Foreign media outlets have compiled the views of several tech executives and investors, stating that while corporate AI budgets are rising rapidly, many companies are beginning to question the actual returns on these expenditures. Discussions surrounding large models, AI agents, and internal computing power consumption are shifting from "using as much as possible" to "whether it's worthwhile."
The first thing businesses feel is cost.
Uber COO Andrew Macdonald said he hasn't yet seen a clear correlation between AI spending and business improvements. He noted that it's difficult for companies to directly link a particular AI metric to "25% more useful features."
OpenAI CEO Sam Altman also said that the most common question he hears from enterprise clients is, "Spending is rising, and productivity seems to be rising too, but where are the increases in revenue and actual efficiency?" He believes that AI is still in a very new stage, and companies need time to explore ways to implement it.
Shift from "multi-use" to "accounting"
The article mentions that the early hype surrounding "tokenmaxxing" is cooling down. This refers to companies and employees using AI computing power and tokens as much as possible, but now more and more companies are starting to pay attention to where this money is actually going.
Mark Cuban believes the problem lies not in the token itself, but in the fact that companies haven't integrated new technologies in the long run. Shruti Gandhi bluntly states that many companies simply issue AI permissions without understanding how employees actually use them or whether they have truly changed the outcome.
- Jason Lemkin believes that more efficient companies will continue to invest more in AI.
- Less efficient and more traditional businesses may be more cautious after prices rise.
- Modal co-founder Akshat Bubna said that a significant portion of internal token spending may be worthless.
Model manufacturers begin to respond
Google CEO Sundar Pichai said he has also heard about budget pressures from several CIOs. He indicated that these concerns will continue to intensify this year.
Pichai also took the opportunity to promote Google's Gemini 3.5 Flash, claiming that this model is faster, cheaper, and suitable for real-world tasks. The article concludes by noting that while some industry insiders still believe AI spending will continue to expand, the market is beginning to ask more directly: what are the actual results of these investments?
