The AI era cannot escape Jevons' paradox either

A friend is curious, why did the price of each million tokens plummet from $60 to $0.3, while global AI spending skyrocketed more than threefold?
Just last week, DeepSeek announced a permanent price reduction for V4-Pro, with discounts up to 99%; Xiaomi MiMo quickly followed suit, increasing token package usage by 5-8 times.
Almost simultaneously, Alibaba Cloud and Tencent Cloud quietly raised their AI computing power and service prices, with some increases exceeding 400%.
The cheaper the tokens, the more expensive enterprise AI bills become; the story of the steam engine is replaying on tokens.
In 1865, British economist William Stanley Jevons discovered a counterintuitive phenomenon: after Watt improved the steam engine, its thermal efficiency greatly increased, which should have reduced coal consumption, but instead, coal consumption across the UK exploded.
Jevons' explanation was very simple: when the efficiency of resource use improves and costs decrease, it will be used in more scenarios that were previously "unaffordable."
Total demand growth far exceeds the savings brought by efficiency improvements.
Today, tokens are perfectly replicating this story.
When each million tokens still sells for $60, AI can only serve high-end scenarios like financial quantification and new drug development that can "afford to spend."
When prices drop to a few cents or even a fraction of a yuan, a large number of scenarios that previously couldn't afford AI—such as customer service systems, programming assistance, personalized education, and mass content production—are fully exploding.
In the future, the price signals of tokens will continue to split: foundational models will keep competing on price, cloud computing power will keep rising in price, and top-tier intelligence will continue to command a premium.
But regardless of how surface prices fluctuate, the overall expenditure will keep rising, and the long-term demand for computing power will remain in short supply—this trend was predicted by Jevons 160 years ago.