"Agent AI" rises, is the CPU making a comeback?

Over the past three years, the narrative of the AI server market has been highly singular: whoever owns more GPUs holds the key to artificial general intelligence (AGI).

However, as AI applications evolve from simple chatbots to “Agentic AI” that can automatically execute complex processes, the fundamental computing power demands of data centers are undergoing a profound “paradigm shift.”

According to Wind Trade Desk, HSBC Securities stated in their latest in-depth industry report that a “rebalancing of computing power” era has begun, with CPUs once again becoming the decision-making hub of AI data centers.

From “chatting” to “doing”: the seismic shift in computing power driven by Agentic AI

Early AI models mainly relied on large-scale GPUs for parallel processing, known as “brute-force aesthetics.” But the workflow of Agentic AI is entirely different—it’s a “super orchestrator.”

HSBC’s research report points out that Agentic AI not only needs to generate outputs but also requires retrieval-augmented generation (RAG) and access to external APIs (such as email systems, web searches, collaborative office software) to complete tasks. These involve complex logical branching, task synchronization, and multi-threaded processing.

Comparative data shows that in Agentic AI tasks, the parallel advantage of GPUs is diminished, while CPUs shine with their flexibility in handling multiple tasks.

According to recent research from Cornell University, about 44% of the computing power in Agentic AI workflows depends on CPUs, which is 3-4 times higher than in traditional AI workflows. CPUs are no longer just system “assistants” but have become the indispensable “brains” of the computing engine.

Why are server shipments expected to continue soaring? Supply far below demand

This structural change in computing power demand directly supports strong growth in server shipments.

HSBC Securities significantly raised their 2026 global server shipment growth forecast to 20% year-over-year (from 4%), far exceeding the market’s general expectation of 10%-15%. Even more astonishing, HSBC believes the market’s “latent demand” could be as high as 60%.

Why hasn’t the market fully met this demand?

The main reason lies in supply chain bottlenecks. The report emphasizes that the current shortages of CPUs and DRAM memory are as high as 30%-40%, with core components like SSDs and power chips also facing 10%-30% supply deficits.

This means that in the next two years, global server shipment growth will be constrained, with a “toothpaste squeezing” release. This ongoing supply shortage results in a large backlog of orders, not only supporting the shipment logic for 2026-2027 but also convincing HSBC that the global server growth cycle could extend into 2028.

Capital expenditure boom: “The Big Beautiful Bill” as an “additional fuel”

Supporting this hardware frenzy are not only technological upgrades but also policy incentives.

According to HSBC’s estimates, with the implementation of the U.S. “The Big Beautiful Bill,” companies purchasing data center assets (including servers) can enjoy 100% bonus depreciation and immediate R&D expense deductions. This provides strong financial incentives for cloud service providers (CSPs) to invest in hardware.

HSBC predicts that major cloud providers (including Google, Microsoft, Amazon, Meta, and Oracle) will have a combined capital expenditure of $743 billion in 2026, an 81% increase year-over-year.

This influx of capital is shifting structurally: in 2025, about 35% of CSPs’ capital expenditure was allocated to servers; by 2027, this will jump to 59%. This resource tilt from “land acquisition and building” to “purchasing computing equipment” precisely benefits the server supply chain.

Who benefits? The “rebalancing” dividend in the supply chain

Market dynamics are shifting from a simple “GPU logic” to a more balanced “compute node value chain.”

HSBC notes that as the value of CPUs in racks increases, server ODMs (Original Design Manufacturers) and key component suppliers will become the main beneficiaries of this computing power rebalancing.

In summary, this computing power arms race has entered the “deep water” stage. The past was “who owns the graphics cards wins”; the future is “who manages computing power distribution efficiently wins.”

The resurgence of CPU demand driven by Agentic AI not only injects long-term certainty into the server market but also outlines a clear hardware upgrade logic for investors.

This insightful content is provided by Wind Trade Desk.

For more detailed analysis, real-time insights, and frontline research, please join【Wind Trade Desk▪Annual Membership】

![](https://img-cdn.gateio.im/social/moments-3dd50f77a6-39610e97d5-8b7abd-ceda62)

Risk Disclaimer and Terms of Liability

Market risks exist; investments should be cautious. This article does not constitute personal investment advice and does not consider individual users' specific investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions herein are suitable for their particular circumstances. Invest at your own risk.