AMD calls CUDA a ’non-event’, says companies are programming at levels where the Nvidia tech doesn’t really matter
AMD has dismissed CUDA, Nvidia’s widely-used GPU computing platform, as a “non-event” for companies in the AI sector. According to Andrew Dieckman, AMD’s corporate vice president and general manager of the data center GPU business group, the importance of CUDA has diminished significantly in industry conversations.
Dieckman stated that discussions about CUDA have become virtually nonexistent in his customer interactions, noting that major companies have moved beyond direct engagement with the technology. Organizations are instead adopting higher-level programming abstractions and diverse serving frameworks that reduce dependence on Nvidia’s proprietary ecosystem.
The AMD executive attributed this shift partly to the emerging efficacy of AI agents, which he suggested has further diminished CUDA’s strategic importance as a competitive moat. This perspective challenges the conventional view that CUDA represents an insurmountable barrier to market competition.
These comments reflect a broader industry trend toward abstraction layers and standardized approaches in AI development, potentially reducing lock-in effects that historically benefited Nvidia’s market dominance. However, Nvidia remains the dominant player in AI hardware, and CUDA continues to be widely deployed in production environments.