The AI compute cycle is not a bubble — it is the largest infrastructure buildout in human history. We examine the structural drivers, key beneficiaries and risk factors across the AI infrastructure stack.
The global AI infrastructure buildout represents a structural shift comparable to the electrification of the 20th century. Unlike the dot-com era, today's AI investment cycle is driven by measurable productivity gains, clear revenue models and a small number of dominant platform companies with the financial capacity to sustain multi-year capital expenditure programmes.
NVIDIA's data centre revenue has grown from $3B in 2022 to over $90B in 2025, reflecting genuine demand from hyperscalers, sovereign AI programmes and enterprise AI adoption. The Blackwell architecture represents a 30x performance improvement over Hopper for AI inference workloads — a step-change that is accelerating, not decelerating, adoption.
The semiconductor supply chain underpinning AI infrastructure is highly concentrated. TSMC manufactures over 90% of the world's most advanced AI chips. ASML supplies 100% of the EUV lithography systems required to manufacture them. This concentration creates both extraordinary moat quality and significant geopolitical risk.
Our base case projects AI infrastructure capex to exceed $1 trillion annually by 2028, with NVIDIA, TSMC, ASML and Broadcom capturing the majority of incremental value. The key risk to this thesis is not competition — it is a sudden reversal in hyperscaler AI return-on-investment expectations.
Research Disclaimer: This article is intended solely for informational, educational and research purposes. Nothing in this article constitutes investment advice or a recommendation to buy or sell any security. BGAM™ scores, scenarios and classifications are model-based analytical assessments for illustrative purposes only.
MORE IN AI INFRASTRUCTURE
Cookie ConsentWe use cookies to understand how visitors use our site. No personal data is shared with third parties. Privacy Policy