SEOUL, November 12 (AJP) - Intel tried to keep the AI optimism alive in Seoul on Thursday as it followed Nvidia onto the conference stage while turmoil at its larger rival cast a gloom over the industry’s growth story.
The Intel AI Summit came days after Nvidia’s event in the same city.
The company that has drawn investment and attention away from Intel’s traditional processor business was facing a leadership crisis: Nvidia’s board was meeting to consider removing Chief Executive Jensen Huang amid scrutiny of his personal conduct.
Intel is hardly in a position to enjoy the rival's ill fortunes. Its shares plunged nearly 29 percent on Oct. 30 in their worst session since 1974 after its fourth-quarter outlook fell short of market expectations. The selloff came despite a 78 percent annual increase in third-quarter data center and AI revenue to $7.3 billion.
Its executives tried to sound bullish about the next phase of AI spending — and the prospect that it could bring central processing units (CPUs) back into the spotlight.
The global market for data center processors could exceed $200 billion over the next few years as AI moves from answering questions to executing tasks, said Anil Nanduri, Intel’s head of AI products and go-to-market for its Data Center Group.
His argument was that autonomous AI agents would require more CPUs, memory and networking equipment alongside the graphics processing units (GPUs) that have powered the investment boom.
“As more models are executed, more models are generated and more tokens are generated, a lot more sandboxing and execution and tool calls are happening,” Nanduri told reporters at a briefing following the summit.
In practical terms, AI systems that carry out assignments must run software, retrieve information and coordinate multiple processes. Intel sees those tasks creating more work for the general-purpose processors at the heart of its business.
GPUs remain essential for training and running large AI models. But as agents interact with applications and one another, CPUs help coordinate their work, execute commands and keep data moving.
That shift could give Intel a fresh opening in an AI infrastructure market dominated by Nvidia.
It could also broaden opportunities for South Korean suppliers. Nanduri said growing machine-to-machine interactions would increase demand for memory and networking infrastructure, extending the investment cycle beyond the accelerators that have captured most of the attention.
Intel Korea CEO Bae Tae-won said CPU supplies could remain tight next year as demand grows. The company was coordinating with its headquarters and Korean partners on supply priorities, he said.
Intel is also seeking a larger role in South Korea’s push to build more of its AI infrastructure at home.
The company is participating in the K-AI Full Stack Alliance, launched Thursday to connect Korean server manufacturers and domestic technology companies through an open AI infrastructure ecosystem.
Asked about the dominance of major technology vendors, Intel executives emphasized open standards and cooperation with local companies.
Nanduri cited Compute Express Link (CXL), a technology that allows computing systems to expand and manage memory resources more flexibly, as one area for collaboration with Korea’s semiconductor industry.
Intel’s pitch also includes an AI accelerator of its own.
At the summit, the company showcased Crescent Island, now known as the Intel GPU D-series, aimed at inference — the process of running trained AI models to produce answers or carry out tasks.
The design uses low-power LPDDR5X memory instead of the high-bandwidth memory found in many leading AI accelerators. Intel’s own card carries 160 gigabytes of memory, while partners can build versions with up to 480 gigabytes. The card is designed around a 350-watt power envelope.
The emphasis is on memory capacity, power efficiency and deployment costs, reflecting the growing importance of how economically businesses can operate AI systems.
Nanduri said Crescent Island was aimed particularly at enterprise customers seeking affordable inference capacity. He did not present it as a universal replacement for competing accelerators.
Intel demonstrated early performance results but gave no firm commercial deployment date during the press briefing.
Beyond data centers, the company is pursuing robotics, factory automation and other forms of physical AI, in which computing systems interpret their surroundings and respond in real time.
Dan Rodriguez, vice president and general manager of Intel’s Edge Computing Group, said the company was working with established Korean technology businesses and startups on applications in manufacturing, healthcare and other industries.
During his keynote, Rodriguez highlighted a partnership with Samsung Medison on AI-assisted ultrasound imaging. The work uses Intel processors and software to automate measurements that traditionally require manual analysis.
He also announced a collaboration on reference architectures for autonomous factories — shared technical blueprints connecting robotics, industrial equipment and computing infrastructure.
Asked by AJP whether discussions with Korean companies had progressed toward commercial deployment and whether concrete timelines were available, Rodriguez pointed to existing partnerships but stopped short of giving specific schedules.
He said Intel was working with companies of different sizes, including startups developing applications for edge computing, which processes data close to where it is generated, and physical AI.
Intel said its Open Edge ecosystem now includes more than 180 qualified AI systems from over 60 hardware partners, alongside more than 160 independent software vendors.
For Korea, Intel’s strategy offers potential openings across memory chips, locally built servers and industrial applications. For Intel, it offers a chance to regain relevance as AI spending spreads to the systems needed to put the technology to work.
The summit supplied a bullish case for that expansion. Commercial timelines and sales will determine how much of it reaches the factory floor and the balance sheet.
Copyright ⓒ Aju Press All rights reserved.