Lisa Su's AI Infrastructure Vision: 3 Stocks for the Next Phase

Lisa Su's AI Infrastructure Vision

• 💡 Lisa Su argues that AI did not peak in 2025 but rather proved its concept, with early return on investment appearing in various businesses.
• 🚀 The next phase of AI is a massive infrastructure cycle, moving beyond just software to address a global compute deficit.
• 📈 As AI users scale from 1 billion to 5 billion, the world needs a 100-fold increase in compute over the next 4-5 years.
• 🛠️ This infrastructure buildout requires significant investment in silicon, memory, networking, power, advanced packaging, and full data center systems.

AMD: Powering AI Compute

• ✅ Advanced Micro Devices (AMD) is positioned at the core of AI compute, transitioning from a CPU story to a full-stack AI platform.
• 📊 AMD's AI accelerator revenue exceeded $5 billion, with significant data center growth and planned deployment of 50,000 GPUs at Oracle by 2026.
• 💻 The company is also driving AI PCs with its Ryzen AI platform, enabling on-device neural processing and local inference.
• 🚗 AMD's embedded segment is gaining traction in physical and automotive AI, powering digital cockpits and real-time systems.

Broadcom: Connecting AI Infrastructure

• 🔗 Broadcom (AVGO) serves as the critical connective tissue for AI at scale, dominating networking, interconnects, and custom silicon.
• 💰 The company reported a projected doubling of AI semiconductor revenue and a substantial $73 billion AI backlog from hyperscalers.
• 💡 Broadcom co-develops custom ASICs and XPUs with major customers like Google and Meta, optimizing for lower power consumption and cost per token.
• ⚡ Its Tomahawk 6 switch addresses AI bottlenecks by efficiently moving data between accelerators, crucial for scaling AI clusters.

TSMC: Manufacturing AI's Foundation

• 🏭 Taiwan Semiconductor Manufacturing Company (TSMC) is the ultimate "picks and shovels" play, manufacturing nearly every serious AI chip.
• 📈 TSMC's advanced nodes (7nm and below) represent three-quarters of wafer revenue, with AI and high-performance computing accounting for over half of total sales.
• 🔬 The upcoming 2nm manufacturing transition in late 2025 will bring higher wafer pricing and further lock in customers due to high switching costs.
• 📦 Its CoWoS advanced packaging technology is essential for AI accelerators and high-bandwidth memory, generating significant revenue streams.

The Future of AI Buildout

• 🎯 The period of 2026-2027 is crucial as AI moves from experimentation into everyday business operations across various sectors.
• 🌐 This shift will lead to productivity gains showing up in earnings, efficiency metrics, and eventually GDP growth.
• 🤝 Together, AMD, Broadcom, and TSMC form the core backbone of a decade-long AI infrastructure buildout, redefining economic growth.