According to a new report from SemiAnalysis published on September 3, 2025, Amazon Web Services (AWS) is deploying a plan to scale up its Trainium chips to multi-gigawatt levels alongside partner Anthropic in an effort to regain the upper hand in the AI cloud computing race. This is a crucial move as the tech giant faces significant pressure from its direct competitor, Microsoft Azure.
Background
About two and a half years ago, analysts warned of a looming "cloud crisis" at AWS. Although the cloud services division remains the crown jewel of the Amazon empire, contributing roughly 60% of the conglomerate's operating profit, AWS seemed to stumble in transitioning quickly to the era of GPU and XPU-accelerated cloud computing. This delay allowed Microsoft Azure to surge ahead and temporarily take the lead in the new wave of AI infrastructure.
Key Developments
To counter this situation, AWS has decided to place a major bet on its custom-designed Trainium chip family by partnering closely with Anthropic. The hardware capacity expansion plan is expected to help alleviate dependency on Nvidia's scarce GPU supply. Designing its own silicon not only optimizes operational costs but also allows AWS to offer more efficient large language model (LLM) training and inference solutions to its partners.
Why It Matters
AWS's custom chip strategy shows that the global AI race is shifting deeply into the core semiconductor infrastructure layer rather than just stopping at application software. For engineers and tech enterprises in Vietnam, this fierce competition promises to cool down AI cloud infrastructure rental costs in the medium term. At the same time, having alternative options to the Nvidia ecosystem will help local developers optimize budgets for operating their own AI projects.