Enterprise technology buyers are increasingly looking beyond Nvidia as they draw up testing schedules for their upcoming hardware cycles. According to a July survey of 170 artificial intelligence infrastructure managers, 39.4 percent of respondents report that they are likely to evaluate non-Nvidia accelerators over the next 12 months. By comparison, 25.3 percent plan to test Nvidia’s upcoming Blackwell GB300 or other next-generation graphics processing units, establishing a 14-percentage-point preference gap for competing chips.
The non-Nvidia hardware being targeted for future evaluation includes a mix of third-party processors and bespoke silicon. Enterprise teams reported plans to assess cloud-specific offerings like Amazon Web Services’ Trainium and Google’s Tensor Processing Units, merchant chips such as AMD’s Instinct series and Intel’s Gaudi line, and internally developed application-specific integrated circuits.
Even with this shift in testing preferences, Nvidia continues to serve as the default standard inside the majority of live production deployments. However, enterprise infrastructure teams are taking deliberate steps to establish operational flexibility across their hardware roadmaps rather than treating Nvidia as their sole viable option.
This evaluation trend aligns with a broader consolidation phase among enterprise infrastructure operations. Organizations are prioritizing the expansion and performance tuning of their established computing platforms over making immediate, large-scale platform migrations. Consequently, the urgency to execute rapid platform swaps has tempered: the share of respondents expecting to change platforms within three months declined from 38.3 percent in June to 28.8 percent in July. That drop in immediate transition plans materialized alongside rising production usage, increased hardware utilization rates, and continued exploration of alternative cloud providers.
