A process technology can now penetrate various computing scenarios.
From "Generative" to "Proxy": "Hybrid Proxy AI Inference" Becomes a New Prominent Field of Study
Compared to the large-scale language models of generative AI that the industry has been emphasizing in the past few years, Chen Liwu clearly defined the next stage of AI development in this speech - "Agentic AI".
The AI of the future will no longer passively generate text or images, but will transform into digital agents capable of understanding goals, planning steps, and proactively executing tasks. During a presentation alongside Perplexity founder and CEO Aravind Srinivas, the two companies showcased their collaborative "Hybrid Agent AI Inference." Using devices equipped with Intel Core Ultra Series 3 processors, the Perplexity browser's AI agent function, "Computer," can simultaneously invoke up to 20 different AI models to perform designated tasks as AI agents.
On-site, an employee of a private equity investment firm was asked to analyze a confidential investment project codenamed "Project Falcon". The AI agent first read the relevant files and analyzed which information was confidential and sensitive data. Then, the appropriate data was sent to the cloud for analysis and processing, while the confidential content was left on the device for analysis.
Chen Liwu stated that while more computations will be performed in data centers in the future, an even greater number of computations are expected to be performed on devices. This architecture, which perfectly combines local security with cloud computing power, will become key to enterprises adopting AI in the future.
Servers and Custom Chips: Xeon 6+ Server Processors and the Future of Neuromorphic Computing
At the data center level, in response to the massive inference demands brought about by proxy AI, Intel officially announced at COMPUTEX 2026...Xeon 6+ server processors built using the 18A processFurthermore, it emphasizes that the computing architecture of future data centers will move from traditional server units to "rack-scale computing" and disaggregated architectures, and has also begun in-depth cooperation with Foxconn.
To this end, Intel has partnered extensively with numerous OEMs, ODMs, and industrial manufacturers, including Siemens, to provide more flexible infrastructure for massive data computing.
On the other hand, Chen Liwu also announced the recent developments of Intel Custom Foundry, clearly pointing out that technology companies including Google and Ericsson have now become its custom chip solution partners.
At the end of his speech, Chen Liwu explained that Intel is collaborating with the University of California, San Francisco (UCSF) and Echo Neurotechnologies on the field of "neuromorphic computing." This computing architecture, which mimics the workings of the human brain, is expected to provide a more efficient and lower-power computing mode than existing AI chips in the future.
Intel returns to the essence of engineering, letting execution determine victory.
The keynote address at COMPUTEX 2026 revealed that, 14 months after Chen Liwu took the helm, Intel's approach has become more pragmatic and precise. It's no longer just about grand pronouncements about the future, but about demonstrating its "execution capabilities" through the full rollout of the Intel 18A process and a complete product line covering PCs, handhelds, edge computing, and cloud data centers.
At this pivotal moment when AI technology is shifting from "dialogue generation" to "agent execution," Intel is attempting to redefine the rules of computing for the next decade through its powerful underlying architecture and hardware-software synergy.
