How to Build a Thriving AI Ecosystem with Lisa Su, CEO of AMD

Executive Summary
Lisa Su, Chair and CEO of AMD, discusses the strategic evolution of high-performance computing (HPC) and the critical role of open ecosystems in the generative AI revolution. The conversation highlights AMD's transition from a low-share server player to a primary AI infrastructure provider through a focus on open standards, strategic acquisitions (Xilinx, ZT Systems), and deep partnerships with hyperscalers and foundries.

Strategic Vision and Market Positioning

• Lisa Su characterizes generative AI as the most significant technology discontinuity of the last 30–50 years, fundamentally shifting compute from narrow PC/mobile applications to ubiquitous infrastructure.
• AMD's core strategy relies on "high-performance computing" as the foundation for AI, specifically leveraging the ability to train models with hundreds of billions to trillions of parameters.
• Su predicts a future of heterogeneous AI engines, where different workloads (edge, embedded, automotive, client) require specialized compute architectures rather than a single "highest performance" chip for all tasks.
• The company has moved from a "what should we be when we grow up" identity crisis a decade ago to a focused portfolio on complex microprocessors, GPUs, and adaptive computing.
• AMD operates as a fabless design company, relying on deep partnerships with foundries like TSMC for advanced node manufacturing and advanced packaging.

Ecosystem, Standards, and Openness

• Su advocates strongly for open ecosystems and interoperability, arguing that closed walls historically hinder rapid technological evolution and innovation.
• AMD co-leads the Ultra Accelerator Link (UALink) and Ultra Ethernet Consortium to establish open industry standards for networking fabrics, allowing competitors to collaborate on infrastructure while differentiating on individual hardware.
• The company believes software accessibility is as critical as silicon; developers must be able to build across a wide range of hardware rather than being locked into a single vendor's proprietary ecosystem.
• AMD's "open" stance extends to customer relationships, where they prioritize long-term roadmaps and consistent execution with hyperscalers like Microsoft, Amazon, Google, Oracle, and Meta.

Supply Chain, Manufacturing, and Government Policy

• The 2020-2021 COVID-19 pandemic created an unprecedented simultaneous demand spike across all semiconductor markets, breaking traditional cyclical hedging strategies.
• The current supply constraint in AI is driven by a lack of advanced packaging capacity and high-bandwidth memory (HBM), areas that were not foreseen by the industry.
• The CHIPS Act (authorizing $280 billion total, with $53 billion spent) is viewed by Su as a critical industrial policy shift that prioritizes U.S. manufacturing and R&D resiliency.
• Su notes that the industry is shifting focus from "efficiency" (extracting every penny) to "resiliency" to build supply chains capable of withstanding future disruptions.
• AMD utilizes a long development cycle approach (3–5 years) for major bets like chiplet architecture and advanced packaging, requiring a "crystal ball" capability to predict technology arcs despite uncertainty.

Acquisitions and Portfolio Expansion

• The acquisition of Xilinx is identified as the largest semiconductor acquisition in history, bringing FPGA and adaptive computing capabilities into the AMD portfolio.
• The acquisition of ZT Systems represents a strategic move into systems integration, adding the "third leg of the stool" (hardware, software, systems) to address large-scale HPC cluster challenges.
• ZT Systems enables AMD to provide turnkey solutions that address complex networking, thermal management, and reliability issues inherent in building massive AI clusters.
• AMD has executed approximately six acquisitions in the last five to six years to round out its product portfolio and capabilities.

Customer Engagement and Startup Relations

• AMD's server market share was less than 1% prior to the Zen architecture launch, requiring a strategy focused on deep customer partnerships and sustained innovation over generations.
• The company is actively engaging with startups to leverage their speed and experimentation, noting that even large enterprises are becoming more open to collaborating with new entrants to avoid falling behind in AI.
• Su offers a direct call for startups needing GPUs to contact AMD, citing the company's capacity to support emerging innovators.
• Barriers to entry for semiconductor startups have lowered due to the availability of world-class foundry capabilities and hyperscaler infrastructure.