AMD (AMD) investor relations material

AI and semiconductor market trends

• Agentic AI adoption is accelerating, driving significant demand for high-performance CPUs and GPUs across data center, client, and gaming applications.

• CPU business grew over 50% in Q1, with Q2 guidance for 70% year-over-year growth, fueled by investments in adaptive compute and end-to-end solutions.

• The total addressable market (TAM) for CPUs is now forecasted to exceed $120 billion by 2030, with agentic AI expected to comprise over 50% of this market.

• Agentic AI workloads require high core count and high-performance CPUs, leading to increasing average selling prices (ASPs) over time.

• The shift from AI training to inference is transforming CapEx allocation, with agentic inference workloads driving diverse CPU demand.

Market segmentation and competitive positioning

• The CPU market is segmented into traditional compute, head node (CPU-GPU communication), and agentic AI, with agentic AI as the fastest-growing segment.

• Traditional CPU growth remains steady, while head node and agentic AI segments are expanding rapidly, with agentic AI expected to dominate future market share.

• x86 architecture is favored for traditional workloads due to ecosystem compatibility, while agentic AI racks prioritize high thread and core count products.

• Security and reliability features in x86 platforms are critical for mission-critical server deployments, supporting strong competitive positioning.

• Value share in CPUs reached 46% in Q1, with continued collaboration with TSMC ensuring supply for future growth.

Supply chain, pricing, and operational planning

• Supply remains tight, especially for advanced nodes, but strong relationships with TSMC and early planning have secured necessary capacity through 2027.

• Two-thirds of CPU revenue growth is attributed to unit expansion, with ASPs expected to rise as agentic AI adoption increases.

• CPU ASPs are influenced by configuration mix, with higher core counts and performance driving higher prices in agentic AI workloads.

• GPU product launches, such as MI450, are on track, with major customers like OpenAI and Meta exceeding initial forecasts for 2027.

• Long-term agreements and strategic planning with suppliers are essential to manage memory cost inflation and ensure supply chain resilience.