Why AMD Stock Popped: 12-Day Winning Rally Signals CPU Renaissance in Agentic AI Push

Since March 2026, Nvidia (NVDA) and Arm (ARM) both announced their entry into the CPU market, the signal that CPUs will become the next major trend has gradually become a market consensus.

However, prior to this, the role of CPUs in data centers had long been marginalized, even facing downward budget pressures due to the crowding out of GPU investments, as the market typically equated computing power with GPUs. This tilt is also reflected in the data. According to IDC, global server market revenue reached $235.7 billion in 2024, but more than 50% of that revenue came from AI-accelerated servers equipped with GPUs.

Looking at the most representative company in the CPU market, Intel (INTC) 's Xeon processors once held over 95% of the data center CPU market share for a long time. However, after peaking in shipments in 2021, they have continued to decline over the past few years, with 2024 shipments falling to their lowest level in 13 years—less than half of their peak.

During this period, due to the waning CPU market, AMD was seen more as a follower of Nvidia in the GPU market rather than an independent beneficiary. Although AMD's market share in the CPU market reached 28.8% by the fourth quarter of 2025, the market turned a blind eye to this.

Now, the market has changed. TD Cowen analyst Joshua Buchalter noted in a report last week that with the rise of AI agents, server CPUs are now seen as "indispensable assets."

The reason is that while GPUs are typically responsible for large-scale parallel computing inference, AI agents need to interact dynamically with their environment—including planning tasks, calling tools, passing data between sub-agents, and assessing task completion. These tasks are better suited for CPUs with complex scheduling capabilities.

This shift is already reflected in pricing: at the end of the first quarter of 2026, both Intel and AMD raised prices on some of their CPU product lines.

A recent report from TrendForce points out that CPUs in current AI data centers are significantly under-provisioned, with the current CPU-to-GPU ratio ranging from approximately 1:4 to 1:8. In the era of AI agents, this ratio is expected to evolve to between 1:1 and 1:2, implying multi-fold structural growth potential for CPU demand.

Currently, as a large number of non-traditional players enter the fray, the CPU market landscape is bound to undergo a dramatic shift. In March 2026, NVIDIA announced it would sell the Vera CPU as a standalone product and introduced the Vera CPU rack, which integrates 256 CPUs per rack for a total of 22,528 cores/45,056 threads and 400 TB of total memory, making this product extremely competitive in the CPU market.

In the same month, Arm announced its first proprietary CPU product, the Arm AGI CPU, with initial partners including Meta (META) , OpenAI, Cerebras, Cloudflare (NET) , and SK Telecom.

Furthermore, cloud service providers are also doubling down on the CPU sector by developing their own designs. For instance, AWS released its project in December 2025 based on TSMC's (TSM) N3-process Graviton5, to be deployed in synergy with its proprietary Trainium 3 AI ASIC; Microsoft (MSFT) launched Cobalt 200 in November 2025; Google (GOOGL) (GOOG) plans to launch the Axion C4A.metal bare-metal version and the next-generation Axion N4A in 2026. These are all cloud-native proprietary CPUs based on the Arm architecture.

While facing intense competition, some analysts pointed out that AMD's share price appeal is declining. Bernstein analyst Stacy Rasgon stated that Wall Street's current expectations for AMD already bake in an assumption of roughly 50% year-over-year growth in server sales, meaning the quarterly results due in early May must deliver on these high market expectations, or its valuation will be significantly discounted.

Furthermore, while AMD's market share in the GPU market is low, its importance remains significant. Rasgon said that although AMD's GPU sales volume is much smaller than NVIDIA's, the market believes this segment is gaining stronger upward momentum for AMD, and focus should remain on whether the Instinct MI400 series GPUs can truly break into the market.

Beyond the aforementioned AMD, NVIDIA, Arm, and Intel, it is important to monitor which companies may benefit from the current trend of major cloud service providers developing in-house CPUs.

As a global chip foundry giant, TSMC remains one of the primary beneficiaries; whether it is NVIDIA's Vera, AMD's EPYC, or the proprietary chips of cloud giants, all rely on TSMC's 2nm/3nm process nodes.

ASML (ASML) As the most upstream player in the global chip supply chain and a key supplier to the semiconductor industry, its lithography capacity directly dictates the global supply ceiling for advanced process chips, and the company will directly benefit from the expansion of CPU market demand.

Furthermore, attention should be paid to custom AI chipmaker Broadcom (AVGO) and Marvell (MRVL) , as Broadcom is a core partner for Google's TPU and Meta's inference chips, and Marvell is a key collaborator with Arm; both are set to benefit from the expansion of demand.

This content was translated using AI and reviewed for clarity. It is for informational purposes only.