Nvidia is repositioning its competitive advantage beyond GPUs to the entire system built around them, according to an analysis by TechCrunch following the company’s better-than-expected Q2 earnings report.
For years, the prevailing narrative around Nvidia was that its GPUs were the sole supply for AI labs, generating massive profits. But as hyperscalers (Amazon, Google) begin developing their own chips, the question arises whether Nvidia’s advantage is sustainable.
Nvidia’s answer is that its advantage does not lie in GPUs alone. The Vera Rubin architecture is currently rolling out, combining Rubin GPUs with the Vera CPU, Groq 3 LPX (an inference accelerator), along with dedicated storage and networking racks. If the GPU is the engine, the rest is the entire car — and Nvidia is leading the way in building that complete vehicle.
According to Jason Hardy, Nvidia’s Vice President of Storage: “Vera is critical because you can only attach so much memory to a server.” As data centers scale, delivering data to GPUs on time is no simple feat. The Vera CPU helps accelerate these operations by up to 3 times, enabling maximum utilization of flash without bottlenecks.
OpenAI’s approach with its Jalapeño chip reflects the same logic: designing chips to minimize data movement and keep the entire workload within an integrated system. The difference is that Jalapeño is OpenAI’s custom chip, whereas Vera Rubin is Nvidia’s off-the-shelf ecosystem.
This opens up a new competition on a different level: building a rival GPU is only part of the battle; what matters more is making the entire system run efficiently. And at this early stage, Nvidia holds a clear lead.