Survey: 79% of Enterprises’ GPUs Are Half-Idle
A VentureBeat Intelligence survey found that 79% of 218 enterprises running AI in production used their GPUs at 50% capacity or less between July and September. In the September survey, the figure was 81%, while 31% of respondents said their GPUs were doing useful work no more than a quarter of the time.
Despite low GPU utilization, 70% of respondents reporting underused hardware said they plan to adopt a new, additional or replacement AI compute platform within 12 months. In the September wave, 63% of 102 qualified respondents planned a platform change or addition. Only 10% had a finance-reviewed return on investment for their existing infrastructure, and 12% reported using specialized AI cloud providers such as CoreWeave or Lambda.
The findings point to a mismatch between GPU utilization and continued AI infrastructure investment. Some planned changes may replace unsuitable systems rather than add capacity. The trend could create opportunities for providers offering better workload optimization or alternative compute models. Future GPU utilization and spending data will show whether enterprises can reduce idle capacity and demonstrate clearer returns.
Neutral
The survey describes enterprise AI compute usage and purchasing plans, not cryptocurrency prices, blockchain activity or token demand. Its direct impact on crypto markets is therefore limited, supporting a neutral classification. In the short term, the findings could influence sentiment around AI infrastructure companies and GPU cloud providers, but they offer no clear trading signal for major crypto assets. Traders may watch related technology stocks, AI-sector sentiment and broader risk appetite for indirect spillovers.
Over the longer term, persistent GPU underuse alongside planned platform changes could shift investment toward more efficient compute providers or workload optimization. That may affect companies supplying AI infrastructure, including firms with ties to crypto mining, but the survey does not establish a direct link to token prices. As with past AI-related headlines, any crypto-market reaction would likely depend more on broader technology sentiment and macroeconomic conditions than on this data alone. No specific change in crypto market stability or trading activity can be inferred from the survey.