AI bubble risk grows as power constraints threaten GPU boom
AI bubble risk intensifies as Google’s Gemini 3 launch spurs record GPU demand, overturning the “scaling wall” thesis and driving investors to pour capital into AI infrastructure. Nvidia reports off-the-charts Blackwell sales and continuous utilization of older A100 GPUs, reflecting a cascading use model of training, inference and legacy workloads. However, data centers, which power this boom, face a power supply bottleneck: gas turbine manufacturers are booked until 2030, stalling new capacity. Without increased energy supply, investing in power-hungry next-gen GPUs or data centers has diminishing returns, raising the specter of an AI bubble burst. Data centers now contribute 4% of US GDP but account for 93% of growth. Capital expenditure on AI rivals past bubbles, with Microsoft’s capex near 50% of sales and startups almost entirely AI-focused. Traders should monitor energy constraints as a key indicator of AI bubble risk.
Neutral
While escalating AI infrastructure investments and potential power shortages could indirectly influence GPU availability, the direct effect on cryptocurrency markets is limited. Crypto mining has largely shifted to ASIC hardware for major coins like Bitcoin and Ethereum’s move to proof-of-stake reduces GPU demand. Historical instances, such as GPU shortages during crypto bull runs, had only temporary effects on mining profitability and GPU prices, without significantly altering market trends. Consequently, traders should view this AI-focused bubble risk as neutral for crypto assets, noting that energy constraints may affect mining costs marginally but are unlikely to drive major market movements in either direction.