Dead Money (www.wheresyoured.at)

The recent analysis by Fidelity's Jurien Timmer reveals troubling trends in the AI industry's capital expenditures, describing substantial amounts of investment in GPU hardware as "dead money." Timmer notes that both token spending and GPU lease rates have stagnated, indicating a disconnection between the sales of AI hardware and the actual demand for AI services. Morgan Stanley's findings suggest that a significant portion of GPUs sold may lack the necessary infrastructure for deployment, creating doubts about the legitimacy of the purported AI boom. This raises serious concerns about the viability of investments from major hyperscalers like Microsoft, Amazon, and Google, which collectively require up to $3 trillion in annual AI revenues by 2030 to justify their expansive capital expenditures. The implications for the AI/ML community are profound. The current narrative—where hyperscalers stockpile GPUs in anticipation of future demand—may be misleading if the capital influx is not anchored in grounded demand across a diverse customer base. Instead, a disproportionate share of revenue is funneled back to a few key players like OpenAI and Anthropic, suggesting that the AI market's growth may be more speculative than sustainable. As venture capital remains pivotal for supporting companies' AI compute needs, the industry's future hinges on addressing potential overcapacity and reevaluating investment strategies to develop a robust framework for real, lasting demand.