According to Sina Finance, Amazon CEO Andy Jassy said on November 17 that Nvidia had long been an important partner and that the two companies would continue working together, despite a recent controversy involving Nvidia and Jensen Huang. Jassy said he did not know all the details but understood that the parties were negotiating a resolution. “I can’t foresee a day when we are no longer partners with Nvidia,” he said.
Amazon and Nvidia announced an expansion of their strategic partnership in August. They plan to deploy an additional 2 million Nvidia GPUs across AWS’s global infrastructure from 2027 to 2028 and deepen cooperation on AI factories, CPUs, networking, open models, data processing and robotics.
Amazon is also developing its own chips. It acquired chip designer Annapurna Labs in 2015, and its portfolio now includes Trainium for AI training and inference, Graviton for general-purpose cloud computing, and the Nitro system for security and networking. Amazon’s third-quarter results showed triple-digit year-over-year growth in its chip business, while revenue commitments for Graviton and Trainium each rose more than 20-fold. Jassy said Trainium’s price-performance advantage over alternatives appeals to customers.
In October, Amazon released the open-source TorchNeuron tool, which lets developers run existing PyTorch code on AWS Trainium chips by changing one setting. Amazon said the move gives developers more choice and lets them benefit from lower costs and faster iteration. Some observers said reducing migration costs could create an opening in Nvidia’s CUDA software ecosystem.
Jassy rejected the view that setbacks for competitors would benefit Amazon’s chip business, saying its in-house chips were growing at a triple-digit annual rate and would succeed regardless of external conditions. He said customers choose technologies and services based on their capabilities, not on what is happening around a company. He described Amazon more as Nvidia’s partner than its competitor, despite some business overlap, and said he expected Nvidia’s business to remain strong for the foreseeable future.
Investor and AI expert Guo Tao told First Financial Daily that technology companies can integrate their chips with their own AI models and cloud services. He said this can support end-to-end optimization across hardware, compilers and models, while companies retain general-purpose GPUs for leading-edge training and shift some high-concurrency inference workloads to in-house chips.