The dirty secret behind Big Tech’s AI arms race: Massive hardware investments that are obsolete in 3 years

A new report from investment firm Research Affiliates, which manages approximately $200 billion in assets, exposes a critical paradox in the AI boom. The analysis, led by CEO Chris Brightman, argues that the massive hardware investments by tech giants (hyperscalers) are not traditional capital expenditures but a form of rapid inventory churn. The specialized GPUs and infrastructure powering today's AI models have an effective lifespan of just three years before becoming obsolete, fundamentally altering the economics of the industry.

This creates a new industrial paradigm where companies like Microsoft, Google, and Amazon are engaged in a perpetual arms race. They must constantly 'restock their shelves' with the latest hardware to maintain competitive AI products like large language models (LLMs) and vector search databases. The report contends this transforms these tech behemoths, making their financial models resemble high-turnover retailers more than classic technology or industrial enterprises, with billions in spending locked into a cycle of near-constant hardware replacement.