UBS Sharply Raises AMD Target Price to $670 and Arm to $470
Miles Bennett
UBS lifted its AMD price target from $455 to $670 and Arm from $260 to $470, arguing that agentic AI is creating a new hardware category — standalone CPU racks — where each company owns a distinct lane.
Why did UBS raise both targets so aggressively?
UBS believes the market underestimates a new category: standalone CPU racks — not CPUs supporting GPUs, but dedicated cabinets running agentic AI inference (AI that autonomously plans, calls tools, and completes multi-step tasks).
In plain terms = the AI compute spotlight has been on GPUs; UBS now argues CPUs deserve their own stage.
AMD's target jumped 47%, Arm's 81%. This reflects not a near-term earnings tweak but a full re-sizing of the category's addressable market.
Why does AMD capture the largest share?
UBS assumes a 60/40 split between x86 and ARM architectures in standalone CPU racks, and within x86, AMD takes nearly all incremental demand.
This reflects Intel's current struggles: roadmap and supply uncertainties mean most new orders flow to AMD by default.
UBS raised AMD's 2027 CPU revenue estimate to $23 billion and 2028 to $29 billion, roughly 10% above prior forecasts.
This means → AMD's core-count advantage and the mature x86 software ecosystem form a moat in the heavily multi-threaded workloads that agentic inference demands.
Why is Arm's story different?
Arm's strength lies in low latency and power efficiency, which fits the custom-chip strategies of hyperscalers — companies like Amazon AWS and Microsoft Azure.
But UBS flags a near-term gap: core count and throughput are still limited. First-generation general-purpose AI workloads need to scale across as many cores as possible on a single machine, and Arm has not closed that gap yet.
In plain terms = Arm excels at "fast and frugal," but agentic inference needs "many threads working in parallel." The two demands are not yet fully reconciled.
Does Arm have an underappreciated revenue line?
Beyond selling its own CPUs, Arm earns royalties — a fee on every chip built on the ARM architecture.
UBS expects per-device royalty rates to rise as head-node chips (the most critical compute nodes in a data center) pack in more cores.
In a bull case, per-core royalties could increase more than 2× and up to 3.5×. This means → even if Arm's own CPU market share grows slowly, revenue scales significantly just from others building on the ARM architecture.
UBS accordingly raised Arm's 2030 internal CPU revenue estimate from ~$13 billion to ~$14 billion.
Where does Intel fit in this story?
UBS acknowledges Intel's x86 chips could also benefit, but upside hinges on timely delivery of high-core-count products.
Roadmap uncertainty is the binding constraint — this reflects a market that has not yet regained confidence in Intel's execution.
In plain terms = Intel holds an entry ticket, but whether it gets to play depends on product milestones.
What is the key checkpoint for this thesis?
UBS itself names the test: whether AMD and Arm can deliver results in the standalone CPU rack category is the core premise behind both target raises.
This means → if agentic AI inference demand does not generate standalone CPU rack purchases at sufficient scale, both targets face downward revision.
The signal investors should track: whether standalone CPU racks begin to show up as a quantifiable line item in hyperscaler capital-expenditure disclosures.
Content is for reference only, not financial advice.