Amazon EC2 G7 instances with NVIDIA RTX PRO 4500 Blackwell GPUs now GA

AWS made Amazon EC2 G7 instances generally available, powered by NVIDIA's RTX PRO 4500 Blackwell Server Edition GPUs. AWS quotes up to 4.6x AI inference performance and 2.1x graphics performance over the previous G6 generation, positioning G7 for inference serving, graphics rendering, and analytics workloads rather than the largest training runs (which target higher-end instances).

The generational jump reflects the Blackwell architecture's efficiency gains and gives AWS customers a cost-effective option for production inference — the workload that increasingly dominates real-world AI spend. By offering RTX PRO Blackwell silicon in a mainstream instance family, AWS broadens access to current-gen acceleration without requiring the most expensive flagship GPUs.

The launch is notable in the context of the same week's news that Amazon is exploring selling its own custom AI chips to undercut Nvidia. The two are not contradictory: AWS continues to be one of Nvidia's largest customers while simultaneously building Trainium/Inferentia alternatives — a hedge-and-partner strategy. For developers, G7 is a concrete, available upgrade today; for the strategic picture, it underscores how dependent even Nvidia's would-be challengers remain on Nvidia hardware in the near term. The figures are AWS's own benchmarks, so independent inference results across real models will be the practical test.