UBS and TD Cowen raised Arm's target price to $475 on the same day, citing future revenue from self-developed CPUs.

UBS significantly raises Arm price target from $260 to $470, while TD Cowen raises from $265 to $475; both investment banks maintain Buy ratings on the same day. The core logic is that with the rise of agentic AI, the strategic position of the CPU in data centers is rising, and the GPU no longer monopolizes the investment narrative. UBS estimates Arm's self-developed CPU-related revenue could reach about $14 billion by 2030, but Arm itself admits that the business will not materially impact finances until fiscal 2028.
(Previous Context: OpenAI teams up with Broadcom to unveil first AI chip 'Jalapeño', achieving in 9 months to challenge Nvidia's dominance)
(Background: JPMorgan sharply raises Micron price target to $1,540, surpassing Bank of America's view: AI memory supercycle just starting)

Table of Contents

Toggle

• Agentic AI shifts work from GPU to CPU
• The market is really arguing about how much the self-developed CPU is worth
• Stock price falls, price target rises, investment banks bet on 2030

Key Takeaways

• UBS raises Arm price target to $470 (from $260), TD Cowen raises to $475 (from $265), both maintain Buy ratings on same day
• Core bet: Agentic AI drives CPU demand; UBS estimates Arm self-developed CPU revenue reaches about $14 billion by 2030, roughly 8% of server CPU market
• Arm stock price around $357, down about 12.5% in the past week, but up about 235% year-to-date, market cap about $381 billion

Stock down 12.5% in a week, but two major investment banks raise price targets by over 80% on the same day—this contrast itself explains the issue: the market is buying not this quarter, but the architectural story of 2030. UBS raises Arm price target from $260 to $470 in one go, TD Cowen raises from $265 to $475, both maintain Buy ratings. The core bet is only one: agentic AI (AI agents that can decompose tasks and act sequentially) is rewriting the division of labor between CPU and GPU in data centers.

Agentic AI shifts work from GPU to CPU

In the past few years, the AI investment narrative was almost equal to Nvidia and GPU, but the working pattern of agentic AI is breaking this equation. UBS and TD Cowen have a similar analytical framework: the "thinking tasks" of AI are handled by GPU-intensive computation, but when AI agents start "doing tasks," i.e., actually operating tools, coordinating steps, calling APIs, the CPU's low latency and energy efficiency advantages are key. Arm architecture's strengths lie precisely here.

TD Cowen analyst Sankar believes Arm's self-set FY31 target of $15 billion in AGI CPU revenue is achievable. UBS's forecast is slightly more conservative, estimating about $14 billion by 2030, but still up from the previous estimate of $13 billion. The server CPU market itself is also expanding; UBS estimates the market could grow from about $30 billion in 2025 to about $170 billion by 2030, with Arm's $14 billion accounting for about 8%.

The market is really arguing about how much the self-developed CPU is worth

UBS points out that the core debate over Arm's stock price is not the overall licensing model, but the revenue potential of Arm's self-developed or independent CPU business. Arm architecture has established a beachhead in cloud vendors' AI infrastructure expansion, relying on energy efficiency and low latency—two metrics increasingly important in large-scale AI inference workloads. UBS estimates that by 2030, Arm architecture will capture about 70% of the potential market for about 20 million "head-nodes" (AI server main control nodes).

Notably, Arm itself makes it clear that its internal chip business will not materially impact finances until fiscal 2028 (FY2028, Arm's fiscal year differs from the calendar year). In other words, analysts' current calls of $470, $475 are pricing a business that will only start generating revenue in 2028 and reach full speed by 2030; the implied volatility of such pricing is inherently high.

"The strategic position of CPU in agentic AI architecture may surpass GPU in some workloads."

Stock price falls, price target rises, investment banks bet on 2030

Arm's current stock price is about $357, down about 12.5% in the past week, but still up about 235% year-to-date, with a market cap of about $381 billion. UBS also raised AMD's price target on the same day, listing both AMD and Arm as major beneficiaries of CPU demand driven by agentic AI, placing bets on both with the same direction. Short-term pullbacks are noise for analysts; they are watching the growth curve of the server CPU market expanding from $30 billion to $170 billion by 2030. The next observation point is whether Arm's current fiscal year guidance aligns with the investment banks' rhythm; if the numbers don't match, the $475 price target remains a hypothesis waiting to be validated.

FAQ

Why does agentic AI make CPU more important than GPU?

Agentic AI (AI agents that can decompose tasks and act sequentially) requires a lot of coordination and tool calling during the "task execution" phase. These workloads are latency-sensitive and have high energy efficiency requirements, which are exactly the strengths of CPUs, while the parallel computing advantages of GPUs are relatively limited here.

When will Arm's CPU business have a material financial contribution?

Arm officially stated that its self-developed CPU-related business will not materially impact finances until fiscal 2028 (FY2028). UBS estimates related revenue could reach about $14 billion by 2030, while Arm's self-set target is to achieve $15 billion by FY31.