Foreign media reports that during his speech at Jackson Hole, Federal Reserve Chairman Kevin Warsh, in addition to discussing inflation, also mentioned a less common AI indicator: token prices. Here, token does not refer to crypto assets, but rather the unit of measurement used by AI models when processing data. Many AI companies charge customers based on their usage of token.
A downward price movement is not necessarily a good thing.
The article argues that Walsh does not regard the token price as a new official policy indicator, but rather as a window to observe the commercialization progress of AI. Price changes can reflect quality differences between models, the intensity of supplier competition, pricing strategies, as well as the underlying computing power costs.
token A price drop could correspond to two completely different situations. One is that the model's capabilities continue to improve, but at the same time, the cost of use also decreases, allowing companies to obtain more output for every dollar invested. The other situation is that models become increasingly similar, and suppliers can only compete for orders by lowering prices.
Companies should focus more on returns rather than unit prices.
The article quotes the statement of Gregory Daco, the chief economist of EY-Parthenon, stating that looking solely at token prices is not sufficient, as price signals themselves are not easy to interpret. For corporate financial officers, a more critical question is whether the adoption of AI has led to measurable improvements in efficiency.
The article states that companies also need to see whether AI has improved profit margins or whether it has created new sources of revenue. In other words, the price of token is merely a cost indicator on the input side. A lower price does not automatically mean a higher return on investment.
Pricing capability becomes a watershed in the industry
The article also quotes the view of technical analyst Luke Lango, stating that to determine whether AI truly creates value, the key lies not only in the growth of usage but also in whether the leading model can maintain its pricing power.
If the advanced models can still maintain a premium, it indicates that the market is willing to pay for the performance differences, and more of the industry's value will remain in the hands of the model providers. Conversely, if prices continue to move closer to marginal costs, the focus of competition will shift towards cost reduction and scale competition.
