AMD (AMD) Morgan Stanley Technology, Media & Telecom Conference 2026 summary

Market outlook and growth strategy

• Strong momentum in high-performance compute and data center business, with 2026 shaping up as an exciting year driven by product launches and deep customer relationships.

• Ambitious financial model targets 35% CAGR over the next 3–5 years and over $20 EPS in that timeframe.

• AI infrastructure investments are equated with productivity and intelligence, with enterprise use cases and demand still in early stages of deployment.

• CapEx discussions and planned builds extend into 2026–2027, reflecting durable and growing enterprise demand.

• Compute market growth is outpacing previous forecasts, with both CPU and GPU demand exceeding expectations.

Strategic partnerships and major deals

• Announced a 6 GW long-term strategic partnership with Meta, including a semi-custom GPU tailored to Meta’s workloads, deepening the relationship and increasing ecosystem footprint.

• Warrants issued to Meta are performance-based, designed for transformational partnerships, and incentivize mutual success.

• Similar multi-generational, gigawatt-scale partnership with OpenAI, with technical alignment and co-validation of installations.

• No plans for similar warrant structures with other customers; Meta and OpenAI relationships are unique.

• Multiple other hyperscalers and AI labs are interested in gigawatt-scale deals, with ambitions to accelerate data center AI business at over 80% CAGR in the next 3–5 years.

Product and technology roadmap

• MI450 series launch is a major step in capability, with strong customer interest and rack-scale infrastructure investments, including the ZT Systems acquisition and Helios rack developed with Meta.

• MI450 and MI455 are positioned as leadership products for both training and inference, leveraging chiplet architecture for workload optimization.

• Rack-scale systems are validated and on track for a smooth ramp, with Q3 revenue and a sharp volume ramp in Q4.

• Networking roadmap includes UALink and Ethernet support, with a focus on open standards and customer choice.

• Venice CPU, built on TSMC 2nm and chiplet architecture, is set for a strong ramp in the second half of the year, with broad customer demand.