UBS Group indicates that the ongoing acceleration of agentic AI is poised to benefit Advanced Micro Devices (AMD.US) and ARM Holdings (ARM.US) from heightened demand for dedicated CPU racks.

While Intel (INTC.US) could also play a role with its x86 chips, its medium-term upside is contingent on improvements in its high core-count and performance roadmap. In the x86 processor arena, AMD stands as Intel's primary competitor.

An analyst team at UBS, led by Timothy Arcuri, stated in an investor note on Wednesday that their stance on AMD is becoming incrementally more positive as dedicated CPU racks gain favor.

This is due to AMD's advantages in core count and multithreading, coupled with x86's long-established and robust software ecosystem for traditional workloads, which are increasingly becoming part of agentic AI workloads.

The analysts now assume a 60% to 40% market share split between x86 and Arm in the dedicated chip segment of the market.

Given Intel's roadmap and supply challenges, they believe AMD is well-positioned to capture the lion's share of this new demand.

UBS significantly raised its price target for AMD from $455 to $670.

The firm also increased its CPU revenue forecast for the 2027 calendar year from $21 billion to $23 billion and raised its 2028 estimate from $27 billion to $29 billion.

Concurrently, UBS substantially lifted its price target for Arm from $260 to $470.

The bank maintains its previous view that by the 2030 calendar year, Arm architecture will capture approximately 70% of a total addressable market of around 20 million head nodes.

Arcuri added that the real debate among investors, in their view, centers on the revenue potential for Arm's dedicated CPUs.

Following a series of expert calls and industry discussions, the bank has become slightly more optimistic about Arm's dedicated CPU prospects.

It has modestly raised its internal CPU revenue expectation for Arm for the 2030 calendar year from about $13 billion to approximately $14 billion.

Arm's core competencies lie in latency and efficiency, which align well with hyperscaler needs.

However, current limitations in core count and throughput are expected to restrict the use of Arm's first-generation internal general-purpose AI in dedicated AI racks, as workloads for such applications need to scale across as many cores as possible on a single machine.

Beyond the opportunity in dedicated CPUs, UBS also anticipates growth in Arm's royalty segment for head node applications.

Arcuri noted that they expect the royalty rate per device to increase, driven by higher core counts and an increase in per-core royalties, which in their bullish scenario could rise by more than 2x or even 3.5x.