Enterprise buyers are increasingly adding non-Nvidia AI accelerators to their evaluation lists, with 39.4% of respondents in a July survey saying they are likely to assess chips from AWS, Google, AMD, Intel, or in-house designs over the next 12 months. That compares with 25.3% who plan to evaluate Nvidia's next-generation GPUs, such as the Blackwell GB300, a 14-point gap that signals a shift toward greater optionality in AI infrastructure strategies.
The findings come from VentureBeat's July VB Pulse survey of 170 AI infrastructure respondents. While Nvidia remains the default choice in most production environments, organisations are building real alternatives into their planning rather than treating Nvidia as the only option worth considering. The survey also shows that enterprises are focusing on optimising their existing AI infrastructure before making major platform changes, with the share expecting a platform change within three months falling from 38.3% in June to 28.8% in July, even as production adoption and accelerator utilisation rose.
Interest in Nvidia alternatives is particularly strong among senior decision-makers. Among C-suite respondents, the share likely to evaluate non-Nvidia accelerators rose from 42.9% in June to 57.1% in July, while among final decision-makers it increased from 35.4% to 50%. Smaller organisations are also driving the trend: firms with 251 to 1,000 employees saw the share rise from 41.4% to 53.2%, and those with 101 to 250 employees from 33.3% to 57.7%.
The survey also highlights growth in production adoption of major cloud platforms. Microsoft Azure posted the largest increase, with the share of respondents reporting it in production jumping from 29% in June to 47.1% in July, an 18.1 percentage-point rise. Google's Gemini remained the most-used platform, with production adoption rising from 41.1% to 47.6%, narrowly ahead of Azure. OpenAI's production adoption increased from 40.2% to 49.4%, while Anthropic's rose from 12.1% to 24.7%.
Enterprises are also running their own GPUs more intensively. The share of respondents operating at half capacity or less fell from 83% in June to 69% in July, while the share above 50% utilisation rose from 13% to 23%. The definition of infrastructure effectiveness is becoming more operational, with the share selecting uptime and reliability as important measures increasing from 42.1% to 51.2%, and throughput rising from 21.5% to 24.7%.
Selection criteria are shifting toward workload-level scrutiny. Integration with existing cloud and data stacks remained the top factor, holding steady at around 40%. The share prioritising performance increased from 24.3% to 35.3%, while cost per million tokens rose from 7.5% to 15.9%. By contrast, the share selecting broad total cost of ownership as a leading factor fell from 34.6% to 21.8%, suggesting buyers are focusing on specific operational metrics rather than general infrastructure planning.
The survey indicates that enterprises are becoming more capable operators with better architectures, but they are also setting a higher bar for what their infrastructure must deliver, with reliability leading the way. The declining urgency for platform changes, combined with rising interest in alternatives, points to a market that is maturing beyond a single-vendor default.