The AI Silicon Shortage Explained: TSMC, Nvidia CPO, Memory Crisis & What Comes Next

The podcast analyzes the AI silicon shortage impacting the tech industry, highlighting wafer fab capacity constraints at TSMC and memory supply issues. Demand for AI accelerators is outpacing supply, squeezing consumer electronics. TSMC's role as "kingmaker" in allocating resources and prioritizing stable, high-volume customers like Apple and Nvidia is examined. The discussion covers the impact of memory constraints on smartphone production and the potential for double ordering to skew demand visibility. Co-packaged optics (CPO) and multi-source agreements (MSAs) are also discussed, with an analysis of Nvidia's evolving strategy and the dueling standards in the optical interconnect space. The panel also debates signals that might indicate a peak in AI demand.

Outlines

Part 1: The Silicon Shortage, Causes, and TSMC

Part 2: Market Impact and Supply Chain Constraints

Part 3: Hyperscaler Spending and GPU Economics

Part 4: Industry Cycles and Demand Monitoring

Part 5: Co-Packaged Optics (CPO) and Nvidia's Strategy

Part 6: Technical Standards and Future Connectivity

Sign in to continue reading, translating and more.

Open full episode in Podwise

SemiAnalysis Weekly

Part 1: The Silicon Shortage, Causes, and TSMC

The Great AI Silicon Shortage: Causes and Contributing Factors

TSMC's Capacity Constraints: AI Demand Outstripping Smartphone Demand

TSMC's Customer Selection and the Importance of Procurement Strategy

Part 2: Market Impact and Supply Chain Constraints

AI's Dominance of Entry Capacity and the Impact on Consumer Electronics

Limited Solutions to the Silicon Shortage and the Insufficient Smartphone Release Valve

HBM Shortages and the Structural Trend of Tight Memory Supply

Part 3: Hyperscaler Spending and GPU Economics

Hyperscalers' Continued Spending and Rising GPU Rental Prices

The Three-Tiered GPU Rental Market and the Economic Useful Life of GPUs

Long-Term GPU Contracts and the Bullish Case for AI Demand

Part 4: Industry Cycles and Demand Monitoring

Monitoring End Demand and the Cyclical Nature of the Semiconductor Industry

Double Ordering in the Semiconductor Industry and the Productivity Boost from Cloud Code

Part 5: Co-Packaged Optics (CPO) and Nvidia's Strategy

Introduction to Co-Packaged Optics (CPO) and its Importance

Nvidia's CPO Strategy: Scale-Out as the Vanguard

Investor Interest in CPO and the Wide Range of Total Adjustable Market (TAM)

Part 6: Technical Standards and Future Connectivity

Dueling MSAs: CPX vs. OCI and the Shift to DWDM

The CPX MSA and the Future of Die-to-Die Connections

The AI Silicon Shortage Explained: TSMC, Nvidia CPO, Memory Crisis & What Comes Next

SemiAnalysis Weekly

Part 1: The Silicon Shortage, Causes, and TSMC

00:00The Great AI Silicon Shortage: Causes and Contributing Factors

The Great AI Silicon Shortage: Causes and Contributing Factors

03:31TSMC's Capacity Constraints: AI Demand Outstripping Smartphone Demand

TSMC's Capacity Constraints: AI Demand Outstripping Smartphone Demand

09:09TSMC's Customer Selection and the Importance of Procurement Strategy

TSMC's Customer Selection and the Importance of Procurement Strategy

Part 2: Market Impact and Supply Chain Constraints

11:40AI's Dominance of Entry Capacity and the Impact on Consumer Electronics

AI's Dominance of Entry Capacity and the Impact on Consumer Electronics

13:34Limited Solutions to the Silicon Shortage and the Insufficient Smartphone Release Valve

Limited Solutions to the Silicon Shortage and the Insufficient Smartphone Release Valve

16:58HBM Shortages and the Structural Trend of Tight Memory Supply

HBM Shortages and the Structural Trend of Tight Memory Supply

Part 3: Hyperscaler Spending and GPU Economics

18:38Hyperscalers' Continued Spending and Rising GPU Rental Prices

Hyperscalers' Continued Spending and Rising GPU Rental Prices

23:38The Three-Tiered GPU Rental Market and the Economic Useful Life of GPUs

The Three-Tiered GPU Rental Market and the Economic Useful Life of GPUs

27:05Long-Term GPU Contracts and the Bullish Case for AI Demand

Long-Term GPU Contracts and the Bullish Case for AI Demand

Part 4: Industry Cycles and Demand Monitoring

31:34Monitoring End Demand and the Cyclical Nature of the Semiconductor Industry

Monitoring End Demand and the Cyclical Nature of the Semiconductor Industry

33:34Double Ordering in the Semiconductor Industry and the Productivity Boost from Cloud Code

Double Ordering in the Semiconductor Industry and the Productivity Boost from Cloud Code

Part 5: Co-Packaged Optics (CPO) and Nvidia's Strategy

39:14Introduction to Co-Packaged Optics (CPO) and its Importance

Introduction to Co-Packaged Optics (CPO) and its Importance

43:05Nvidia's CPO Strategy: Scale-Out as the Vanguard

Nvidia's CPO Strategy: Scale-Out as the Vanguard

45:57Investor Interest in CPO and the Wide Range of Total Adjustable Market (TAM)

Investor Interest in CPO and the Wide Range of Total Adjustable Market (TAM)

Part 6: Technical Standards and Future Connectivity

49:33Dueling MSAs: CPX vs. OCI and the Shift to DWDM

Dueling MSAs: CPX vs. OCI and the Shift to DWDM

53:07The CPX MSA and the Future of Die-to-Die Connections

The CPX MSA and the Future of Die-to-Die Connections