Nvidia's edge shifts from GPUs to full-system orchestration

For years, the Nvidia story was about its state-of-the-art GPUs, which became immensely profitable as the industry scaled out. However, with hyperscalers like Amazon and Google building their own chips, investors have wondered how durable that advantage is, especially after Nvidia's market cap grew 10x between the start of 2023 and mid-2025 before taking a more modest trajectory. The earnings call this week is presenting a new narrative: as AI compute grows into the gigawatt scale, the challenge of orchestrating a megascale data center at peak efficiency is growing, and Nvidia has built much of the state-of-the-art hardware to handle it. This shift is reflected in the Vera Rubin architecture, where components like the Vera CPU are specialized for data orchestration rather than just churning through tokens, aiming to drive tokens-per-watt lower. The common logic between Nvidia's approach and OpenAI's Jalapeño chip design is the focus on smarter traffic control instead of just more processor cycles. This opens up a new layer of infrastructure for companies to compete over, where the ability to make the entire system work efficiently is becoming more important than the GPU itself, and Nvidia appears to have an early lead.