Goldman Sachs Sounds Alarm for AI Investors Amid Falling Token Prices

Goldman Sachs has sounded an alarm for investors in the AI sector, warning that falling prices for artificial intelligence tokens could lead to an infrastructure risk. The bank's Delta One desk highlighted a recent decline of 29% in token prices to new lows, with the industry benchmark from Silicon Data falling to $0.97 per million tokens.

This drop in prices may seem like a boon for customers and could accelerate adoption, but Goldman warns that it also reduces revenue generated from each unit of AI activity. The bank fears that if prices continue to fall faster than overall usage grows, demand for expensive data-center capacity may eventually disappoint.

Goldman Sachs pointed out that the industry is still spending heavily on infrastructure, with Microsoft's capital expenditures reaching $41 billion in its latest quarter and Alphabet expecting roughly $175 billion of 2026 investment. The bank emphasized that periods of compute oversupply become plausible unless future models produce major improvements that stimulate substantially more consumption.

The critical question for investors is whether token consumption grows faster than token prices decline, making AI utilization, cloud revenue growth, and data-center returns increasingly important metrics to watch.