Is the AI spending bubble starting to shrink? Some large cloud customers are turning to more affordable model options

A game around AI pricing and market share is unfolding.

As cheaper AI models emerge, enterprise customers are reassessing their AI spending. Palo Alto Networks initially used cutting-edge models like OpenAI's, but later found that the open-source model from DeepSeek could accomplish the same tasks at only 5% of the cost. Meanwhile, the proliferation and price decline of NVIDIA AI chips have also accelerated this trend.

Although there are signs of a slowdown in AI spending in the short term, the long-term decline in costs may stimulate growth in AI demand, aligning with the "Jevons Paradox" in economics. Industry giants like Microsoft firmly believe that lower costs will expand the overall market for AI models.

As AI technology matures and price competition intensifies, AI investments are facing a more rational market test, and investors need to pay attention to the potential impact of this trend on technology stock valuations.

The Rise of Low-Cost AI Models: A Shift in Enterprise Spending Strategies

Palo Alto Networks began incorporating AI chatbots into its products last year to help customers obtain answers regarding their cybersecurity. Initially, these tools relied on cutting-edge models from OpenAI and other companies.

However, CEO Nikesh Arora recently discovered that the open-source model from Chinese developer DeepSeek can accomplish the same tasks at 5% of the cost of OpenAI's model. The company's CEO Nikesh Arora stated:

"For tasks where we are gaining efficiency and reducing costs, I don't think the IQ of the model needs to be higher than it is now... I don't want to pay a dollar for that; I'd rather pay five cents."

This shift is not an isolated case. Many companies that pay for AI software for customer support, software programming, or analytics are gaining greater benefits due to the release of low-cost AI models from DeepSeek and OpenAI. Additionally, NVIDIA server chips used by cloud customers to run AI have become easier to obtain at lower prices over the past year.

The Information analysis suggests that these developments have generally lowered the costs of AI services. This means that the current slowdown in AI spending among large cloud customers may only be temporary.

Jevons Paradox and Market Dynamics: Long-Term Growth and Short-Term Adjustments

Some companies are slowing their AI spending by turning to cheaper models. For example, TurboTax manufacturer Intuit steadily increased its spending on models from OpenAI and other providers in 2023 and 2024, exceeding millions of dollars per month by last summer. However, Intuit has since shifted to free open-source models and recently released cheaper models from OpenAI, thereby slowing Intuit's growth in AI spending on Azure.

Microsoft and AWS have rebutted the claim that declining AI model costs will hinder their AI businesses. Microsoft CEO Satya Nadella insists that lower costs will lead to an overall increase in the purchase volume of AI models, which is part of the "Jevons Paradox" in economics Nadella said:

"Whenever there is a breakthrough, like what DeepSeek has done... it will bring more demand. That's how the cloud happens."

The Information believes that the decline in AI costs may force software giants like Microsoft and Salesforce to lower the prices of their new AI products (such as Microsoft's Copilot and Salesforce's Agentforce), or they will face challenges from lower-cost competitors.

The price drop may have already impacted cloud service companies in other ways. According to reports citing two people with direct knowledge of the situation, Microsoft has recently reduced the sales quotas for several Azure sales teams by as much as 40%, although it is unclear whether the slowdown in AI customer spending prompted this change.

As AI technology matures and price competition intensifies, the once-rapid investment in AI is facing a more rational market test, and investors need to closely monitor the potential impact of this trend on technology stock valuations