Nvidia's competitive advantage in artificial intelligence infrastructure is no longer confined to its graphics processing units, according to an analysis published by TechCrunch on Wednesday. The company has established a commanding position in the hardware and systems needed to orchestrate AI compute at the gigawatt scale, even as it faces mounting competition from hyperscalers like Amazon and Google in the GPU market itself. As data centers grow larger and faster, managing compute efficiently has become an increasingly difficult challenge, and Nvidia has developed much of the cutting-edge technology required to address it.
The company is currently deploying its Vera Rubin architecture, which combines the Rubin GPU with a suite of additional units including the Vera CPU, the Groq 3 LPX inference accelerator, and comparable racks for storage and networking. These specialized systems focus on ensuring everything outside the GPU operates as efficiently as possible, rather than processing tokens directly. The Vera CPU in particular addresses the challenge of orchestrating data movement as memory capacity scales alongside computing power. Getting data to the GPU at precisely the right moment isn't simple, and as companies push to lower tokens-per-watt, the importance of this traffic management has become clear.
Jason Hardy, Nvidia's VP of storage technology, told TechCrunch that "we saw upwards of 3x improvement in these operations, where the Vera CPU is allowing for acceleration." He explained that the system now enables flash storage to reach its full potential without bottlenecks. The analysis notes that other companies are tackling similar problems through different methods: when OpenAI created its Jalapeño chip, a primary goal was "to minimize data movement and communication delays" by keeping entire workloads within one connected system. Both approaches prioritize efficiency through smarter traffic control rather than simply adding more processor cycles, opening up an entirely new layer of infrastructure for competition.
The report explains that operating a megascale data center at peak efficiency remains extraordinarily difficult despite frequent discussion of compute as a commodity, and this challenge intensifies as deployments expand. The competition has moved to a new level where constructing a rival GPU matters less than making the entire system function efficiently. While this focus on data orchestration doesn't automatically guarantee Nvidia's dominance—the company must compete with rival chipmakers and hyperscalers just as it has with GPUs—it currently appears to hold a substantial lead in the early stages of this shift. After Nvidia's market cap grew tenfold between early 2023 and mid-2025, its shares followed a more modest path over the past year due to concerns about GPU competition, but the company's earnings this week have prompted investors to recognize that its advantage extends far beyond graphics processors. The fundamental question for hyperscalers building their own chips is whether vertical integration alone can match the breadth of Nvidia's orchestration ecosystem, or whether best-of-breed components still require specialized coordination that only a platform provider can deliver at scale.