Bank of America Upgrades Intel on Agentic AI CPU Opportunity

What happened

Bank of America analyst Vivek Arya issued a rare upgrade of Intel from underperform to buy, raising the price target to $135 from $96. The firm also increased its forecast for the server CPU market in 2030 to nearly $170 billion from $125 billion, representing a 37% compound annual growth rate. The upgrade is based on the emergence of agentic AI-systems that plan, reason, retrieve information, and execute multi-step tasks-which requires more CPU-intensive workloads for orchestration, memory management, API calls, and tool execution. AMD retained its buy rating with a price target increase to $560 from $500, and remains Bank of America's top CPU pick due to its market position and upcoming EPYC Venice CPU with up to 256 cores. Nvidia remains the firm's top overall AI sector pick.

Why it matters

For AMD investors, this analysis validates the company's CPU franchise in the evolving AI infrastructure landscape. While the first wave of AI spending centered on GPUs for model training, the shift toward agentic AI systems creates parallel demand for high-performance CPUs that handle sequential, latency-sensitive tasks. AMD's competitive position is reinforced by its upcoming EPYC Venice with 256 cores-significantly more than Intel's anticipated Diamond Rapids at 192 cores or Nvidia's Vera at 88 cores. This core-count advantage positions AMD to capture share in workloads requiring simultaneous management of multiple subtasks. The firm's raised $560 price target and reaffirmed buy rating reflect confidence that AMD can capitalize on both GPU and CPU demand growth, leveraging its incumbency, product pipeline, and performance leadership.

Bigger picture

The AI hardware market is maturing beyond the GPU-centric buildout that initially defined the sector. Bank of America's thesis suggests that agentic AI will expand the addressable market rather than substitute GPUs for CPUs-data centers may need CPU-heavy systems to sit alongside GPU racks for orchestration, retrieval-augmented generation, vector databases, and memory-intensive workflows. The firm forecasts the 2030 server CPU market will include roughly $30 billion in traditional infrastructure, $70 billion in AI computing and head node CPUs, and another $70 billion in AI agentic standalone node CPUs. This diversification benefits multiple players: Nvidia retains leadership due to its full-stack AI infrastructure approach, Arm is expected to reach nearly 50% server CPU value share by 2030, and Intel gains a potential comeback narrative after years of manufacturing delays and market-share losses. For AMD, the expansion validates its dual-pronged strategy of competing in both accelerators (Instinct GPUs) and CPUs (EPYC), positioning the company to capture value across the AI infrastructure stack.

What to watch

Key signals include AMD's EPYC Venice CPU launch and adoption metrics, particularly in agentic AI workloads where core count and memory handling are critical. Investor attention will focus on AMD's Data Center segment growth, already up 57% year-over-year to $5.78 billion in Q1 2026, and whether multi-gigawatt GPU commitments from Meta and OpenAI translate into sustained CPU demand. Competitive dynamics will also matter: Intel's execution on manufacturing improvements, foundry business traction, and upcoming server chips like Diamond Rapids, as well as Arm-based CPU share gains from Nvidia, Amazon, Google, and Microsoft. Broader indicators include data center capital expenditure trends, the pace of agentic AI adoption by enterprises, and any constraints in leading-edge wafers and advanced packaging that could limit supply. AMD's partnership with Samsung on HBM4 supply for MI455X GPUs and TSMC manufacturing capacity will also be important for meeting accelerator demand alongside CPU growth.