Amazon is struggling to compete with Nvidia’s dominant AI chips, facing challenges such as low usage, compatibility gaps, and migration issues that are putting millions of dollars in cloud revenue at risk, according to internal documents and sources. This highlights the difficulty for tech companies like Microsoft, OpenAI, Google, Meta, and Amazon to break Nvidia’s hold on the market.
Stacy Rasgon, a chip analyst, mentioned that no one has been able to make a significant impact in Nvidia’s business, with AWS chips not seeing high volumes in usage.
Amazon Web Services, the leading cloud company, is known for designing its own data-center chips to improve profitability. With the new AI chips Trainium and Inferentia, Amazon aims to provide powerful AI services without relying on expensive Nvidia GPUs. However, adoption of these chips has been low compared to Nvidia’s offerings.
The internal documents reveal that AWS chips face challenges due to customer adaption of Nvidia’s CUDA platform, making it hard for AWS to break into the market. While Amazon executives have been optimistic about the AI-chip efforts, the adoption rate among large customers remains low.
Despite Amazon’s efforts with Trainium, Inferentia, and other chips, Nvidia continues to dominate the AI-chip market. AWS measures the success of its AI chips based on usage, customer feedback, and not necessarily on workloads.
AWS faces a major roadblock in Nvidia’s CUDA platform, which has become essential for AI developers due to its ecosystem and compatibility with Nvidia GPUs. Without achieving parity with CUDA capabilities, AWS’s Neuron platform may not see widespread adoption.
While Amazon works on improving its AI chips and building partnerships with the open-source community, some customers like Snowflake choose Nvidia GPUs over AWS chips due to familiarity with CUDA. This has led to challenges in replacing the demand for Nvidia GPUs with AWS offerings.
Despite these hurdles, Amazon remains focused on expanding its AI-chip business and generating revenue through its cloud services. While facing competition from Nvidia and challenges in customer adoption, Amazon is working on enhancing its AI offerings and open-source support to stay competitive in the evolving market.
Article Source
https://www.businessinsider.com/amazon-nvidia-aws-ai-chip-dominance-gpu-trainium-inferentia-2024-5
