Middle East warfare cannot suppress the narrative of the "AI bull market"! GPUs no longer dominate computing power, and the wave of intelligent agents is igniting the CPU and storage.

With the heavy launch of Claude Cowork by Anthropic, as well as the outbreak of super AI agent tools such as OpenClaw that can autonomously execute tasks in 2026, a wave of AI intelligent agents is rapidly sweeping the globe. The bottleneck of AI computing power architecture is shifting from GPUs that focus on matrix multiplication and throughput to data center CPUs that focus on control flow, task scheduling, and memory/IO coordination, creating a serious supply-demand situation for high-performance CPUs for super large-scale AI data centers. According to financial giants like Morgan Stanley, the narrative of AI computing power investment is shifting from "competition around AI GPU/ASIC single-point computing power" to "AI intelligent agents driving the artificial intelligence full-stack system", and in this shift, data center CPUs and storage chips may be the biggest winners of all. Financial giants like Morgan Stanley and Stifel believe that the two giants of PC and data center CPUs, Intel Corporation and AMD, are in the most advantageous position to benefit from the record-breaking expansion of data center CPU demand, particularly as the demand for high-performance server CPUs continues to grow due to the AI computing power infrastructure. Meanwhile, Wall Street's top analysts believe that storage chip giants will also benefit from exponential CPU demand expansion. Morgan Stanley believes that leading US storage manufacturers, Micron and SanDisk, are also in prime positions. As the benchmark of the South Korean stock market, KOSPI, dominated by Samsung and SK Hynix, hit a historic high amid the deteriorating political situation of Geo Group Inc, and with Taiwan Semiconductor Manufacturing Co., Ltd., known as the "king of chip manufacturing," leading the way for one of the biggest winners of the AI boom, the Taiwanese stock market also hit a historic high. With the latest strong performance outlook of ASML Holding NV ADR and Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR, investors are increasingly convinced that the "AI computing power investment theme" can overcome all market noise. However, the distribution of the value chain around AI computing power infrastructure is also beginning to shift, with the next round of excess alpha returns no longer limited to the strongest leaders in the AI GPU/AI ASIC field but systematically spreading to the full stack AI computing power infrastructure layer, including CPUs, storage, PCBs, liquid cooling systems, ABF substrates, and widespread wafer foundries. In this shift of the main narrative, financial giants like Morgan Stanley believe that CPUs and DRAM/NAND storage chips for data centers may be the core beneficiaries of AI computing power subcategories. Despite the ongoing Geo Group Inc political storm in the Middle East, Wall Street's bullish sentiment towards global stock markets is becoming more fervent. After a series of intense sell-offs initially, Wall Street institutions seem to be blocking out the noise related to war and no longer viewing war as the "core variable determining market direction" as they did in early March, but are largely "ignoring the noise of conflict." Many Wall Street financial giants directly attribute the current rise in stock markets to the upward trend in corporate profit expectations, especially in tech companies closely related to the exploding demand for AI computing power infrastructure. In the context of the Middle East conflict, the stock market is currently in the most profound phase of an upward trend from a temporary low point to a new high. Investors who have experienced this latest round of global stock market movements have undoubtedly learned that not every headline related to the Geo Group Inc should be treated as a trend in itself. The market has clearly demonstrated the core logic of "Trump's retreat in the face of adversity," the significant upward trend in earnings for tech companies driven by AI computing power infrastructure, and the eventual return of the US, Iran, and Israel to the negotiating table over the past week. The era of GPU dominance is temporarily coming to an end! As the wave of AI intelligent agents approaches, the demand for CPUs and storage chips continues to explode. The construction of AI data centers is pushing Intel Corporation's data center CPUs into a state of high demand. The longest lead times for high-performance server CPUs in demand at Intel Corporation have stretched to six months, and the prices of these high-performance server-level CPUs for data centers have increased by 10% so far this year. This is why the stock price of Intel Corporation, which has been languishing for a year and a half, has skyrocketed by over 80% this year, setting a new all-time high since 2000. In the intelligent agent network, a large amount of workload is not only spent on token generation on GPUs but also on CPU-dominated tasks such as Python interpretation and execution, web crawling, database retrieval, RAG indexing, lexical processing, task queue scheduling, RPC/IPC communication, KV state updating, and more. This means that what determines user experience is increasingly not just the peak computational power of a single GPU, but whether the CPU has enough cores, thread concurrency, cache levels, memory bandwidth, and PCIe/CXL/interconnect scheduling capabilities to support high-frequency tool calls and high-density task switching. If the CPU cores, memory subsystem, or I/O scheduling are insufficient, the GPU will experience a collapse in utilization due to data preparation, task coordination, and system waiting. Therefore, it is undeniable that the bottleneck of AI computing power architecture is shifting from GPUs focused on matrix multiplication and throughput to data center CPUs focused on control flow, task scheduling, and memory/IO coordination. The root of this change lies in the essential shift in the paradigm of workloads. CPUs are no longer just general-purpose computing chips but the control plane processor, system orchestration engine, and resource scheduling center of the intelligent agents, becoming the new bottleneck determining system throughput, latency, and resource utilization. The latest predictions from Morgan Stanley show that the explosion of intelligent agents marks a structural shift from computation to orchestration, leading to an additional $32.5 billion to $60 billion increment in the CPU market size by 2030. TrendForce's forecast report shows that in the era of AI intelligent agents, the CPU:GPU ratio may increase from the traditional 1:4 to 1:8 in AI data centers, to a range of 1:1 to 1:2. Furthermore, Morgan Stanley's forecast data also shows that by 2030, there will be an additional demand for 15 to 45 EB of DRAM storage chips, accounting for 26% to 77% of the total industry supply in 2027. The Morgan Stanley analyst team emphasizes that storage chips are becoming one of the most "profitable" layers in the AI computing power infrastructure system. Whether it's host-level DRAM, memory interface chips, or CXL extensions and tiered storage systems, they will all become important vehicles for capturing long-term value, as storage chips are no longer just capacity configuration options but core components that directly determine the efficiency and throughput of AI system workloads. Since 2026, DRAM/NAND storage chips have been continuing to surge. The latest memory price survey by the market research firm TrendForce shows that the overall price of conventional DRAM is expected to increase by 58%-63% in the second quarter of 2026, followed by an even steeper increase of 70%-75% from the first quarter's rise of nearly 100% due to the dominance of AI training/inference and extensive data center-related demand. Wall Street's core judgment on the supply-demand dynamics of storage chips currently suggests that the mismatch between DRAM/NAND Flash supply and demand may continue until around 2028, and the market is still underestimating the profit growth trajectory of this round of storage chip manufacturers. This is why the stock price of SanDisk has skyrocketed since early 2026, experiencing an extreme level of revaluation over the past year. This is essentially not just emotional speculation but a reevaluation of the valuation of "AI-driven storage." SanDisk's stock price skyrocketed by an astonishing 580% throughout 2025 and has continued to exhibit epic-level growth since the beginning of this year, surging by 300% and cementing its position as a stock market leader in AI computing power infrastructure. No matter how strong GPU/TPU computing power is, without HBM providing bandwidth, enterprise-level NAND and high-capacity HDD for training checkpoints, vector databases and inference data lakes, the utilization and efficiency of the entire AI computing power infrastructure cluster cannot be maximized; hence, global capital markets are willing to give higher valuations to the storage chain. This is because they have three leverage points: volume increase, price increase, and long-term supply constraints, rather than just single revenue growth. The entire AI value chain is reshaped by intelligent agents! The dividend of AI computing power infrastructure diffusion extends to data center CPUs and storage A team of analysts led by Morgan Stanley's senior strategist Joseph Moore released an investor report on Monday, stating, "The strength of CPUs brings obvious beneficiaries - Intel Corporation and AMD - and the exponential expansion of server CPU demand is crucial for the prospects of both companies to a considerable extent." "Between the two, we prefer AMD; additionally, at this point in time, we believe that storage chip manufacturers have a significantly better risk-return ratio, as the storage theme can be considered one of the direct beneficiaries of expanding CPU demand. Compared to directly investing in Intel Corporation or AMD, investing in popular storage chip stocks like Micron and SanDisk to capture the surge in CPU demand offers a better risk-return ratio." stated the team led by Morgan Stanley's Joseph Moore. Morgan Stanley maintains its "Overweight" rating on Micron and has a target stock price of up to $520. They have also raised Intel Corporation's target price from $41 to $56 and revised their 2027 earnings per share estimate from $0.97 to $1.34 - a forecast that is much higher than the market's general expectation of $1.03. Intel Corporation is scheduled to announce its first-quarter financial results for 2026 after the U.S. stock market closes on April 23. The market generally expects an adjusted EPS of $0.01 and revenue of around $12.42 billion. The analysts, including Moore, stated, Furthermore, we still maintain the view that Intel Corporation's broader investment assumptions revolve around its wafer foundry business, and we remain skeptical about this. We do expect to see some initial signs of growth interest, but in our view, it still remains uncertain to form a DCF positive wafer foundry business. However, we are indeed interested in the Terafab collaboration and are curious to see what this partnership will look like in terms of economic expansion for both parties. Earlier this month, it was revealed that Intel Corporation had joined Elon Musk's "Terafab epic chip manufacturing project" and is working with Musk's SpaceX, xAI, and Tesla, Inc. to "assist them in redefining silicon wafer manufacturing technology." On the other hand, the analyst team from Stifel, another Wall Street giant, stated that Intel Corporation and AMD are the biggest winners in the trend of data center CPU supply shortages, with AMD's target price being significantly increased from $280 to $320, while Intel Corporation's target price were raised from $42 to $65. "We still remain optimistic about AMD stock because Meta and OpenAI have made major strategic commitments at the teravolt level, and data center CPUs are experiencing counter-seasonal expansion, despite increasingly deteriorating supply constraints, coupled with the upcoming MI450/Helios; these factors together constitute what we believe to be a stronger incremental tailwind, especially relative to a long-term target EPS of over $20 per share from the management." said the Stifel analyst team. "On the other hand, Intel Corporation is a relatively different and longer-term investment logic: the progress of 18A advanced process yield and the repositioning of the wafer foundry business led by Pat Gelsinger are worth investors' attention, but short-term gross margins are still anchored in the low to mid-30% range, and the AI computing power strategy is (at least for now) basically absent, which continues to steer us towards other opportunities for the chipmaker, such as the strong growth opportunities brought by the surge in CPU demand." the team explained. Coming from Bank of America Corp, the stock analyst team recently stated that the rapid popularity of AI intelligent agents by Anthropic, especially in the enterprise management sector, continues to ignite an "AI agent workflow storm", which is a significant reinforcement of the entire AI infrastructure investment logic. Bank of America believes that as long as the demand for large models and intelligent agent computing power continues to grow exponentially, the most stable investment returns are likely to flow towards the "leaders providing AI computing power infrastructure." This excess investment return is shifting beyond the leaders in GPU/ASIC and expanding further to CPU and storage, the two new bottleneck segments. The urgent need for enterprises to improve efficiency and reduce operational costs has driven significant progress in the two core categories of AI application software - generative AI applications and the widespread use of AI intelligent agents. The AI intelligent agents (or AI agents) that autonomously perform various tedious and complex tasks are likely to be the ultimate trend in AI applications for the next decade, and the emergence of AI intelligent agents signals the evolution of artificial intelligence from an information assistant tool to a highly intelligent productivity tool. MarketsandMarkets' latest research shows that the AI intelligent agent market size could reach $53 billion by 2030, indicating a staggering CAGR of 46% that started in 2025. Furthermore, Anthropic plans to invest over $100 billion on Amazon.com, Inc.'s cloud computing service platform over the next ten years and plans to deploy approximately 1 gigawatt of computing power initially, gradually expanding to 5 gigawatts. At the same time, Anthropic has officially announced a multi-gigawatt next-generation TPU production capacity agreement with Alphabet Inc. Class C and Broadcom Inc., expected to be launched starting from 2027. Bank of America Corp's analyst team stated that Anthropic's active expansion of AI computing power infrastructure indicates that as long as the demand for large models and enterprise-level intelligent agents continues to grow exponentially, the first to realize revenue and actual orders are often not the model companies themselves but the "chip giants who sell the shovels" - especially leaders in GPUs, CPUs, DRAM - NAND storage chips, traditional cloud infrastructure platforms focusing on "cloud-based AI computing resources," and leaders in the "new cloud" like CoreWeave.