UBS raises AMD and Arm price targets as CPU demand for agentic AI systems takes off

UBS Raises Price Targets on AMD and Arm Amid Surging CPU Demand

UBS has dramatically increased its outlook on two major chipmakers, raising its price target on AMD to $670 from $470 and Arm Holdings to between $455 and $470 from $260, signaling strong confidence in both companies as key beneficiaries of the evolving AI chip market1

1

2

2

. The Swiss investment bank maintained its buy rating on both stocks, with analyst Timothy Arcuri pointing to accelerating traction in standalone CPU deployments as the primary driver behind the upgrades1

1

. The revised targets imply upside potential of approximately 28-29% from recent closing prices, reflecting Wall Street's growing recognition that AI-driven workloads extend far beyond graphics processing units.

Agentic AI Systems Drive Architectural Shift in Semiconductor Markets

The surge in CPU demand stems directly from the emergence of agentic AI systems—autonomous platforms designed to act independently without constant human intervention. This architectural evolution has created an unexpected windfall for AMD and Arm, with both stocks soaring over 150% in just three months1

1

. UBS now projects a 60/40 split between x86 and Arm architectures in the standalone segment, a significant shift that reflects how hyperscaler needs are reshaping chip design priorities2

2

. For AMD, which both designs and manufactures CPUs, this translates to projected server revenues of $23 billion in 2027, $29 billion in 2028, and potentially $50 billion by 20302

2

.

AMD's Advantage in Core Count and x86 Software Ecosystem

AMD holds distinct advantages in standalone agentic AI racks, which tend to be more core and throughput-driven compared to traditional AI infrastructure. Arcuri noted that AMD benefits from "its advantage in core count, multithreading, and x86 [processor's] historically strong software ecosystem for traditional software workloads that are now increasingly part of the agentic AI workloads"1

1

. UBS expects AMD to capture an outsized share of incremental demand, particularly given Intel's roadmap and supply challenges2

2

. Of 53 analysts covering AMD, 45 give it a buy or strong buy rating, underscoring broad market confidence1

1

.

Arm's Power-Efficient Architecture Aligns with Hyperscaler Needs

Arm Holdings, which came public in September 2023, brings a different competitive edge centered on latency and efficiency—qualities that align well with hyperscaler needs1

1

. UBS maintains that Arm architectures will represent approximately 70% share of a roughly 20 million unit head-node market by 20302

2

3

3

. The firm raised its estimate for Arm's internal CPU revenue to approximately $14 billion in 2030, reflecting about 8% share of UBS's $170 billion total addressable market estimate2

2

. Arm has secured major hyperscaler customers including Meta, OpenAI, and ByteDance3

3

.

Market Size Expectations and DRAM Supply Constraints

While investor enthusiasm has pushed some total addressable market estimates toward $300 billion by 2030, UBS takes a more measured view. Arcuri cautioned that DRAM supply constraints will likely prevent such aggressive outcomes, suggesting that "DRAM will remain too constrained to enable a TAM that large by 2030"2

2

. Current limitations in core count and throughput will also restrict use of Arm's first-generation internal product for standalone AI racks, where workloads need to scale across as many cores on a single machine as possible3

3

. However, Arm's roadmap of 200 to 500 cores could expand into broader use cases over time, though this likely requires second or third generation products3

3

. Arm CEO Rene Haas has emphasized the widespread integration of CPUs across applications, describing them as integral to the application space3

3

. With 19 analysts revising earnings upwards and more than half of the 40 analysts covering Arm issuing buy ratings, the market appears aligned with UBS's optimistic stance on both chipmakers' positioning in standalone CPU deployments1

1

3

3

.