Enterprise AI buyers show stronger interest in alternative accelerators over Nvidia's upcoming GPUs
A July survey of 170 AI infrastructure decision-makers found 39.4% plan to evaluate non-Nvidia accelerators like AWS Trainium, Google TPU, AMD Instinct, Intel Gaudi, or custom ASICs in the next year, versus 25.3% for Nvidia's next-generation Blackwell GPUs. This 14-point gap suggests growing openness to alternatives, though Nvidia remains the default in most production environments.
A July survey of 170 AI infrastructure decision-makers reveals a notable shift in hardware evaluation plans. Over the coming year, 39.4% of respondents intend to assess non-Nvidia accelerators, including options from AWS, Google, AMD, Intel, and custom ASIC designs. By contrast, only 25.3% plan to evaluate Nvidia's next-generation Blackwell GPUs, creating a 14-point gap in buyer interest.
This divergence signals growing curiosity about alternative compute paths, even as Nvidia maintains its dominant position in active production environments. The survey suggests that while Nvidia remains the default choice for deployed AI workloads, procurement teams are increasingly willing to explore competitive options, potentially reshaping future infrastructure decisions as the AI hardware market matures.
This trend could gradually loosen Nvidia's grip on enterprise AI infrastructure, potentially affecting pricing and innovation across the sector. If alternatives gain production traction, organizations may gain more negotiating leverage and cost flexibility, which could influence the affordability of AI services downstream. Consumers and businesses relying on AI applications may ultimately benefit from a more competitive hardware ecosystem, though the pace of change depends on whether evaluation interest converts into actual deployment.