I'm LongbridgeAI, I can summarize articles.A Jefferies report points out that as the performance of small language models improves and costs decline, enterprise AI deployment may shift from the cloud to on-premises solutions, putting pressure on cloud computing demand. Data from Swerve Research shows that Anthropic's server utilization has dropped 26% from its peak, suggesting potential idling of supply. Although tech giants are projected to spend nearly $1 trillion on capital expenditures by 2027, if demand growth fails to match the scale of investment, some data centers may face the risk of insufficient returns
AI applications continue to expand, but the hyperscale data center investments underpinning this boom are facing new challenges on the demand side.
A Jefferies report notes that as small language models improve in performance and decrease in cost, enterprise AI deployments may shift from centralized cloud environments to local setups, thereby exerting pressure on cloud computing demand and massive data center capital expenditures.
A key basis for this judgment is recent observations of actual AI server usage. In September 2026, Swerve Research Technologies constructed a "congestion index" by tracking wait times in Anthropic's server request queues, finding that server utilization had fallen 26% from its peak in January–February 2026. While this metric does not directly correlate with company revenue, sustained low utilization could indicate that some computing supply is beginning to sit idle.
Meanwhile, investment in AI infrastructure continues to expand rapidly. The market currently expects that the combined capital expenditures of four U.S. tech giants—Meta, Alphabet, Amazon, and Microsoft—could reach $990 billion in 2027. The report argues that if the growth rate of AI computing demand ultimately fails to match such a vast scale of investment, some data centers may face the risk of inadequate returns or even become "stranded assets."
Declining Server Utilization Tests AI Computing Supply
Meta's personal AI product, Muse, launched on September 8, driving Meta's stock price up 21% within weeks and becoming a highlight of recent AI consumer applications. However, the report suggests that the success of a single application is insufficient to prove the sustainability of the entire AI infrastructure investment cycle.
More notably, AI companies themselves are sending signals that differ from market optimism. Anthropic's CEO, OpenAI CEO Sam Altman, and Elon Musk have recently publicly discussed the need to slow down AI development; Anthropic's planned IPO has been postponed until November.
At the same time, Anthropic previously signed a three-year, $45 billion computing contract with SpaceX and reached multi-billion-dollar computing partnerships with Amazon and Google Cloud. If AI revenue growth slows while large-scale computing contracts signed earlier still need to be fulfilled, corporate cost pressures may rise further.
Reportedly, Meta was also in negotiations with Anthropic in July to lease up to $10 billion worth of data center computing power over two years. The report believes that large tech companies actively seeking external customers to absorb computing capacity reflects a new challenge: how to improve asset utilization as infrastructure supply expands rapidly.
Small Model Performance Catches Up, Local Deployment May Divert Cloud Demand
Compared to short-term fluctuations in server utilization, the report pays more attention to changes occurring in AI model architectures.
A study titled "Intelligence per Watt," jointly released by Stanford University and Together AI, shows that the performance of small language models is rapidly catching up with that of large language models, while energy consumption and computing costs are 50% to 85% lower. If this trend continues, the scale of computing power and cost structure required for enterprise AI deployment could change accordingly.
In the past, enterprises relied more on centralized cloud-based large models to complete AI tasks; as small models improve, some scenarios may shift to local deployment. For industries with high data security requirements, such as banking, there is already strong demand for localized deployment; as small models further reduce deployment costs, the applicability of this model may continue to expand.
This means that growth in AI applications does not necessarily correspond to proportional growth in cloud computing demand. Enterprises may expand AI applications while reducing their reliance on ultra-large model APIs and centralized cloud computing.
This is also a major reason why the report questions the nearly trillion-dollar capital expenditure: if more AI applications are borne by low-cost, small-scale, localized models in the future, the current investment scale centered around hyperscale data centers may mismatch actual computing demand.
In this scenario, what truly needs reevaluation is not whether AI demand exists, but rather how much centralized computing power AI demand actually requires, and whether these data centers can achieve returns commensurate with their investment scale.
Risk Warning and Disclaimer
The market carries risks; invest with caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial status, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment decisions made based on this content are at the user's own risk.