The AI intelligent body frenzy detonates the CPU super cycle. The two major x86 chip giants join forces with Qualcomm (QCOM.US) to head towards the wild bull market.

With the explosive launch of the Claude Cowork by Anthropic, as well as super AI agents like OpenClaw that can autonomously execute tasks, the wave of AI intelligent agents is rapidly sweeping the globe in 2026. The bottleneck in AI computing power architecture is shifting from GPUs with matrix multiplication and throughput as the core, to data center CPUs with control flow, task scheduling, and memory/IO coordination as the core. High-performance CPUs for ultra-large scale AI data centers are facing a severe supply shortage. As AI intelligent agents sweep the globe, the mainstream focus of AI computing power investment is shifting from a "competition for computing power centered around GPUs" to an "all-stack computing power system driven by AI intelligent agents". The next round of excess Alpha returns will no longer only belong to the strongest leaders in the AI GPU/ASIC domain, but will systematically expand to include CPU, storage, PCBs, liquid cooling systems, ABF substrates, and extensive wafer foundry services in the full-stack AI computing power infrastructure layer. In this transformation of the AI narrative, CPUs, optical interconnects, and storage chips may emerge as the biggest winners. Against the backdrop of the surge in CPU demand, the stock prices of the two x86 architecture CPU giants - Intel Corporation (INTC.US) and AMD (AMD.US) - have recently skyrocketed to new historic highs. At the same time, Qualcomm (QCOM.US), which is returning to the data center CPU market, has seen its stock price surge by 60% in the past 10 trading days, highlighting investors' increasing bullish sentiment towards the "key bottleneck" in AI computing power. The well-known investment research firm GF Securities recently released a research report stating, "Data center server CPUs are entering a super cycle," with AMD (AMD.US), Intel Corporation (INTC.US), and Qualcomm (QCOM.US) likely to be the main beneficiaries. The "AI bull market" narrative cannot be restrained by the Middle East conflict! GPUs are no longer the only dominant force, as the trend of intelligent agents is driving CPUs. GF Securities analysts write in a recent report, "As the core coordinator of AI infrastructure construction, server CPUs have received significant attention in recent months. This trend is driven by the wave of agent-based AI/inference demand, driven by breakthroughs from OpenClaw and Anthropic, the latter of which saw its annual recurring revenue (ARR) soar to $44 billion in April. For example, in the latest earnings call, AMD expects the annual compound growth rate of the total addressable market (TAM) for server CPUs to exceed 35% by 2030, while Intel Corporation indicates that the shift to agent-based AI is driving unprecedented structural demand for CPUs, with the actual demand ratio in data center CPUs tightening from 8:1 in the training stage to 4:1 in the artificial intelligence inference stage." "Based on our estimates, assuming inference accounts for 90% of AI workloads, and assuming a GPU to CPU ratio in AI servers of 2:1 by 2030, we expect the total addressable market for server CPUs to grow by 54% in 2026/2027, and possibly reach a market value of $135 billion in 2030, compared to just around $26 billion in 2025 - representing a five-year compound annual growth rate (CAGR) of up to 38%. In addition, we expect the addition of server CPU demand in 2026/2027/2028 to be 6.7 million, 7.6 million, and 6.3 million units, respectively, with a total demand of 30 million, 38 million, and 44 million units during the same period, implying a three-year CAGR of approximately 23%," write the analysts at GF Securities. In a further analysis report, GF Securities points out that x86 architecture CPUs are still the most mainstream instruction set architecture in the data center server domain, and AMD and Intel Corporation are undoubtedly set to be the main beneficiaries of the CPU super cycle. The analysts explain, "From a fundamental perspective, Intel Corporation benefits from continued capacity acceleration and a significant increase in CPU prices in the second quarter, as we already explained in our report dated April 16. For AMD, with a robust product update and iteration roadmap and relatively ample capacity, we expect its market share to continue to rise. We currently anticipate that AMD's DCAI business under Intel Corporation and data center server CPU revenue will each grow by 39% in 2026." Intel Corporation may also gain a larger market share in the next generation of AI server CPUs, as its Xeon 6 central processor may be used in conjunction with NVIDIA Corporation's upcoming mass-produced Rubin NVL8 power cluster featuring a self-developed ARM architecture CPU. Some of the Vera Rubin's may be equipped with x86 architecture CPUs, while the others may have NVIDIA Corporation's self-developed Vera CPU based on ARM architecture; AMD's Venice CPU may be deployed in its next-generation Helios rack system and 8-GPU AI server power clusters. Despite x86 being the mainstream architecture, GF Securities believes that server CPUs based on ARM architecture may also experience accelerated growth, through NVIDIA Corporation's in-house ARM architecture server CPUs - Vera series power clusters and specialized application-specific integrated circuits (ASICS) such as Amazon.com, Inc.'s Graviton 5, Alphabet Inc. Class C's Axion 2 - all built on ARM architecture. The analysts add, "NVIDIA Corporation has launched independent Vera CPU racks (256 CPUs) for data center customers, with early adopters including Alibaba Group Holding Limited Sponsored ADR, CoreWeave, Meta, and Oracle (Oracle Corporation). Within the Alphabet Inc. Class C ecosystem, the GPU to CPU ratio of the TPU power system is adjusting from 4:1 to 4:2. Combining the stronger TPU computing demand outlook after 2026, we expect accelerated shipment of Axion 2. Meanwhile, Amazon.com, Inc. AWS's in-house Graviton 4/5 is also driving with Teton Max racks (18 CPUs) and Meta's independent Graviton racks. As for Arm itself, the company has raised its AGI CPU demand to $20 billion." GF Securities analysts also believe that Qualcomm, which recently announced its re-entry into the data center CPU market, is "actively developing" its 2028 line of data center CPU products, and is expected to receive more information on Investor Day on June 24th. The analysts add, "In an optimistic scenario, assuming Qualcomm obtains a 30% market share of the ARM architecture CPU market in 2028 (meaning about 4 million units, with an average selling price of $3,000 and a net profit margin of about 30%), the net profit it brings will be $3.6 billion, which is equivalent to a significant increase of about 30% in its full year non-GAAP profit." It is understood that Qualcomm's latest technological developments in data center CPUs and AI chips show that it is strategically expanding from a traditional smartphone chip design company to a supplier of data center AI infrastructure. Qualcomm has announced its re-entry into the data center CPU market, a clear departure from its previous focus solely on the mobile SoC sector. The company has confirmed it is developing custom data center CPUs that can work in conjunction with NVIDIA Corporation's AI GPU accelerators such as NVLink Fusion interconnect, to meet the requirements of future AI servers. It's worth noting that in the mid-2010s, Qualcomm launched the Centriq 2400 server processor based on the ARM architecture, which had a high core count and was designed for cloud computing and high-throughput workloads, indicating its experience in data center CPU design. In terms of data center-level AI accelerators, Qualcomm has launched the AI200 and AI250 accelerators designed for large-scale AI inference workloads, these high-performance AI chips are designed specifically for data center AI inference workloads and are planned for commercialization in 2026 and 2027, signaling Qualcomm's efforts to establish an AI chip product line that works in coordination with CPUs, to play a role in new AI workloads such as AI inference and agentic AI. Based on Qualcomm's exclusive Hexagon NPU architecture, these chips optimize memory capacity and energy efficiency, and can be configured as data center rack-scale solutions, indicating Qualcomm is catching up with NVIDIA Corporation and AMD in deploying large-scale AI computing infrastructure while striving to create a low total cost of ownership (TCO) AI inference platform.