Survey shows enterprises GPU idleness and AI platform expansion plans coexist

AT A GLANCE

The survey says that 79%enterprises GPU are idle at least half the time, but 70%still plan to expand or replace their AI computing platforms.

Article

Survey: 79%enterprises GPU idle at least half the time, 70%still plan to expand AI platforms

Beating AI News Flash, a survey for the third quarter of 2026 compiled by VentureBeat Intelligence shows that 79%of enterprise respondents said their GPU computing capacity was idle at least half the time, yet 70%of surveyed enterprises plan to introduce, expand, or replace 12 within the next AI months.

Return on investment assessments are lagging as well. The survey shows that among the 64 respondents planning to adopt a new AI platform, 58 lacked finance-department-reviewed return on investment (ROI) data, accounting for 91%. In addition, while these enterprises’ own GPU utilization remains insufficient, they continue to use API to call proprietary models from companies such as OpenAI and Anthropic, resulting in both idle hardware resources and model-calling expenses.

As AI Agent becomes more widespread, the Token consumption and inference costs generated by multi-step tasks are attracting further attention. Gartner forecasts that by 2028, the inference cost of a single Agent workflow will rise by more than 5 times.

The report points out that enterprise AI competition is shifting from competing for computing capacity to improving resource utilization and controlling the cost of individual tasks. Measures include moving routine tasks such as classification and summarization to open models, optimizing context caching, and reducing unnecessary multi-Agent collaboration. However,GPU idleness may also result from enterprises reserving computing capacity for demand peaks or planning to replace older platforms; the survey data itself cannot directly prove that a AI bubble has formed.

Original link https://m.theblockbeats.info/flash/373741

Key points

A survey for the third quarter of 2026 compiled by VentureBeat Intelligence shows that 79%of enterprise respondents said their GPU computing capacity was idle at least half the time.
70%of surveyed enterprises plan to introduce, expand, or replace 12 within the next AI computing platforms.
Among 64 respondents planning to adopt a new AI platform, 58 lacked finance-department-reviewed ROI data, accounting for 91%。
These enterprises’ own GPU utilization remains insufficient, yet they continue to use API to call proprietary models from companies such as OpenAI and Anthropic.
Gartner forecasts that by 2028, the inference cost of a single Agent workflow will rise by more than 5 times.
Measures mentioned in the report include moving routine tasks such as classification and summarization to open models, optimizing context caching, and reducing unnecessary multi-Agent collaboration.
The report points out that GPU idleness may also result from enterprises reserving computing capacity for demand peaks or planning to replace older platforms.

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

In simple terms, some enterprises’GPU are not being fully utilized, but enterprises still plan to continue building or replacing their AI computing platforms. At the same time, many respondents planning to adopt new platforms still lack finance-reviewed return data. Enterprises may also simultaneously bear the cost of idle hardware and proprietary-model API call fees.

Why it matters to readers

This suggests that enterprise AI construction may be shifting from simply adding computing capacity toward improving GPU utilization and controlling the cost of model calls and Agent tasks.

GPU is the computing hardware used to run AI models; low utilization means the hardware is not performing computing tasks for substantial periods of time.ROI means return on investment, used to measure whether an investment generates sufficient returns.

Risks and unknowns

The original text is a professional media outlet’s paraphrase of the survey and does not provide the original VentureBeat Intelligence report, the complete sample size, or the respondent composition.
Survey data cannot directly prove that a AI bubble has formed.
GPU The specific reasons for being idle and their proportions are unclear.
Gartner's forecast is a judgment about the future; whether actual inference costs will increase and by how much remains unknown.

Survey shows enterprises GPU idleness and AI platform expansion plans coexist

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center