Wall Street’s AI trade is shifting again. After years in which GPUs dominated nearly every artificial-intelligence investing conversation, Bernstein is now turning more bullish on the companies powering the other side of the data-center equation: server CPUs.
Bernstein raised its price targets on Arm, AMD and Intel, arguing that the rise of agentic AI could sharply increase the need for CPUs inside AI data centers. According to Seeking Alpha, the firm lifted its server CPU total addressable market forecast for 2030 to $223 billion, up from a prior estimate of $137 billion. That would represent roughly six times the estimated $37 billion server CPU market in 2025.
The call matters because it broadens the AI infrastructure story beyond Nvidia-style accelerators. GPUs remain critical for training and inference, but agentic AI workloads may require far more CPU orchestration, memory management, scheduling and data movement than earlier chatbot-style AI applications.
Bernstein Sees a “CPU Renaissance”
Bernstein’s latest view is built around a simple but powerful idea: AI is becoming more complex, and that complexity requires more general-purpose compute.
Investors Business Daily reported that Bernstein raised AMD’s price target to $600 from $525, Arm’s price target to $500 from $300, and Intel’s price target to $100 from $65. Bernstein maintained Buy ratings on AMD and Arm while keeping Intel at Neutral.
The market reaction was immediate. According to Investors Business Daily, AMD stock rose to $512.48, Arm jumped to $418.88, and Intel climbed to $121.10 after the note.
The phrase “CPU renaissance” is becoming increasingly common in semiconductor research because AI workloads are moving from simple prompt-response systems toward multi-step agents. These agents can retrieve information, plan actions, call tools, interact with databases, manage memory and execute tasks across software environments. That process can put much more pressure on CPUs.
Why Agentic AI Changes the CPU Story
Agentic AI is different from traditional generative AI. A chatbot may take a prompt and return an answer. An AI agent may break a task into steps, search external data, run code, call APIs, store intermediate results and revise its plan. That workflow requires constant coordination between compute resources.
A recent academic paper on agentic AI workloads found that tool processing on CPUs can account for up to 90.6% of total latency in some workloads, while CPU dynamic energy can represent up to 44% of total dynamic energy at large batch sizes. The authors argued that CPU-side bottlenecks can significantly affect latency, throughput and energy efficiency in agentic AI systems.
That helps explain why analysts are looking again at server CPU suppliers. GPUs remain essential for model execution, but CPUs often manage the orchestration layer around AI agents. If agentic systems become mainstream across enterprises, cloud platforms may need more CPUs per GPU than they used in earlier AI architectures.
Yahoo Finance previously reported that Northland analyst Gus Richard sees the CPU-to-GPU ratio shifting from roughly 1:8 for training to 1:4 for inference and potentially 1:2 for agentic AI. TrendForce has also been cited as expecting agentic AI deployments to move toward CPU-to-GPU ratios between 1:1 and 1:2, compared with current AI data-center ratios of roughly 1:4 to 1:8.
Arm May Be the Biggest Sentiment Winner
Arm was the standout in Bernstein’s target increases, with the firm lifting its target to $500 from $300. That reflects growing investor enthusiasm around Arm’s role in power-efficient data-center computing.
Arm’s appeal is tied to its licensing model and energy efficiency. Cloud providers and chip designers can use Arm architecture to build custom processors optimized for specific workloads. In a world where AI data centers face power constraints, cooling challenges and rising electricity demand, efficiency becomes a major selling point.
Arm’s advantage is not that it directly competes with Nvidia’s GPUs. It is that Arm-based CPUs could become more widely used alongside accelerators in agentic AI systems, cloud infrastructure and custom silicon designs.
The risk is valuation. Arm has already been treated as a premium AI infrastructure stock, and Bernstein’s target increase adds to already high expectations. Investors will want proof that agentic AI demand translates into licensing growth, royalty expansion and data-center share gains.
AMD Gets a Broader AI Narrative
AMD also benefits from the server CPU renaissance thesis. The company is already viewed as Nvidia’s most visible challenger in AI accelerators, but its EPYC server CPUs are equally important to its data-center strategy.
Bernstein’s higher $600 price target suggests investors may need to value AMD as both a GPU competitor and a CPU beneficiary. That is a stronger story than GPU upside alone.
AMD has been gaining attention because agentic AI may require both CPUs and GPUs. Seeking Alpha previously reported that AMD shares rose after company executives highlighted how agentic AI is driving demand for high-performance CPUs and GPUs.
That dual exposure matters. If cloud providers increase CPU attach rates around AI accelerators, AMD could benefit through EPYC demand. If customers also adopt AMD Instinct accelerators, the company gains from both sides of the system.
The challenge is execution. AMD still competes with Nvidia in accelerators, Intel in CPUs and Arm-based custom silicon across hyperscale environments. To justify a higher valuation, AMD needs strong data-center revenue growth, margin discipline and continued customer adoption.
Intel Gets a Boost, but Bernstein Stays Neutral
Intel’s target increase to $100 from $65 is notable because it reflects renewed optimism about server CPU demand. However, Bernstein kept Intel at Neutral, which signals that the firm sees upside from the market expansion but remains cautious on execution.
That distinction is important. Intel has one of the deepest histories in server CPUs, and a CPU-heavy AI architecture could support its turnaround. Reuters recently reported that Intel forecast second-quarter revenue above expectations, helped by demand for server processors used in AI data centers. Intel’s Data Center and AI segment generated $5.1 billion in first-quarter revenue, ahead of analyst estimates.
Intel also has a potential manufacturing advantage if it can successfully execute its foundry roadmap. But the company still faces major challenges: competition from AMD, pressure from Arm-based designs, large capital-spending needs and the difficulty of regaining investor confidence after years of execution problems.
For Intel stock, Bernstein’s view is supportive but not a clean bull signal. The server CPU market may grow substantially, but Intel still has to prove it can capture profitable share.
Nvidia Is Watching the CPU Opportunity Too
The CPU renaissance is not only an AMD, Arm and Intel story. Nvidia is also moving more directly into CPUs.
Reuters reported that Nvidia has begun pitching its new Vera CPU to Chinese customers, describing it as a standalone CPU designed for agentic AI tasks. Reuters said the chip is expected to be available initially by August 2026 and could compete more directly with Intel and AMD in CPU workloads.
That shows why the CPU opportunity is becoming strategically important. Nvidia already dominates AI accelerators, but if the CPU layer becomes more valuable, Nvidia has an incentive to capture more of the total system architecture.
For investors, this reinforces the broader point: AI infrastructure is becoming a full-stack competition. The winners may not be determined only by who sells the fastest GPU, but by who can deliver the best combination of CPUs, GPUs, networking, memory, software and power efficiency.
What Investors Should Watch Next
The first key factor is whether agentic AI moves from hype to real enterprise deployment. If companies adopt autonomous agents at scale, CPU demand could rise meaningfully.
The second factor is the CPU-to-GPU ratio. Investors should watch whether cloud providers and AI infrastructure companies begin adding more CPUs per accelerator than in prior AI architectures.
The third factor is hyperscaler capex. Microsoft, Amazon, Alphabet, Meta, Oracle and other major cloud players will determine how much server CPU demand actually materializes.
The fourth factor is pricing power. A larger CPU market only helps shareholders if AMD, Arm and Intel can capture profitable revenue.
The fifth factor is competition from Nvidia and custom silicon. If Nvidia’s Vera CPU or hyperscaler in-house chips gain traction, the upside for traditional CPU suppliers may be shared more widely.
Bottom Line: Agentic AI Is Making CPUs Matter Again
Bernstein’s bullish call on Arm, AMD and Intel reflects a major shift in the AI investing narrative. The market is no longer focused only on GPUs and memory. CPUs are coming back into focus because agentic AI workloads require orchestration, tool use, planning, memory management and system-level coordination.
Arm may benefit from power-efficient architecture and licensing growth. AMD may benefit from combined CPU and GPU exposure. Intel may benefit from renewed server CPU demand, though execution risk remains higher. Nvidia, meanwhile, is already moving to capture part of the same opportunity.
The AI trade is getting broader and more complex. For investors, that creates opportunity—but also raises the bar. The next phase of AI infrastructure may reward companies that solve entire data-center bottlenecks, not just those that sell one category of chip.
FAQ
Why did Bernstein raise price targets on Arm, AMD and Intel?
Bernstein raised targets because it expects agentic AI to significantly increase server CPU demand. The firm lifted its 2030 server CPU market forecast to $223 billion, up from $137 billion.
What are Bernstein’s new price targets?
Bernstein raised AMD’s target to $600 from $525, Arm’s to $500 from $300, and Intel’s to $100 from $65. The firm kept Buy ratings on AMD and Arm and a Neutral rating on Intel.
Why does agentic AI need more CPUs?
Agentic AI systems handle multi-step workflows such as planning, retrieval, tool calls and memory coordination. Those tasks often rely heavily on CPUs, making server CPU demand more important than in simpler chatbot-style AI systems.
Which stock benefits most from the CPU renaissance?
Arm may benefit from power-efficient architecture, AMD from combined EPYC CPU and Instinct GPU exposure, and Intel from renewed server CPU demand. The best-positioned company depends on execution, valuation and customer adoption.
Is Nvidia affected by the CPU boom?
Yes. Nvidia is also moving into CPUs with its Vera processor, which Reuters reported is being pitched to Chinese customers for agentic AI workloads.
Sources: Seeking Alpha | Bernstein | Investors Business Daily | Reuters | Yahoo Finance | TrendForce | arXiv
Disclaimer: This article is for informational purposes only and does not constitute financial advice. Always conduct your own research before making any investment decisions.
