Semiconductors

The Semiconductor Industry Is Being Reshaped by AI Demand

The global semiconductor market is projected to surpass $1 trillion by 2030, with AI-related chips representing the fastest-growing segment. The insatiable demand for AI training and inference compute has created the largest capital expenditure cycle in semiconductor history. Nvidia alone generated over $130 billion in data center revenue in 2025, yet demand continues to outstrip supply.

The competitive landscape is shifting. While Nvidia dominates GPU-based AI compute today, a new generation of purpose-built AI chips is emerging. Application-specific integrated circuits (ASICs) designed exclusively for transformer inference workloads promise dramatically better performance-per-watt and cost-per-inference than general-purpose GPUs. Companies like Etched are challenging the GPU paradigm with silicon that trades generality for orders-of-magnitude efficiency gains on the workloads that matter most.

The supply chain is equally dynamic. Geopolitical tensions have accelerated semiconductor reshoring efforts, with the U.S. CHIPS Act deploying $52 billion to rebuild domestic manufacturing capacity. Simultaneously, advanced packaging, memory architecture, and interconnect technologies are becoming as important as the logic chips themselves in determining system performance.

Uhlig Capital's semiconductor exposure centers on Etched ($500M raise at $5B valuation), accessed through our investment in Align Ventures' Fund I. Etched's approach - building ASICs optimized exclusively for transformer inference - represents a contrarian bet that specialization will outperform generalization in the most valuable AI compute workloads.