According to CoinMeta, AI renowned podcast host Dwarkesh Patel has made an unusual prediction that in the coming years, the rent for computing power may increase by more than 10 times. He pointed out that as models become smarter, the work that can be accomplished with the same amount of GPU becomes more valuable, and AI companies are more willing to pay high prices to secure computing power. Taking H100 as an example, if its computing power could continuously support a software engineer at human level, its annual rent could theoretically exceed $250,000, which is about 15 times the current spot price. There are already signs of price increases in the market; the one-year rent for H100 rose from $1.70 per hour in October 2025 to $2.35 per hour in March 2026, an increase of nearly 40%. Epoch AI estimates that global AI computing power will grow by about three times in 2025, but the demand for computing power from leading laboratories is growing even faster.