$NVDA Ultrafast and Impacts OpenAI's GPT-6 Astra runs on Nvidia GPU and not ASICs. How they did it will have a big impact to AI Clouds and Neoclouds. “NVIDIA’s deep investment in tooling and documentation has enabled us to make our models exceptionally good at programming Blackwell and Rubin GPUs” (1). Note OpenAI is using GPT it do programming directly on Nvidia GPUs, likely in PTX. This skips all the layers of Cloud Software as the application is programmed directly onto Nvidia GPUs. In the future this makes bare metal GPUs + Kubernetes the industry standard that AI Natives work the model that enables OpenAI to program directly on Nvidia GPUs will spread. All this human written Cloud Software that optimizes inference performance? OpenAI does their inference stack now using AI to
View on X ↗