[Real-World Asset Management] CPUs in the Spotlight in the Age of Agentic AI... Promising Investment Picks

"After GPUs, CPUs Take Center Stage"
Brokerages Say "AI Investment Focus Is Shifting"
Potential for CPU Supply Shortages

The investment trend in the artificial intelligence (AI) semiconductor market is shifting once again. As AI evolves beyond the training stage into inference and agentic AI, demand, once heavily concentrated on GPUs (graphics processing units), is now spreading to CPUs (central processing units).

Spread of Inference and Agentic AI... Agentic AI Takes Command of Entire Work Process

Recently, following Intel's first-quarter earnings surprise, the stock prices of CPU companies have shown a pronounced upward trend. In particular, Intel's Data Center and AI (DCAI) revenue increased by 22% compared to the previous year.

This surge in revenue has been accompanied by a sharp rise in stock price. Although Intel was removed from the Dow Jones 30 Industrial Average in 2024 after 25 years, its stock price has climbed approximately 170% so far this year on growing expectations for CPU demand. In April alone, the stock price soared by as much as 119%.

Jaeseung Kim, a researcher at Hyundai Motor Securities, explained, "With DCAI revenue growing by 22% year-on-year, the expansion of the CPU's role and the growth in demand have become apparent. As AI advances from training to inference and agentic AI, the role of the CPU in orchestrating work is coming back into the spotlight." Whereas early AI investments focused on training, emphasizing GPUs, HBM (high-bandwidth memory), and high-speed networks, the transition to the service stage has increased the proportion of non-computational tasks such as model selection, data retrieval, and tool invocation.

"CPU Supply Shortage" ... Price Increases Accelerate

In agentic AI, the process of "decision-making → tool selection → execution → verification" is repeated, elevating the importance of CPUs in coordinating tasks. Jaegoo Kang, a researcher at Hana Securities, analyzed, "In the inference market, CPUs play the role of a 'traffic controller' in AI systems, supplying data to keep GPUs utilized and maximize efficiency." Jaehwan Park, a researcher at Eugene Investment & Securities, also noted, "In the era of AI agents, CPUs take on the core role of orchestrating all tasks."

The structure of data centers is also changing. Analysts suggest that the CPU-to-GPU ratio, previously at 1:4 to 1:8, could decrease to as low as 1:1 or 1:2 in an agentic AI environment.

Along with demand, supply issues are also coming to the fore. CPUs are structurally more challenging to ramp up production than GPUs, making it likely that bottlenecks will recur during supply expansions. Server CPUs rely on TSMC's advanced process nodes and on advanced packaging technologies such as CoWoS and SoIC that bundle multiple chips, but these production capacities are already saturated due to strong demand for GPUs and smartphone semiconductors.

Yonghee Han, a researcher at Growth Research, noted, "CPU lead times have increased from the previous 1-2 weeks to 8-12 weeks, with some products facing wait times of over six months. Intel and AMD are raising server CPU prices by about 10-15%." Junhyuk Ko, a researcher at Shinhan Investment, stated, "With the paradigm shift in AI computation, CPUs have entered a phase of structural supply shortage. As the CPU-to-GPU ratio narrows to 1:1, the trend of price increases is gaining momentum."

CPU Investment Spreads Across the Value Chain

Investor interest is expanding beyond CPUs to include substrates, packaging, and testing—the entire back-end value chain. Kang emphasized, "Although CPUs are becoming increasingly important, it is difficult to conclude that the benefits will go directly to Intel. AMD may be more advantageous in the traditional data center AI transition, and Arm-based CPUs are likely to spread more rapidly in large-scale AI inference data centers."

He added, "Nvidia CPUs adopt Arm-based designs, and hyperscalers are also expanding their own Arm-based CPUs. As AI inference grows within the cloud, some x86 CPU demand may be absorbed by Arm-based CPUs."

Han noted, "CPU supply shortages are not just an issue for CPU vendors, but the benefits could spread to memory, flip-chip ball grid array (FC-BGA) substrates, packaging, and testing." For server CPUs, high-spec FC-BGA substrates and advanced packaging technologies are essential, and securing memory bandwidth to support CPU performance is also required.

From a domestic investment perspective, it is necessary to focus more on value chain companies than on CPU design companies. Specifically, Samsung Electro-Mechanics and Daeduck Electronics, which produce FC-BGA substrates, as well as companies involved in back-end processing and testing, are being highlighted as major beneficiaries. Expectations are being reflected in substrate companies as well. Euiyoung Ko, a researcher at iM Investment & Securities, commented on Simmtech, saying, "With the spotlight on agentic AI, rising CPU demand and the launch of Nvidia's Vera CPU rack could further expand upside for SOCAMM-related revenue."

This content was produced with the assistance of AI translation services.