Billionaire entrepreneur Elon Musk has weighed in on the AI hardware battle, saying that AMD’s accelerators are “quite good” for running small to medium-sized artificial intelligence models.
At the same time, Musk acknowledged that NVIDIA remains the top choice for massive training workloads, reflecting the current balance of power in the AI accelerator market.
Musk’s xAI Uses Both AMD and NVIDIA #
Musk’s AI company, xAI, has already deployed AMD Instinct MI300/MI300X accelerators for part of its AI infrastructure. These GPUs power inference, fine-tuning, and medium-scale foundational models, where AMD provides strong throughput and cost efficiency.
For large-scale model training, however, xAI still relies primarily on NVIDIA GPUs, which dominate the high-end market. This “division of labor” highlights how enterprises are leveraging AMD for certain workloads while sticking with NVIDIA for critical training at scale.
Why NVIDIA Still Leads in Large-Scale AI Training #
NVIDIA’s edge comes from its CUDA ecosystem, a closed-loop platform built over years that integrates hardware, software, and developer tools. This ecosystem creates high switching costs for developers, making it difficult for competitors to displace NVIDIA in the enterprise AI space.
Meanwhile, AMD has long been in catch-up mode. Despite significant progress in hardware, it has struggled to match CUDA’s ease of use and developer adoption.
AMD’s Progress: ROCm and Instinct Hardware #
The tide is slowly shifting. With the continuous improvement of the ROCm software stack, AMD is enhancing compatibility, usability, and developer experience. This, combined with the performance of its Instinct accelerators, is helping AMD gain ground.
- MI300/MI350: Already adopted by some organizations for inference and medium-scale AI tasks.
- MI450 (upcoming): Positioned as a direct competitor to NVIDIA across both training and inference workloads.
- Annual update cadence: AMD now commits to a yearly release cycle for its Instinct series, keeping pace with AI market demands.
Market Reality: Partnerships and Adoption #
Despite AMD’s progress, NVIDIA holds stronger partnerships with tech giants such as Microsoft, Meta, and Google, whose infrastructures are deeply tied to NVIDIA GPUs. AMD has made inroads into some data centers but still has a long way to go in terms of adoption and ecosystem integration.
Musk’s public endorsement acts as a rare validation effect for AMD, boosting its credibility as a viable alternative in specific workloads.
Conclusion: A Shifting AI Hardware Landscape #
Elon Musk’s comments reflect both recognition and reality:
- AMD is becoming a serious player in inference and small-to-medium model workloads.
- NVIDIA remains dominant in large-scale training thanks to CUDA and deep ecosystem lock-in.
As AI applications diversify, competition between AMD and NVIDIA will only intensify. With AMD’s accelerated product cadence and ROCm improvements, the next few years could bring a much more competitive AI hardware market.