AMD (AMD) investor relations material

Market opportunity and growth outlook

• AI market opportunity is rapidly expanding, with CPU TAM projected to exceed $200 billion by 2030 and expectations to capture over 50% market share in the next 3–4 years.

• Agentic AI workloads are driving significant CPU demand, with agentic applications expected to comprise about 50% of server TAM in the coming years.

• Accelerated business and multi-service compute are key growth areas, with large-scale deployments from major customers like Anthropic, OpenAI, and Meta.

• The company is confident in its ability to scale manufacturing and supply chain capacity to meet projected demand through 2028, with close coordination across the ecosystem.

• Demand projections are based on holistic analysis of customer workloads, adoption rates, and capital availability, with current market growth outpacing previous high expectations.

Product innovation and roadmap

• Venice and Florence CPUs are optimized for AI workloads, with Venice showing strong customer interest and broadening workload support.

• Helios systems will begin shipping in September, ramping through Q4 and into the first half of next year, aligning with major data center build-outs.

• MI450 and MI500 accelerators are central to large-scale AI deployments, with MI500 marking the start of a transition to optical networking.

• Chiplet architecture provides flexibility in memory bandwidth and capacity, supporting performance and TCO optimization for future products.

• ROCm.ai software platform is positioned as a major leap, enhancing developer productivity and platform accessibility, with significant improvements expected over the coming months.

Strategic partnerships and customer engagement

• Strategic deals with Anthropic, OpenAI, and Meta involve multi-gigawatt deployments, with Anthropic's MI450 engagement targeting up to 2 GW, mostly delivered in 2027.

• Collaborations focus on deep, multi-generational relationships, with ongoing discussions for future products like MI500 and MI600.

• Partnership with Cerebras is progressing well, with initial deployments expected by year-end and strong integration of disaggregated workflows.

• Claude and Codex models are being optimized for the platform, leveraging open source and abstraction strategies for rapid adoption.

• AI usage is ramping internally, accelerating development cycles and productivity across hardware and software teams.