Previously, the prevailing perception of Nvidia centered on its near-monopoly during the initial years of the AI boom, being the sole provider of cutting-edge GPUs. This led to immense profitability as the industry scaled. However, in recent years, major hyperscale players like Amazon and Google have begun developing their own chips, challenging Nvidia’s singular position and prompting investor questions about the long-term sustainability of its competitive edge.
While this narrative is compelling and largely accurate, Nvidia’s shares, after a tenfold market cap increase between early 2023 and mid-2025, have seen a more moderate growth trajectory over the past year due to increasing GPU competition.
Following the company’s earnings report on Wednesday, a new understanding has emerged: Nvidia’s competitive advantage extends significantly beyond just GPUs. As AI computing power escalates to gigawatt scales, managing and coordinating these vast resources has become an extraordinarily intricate challenge. Nvidia, unsurprisingly, has engineered much of the advanced hardware essential for this orchestration, providing a substantial lead in the peripheral systems surrounding the GPU, even as direct GPU competition intensifies. Despite discussions portraying compute as a commodity, operating a hyperscale data center at peak efficiency remains incredibly complex, a difficulty that only grows with larger and faster deployments.
System by System: Beyond the GPU
This broader capability is evident in the specific components Nvidia is currently marketing. The company is deploying its Vera Rubin architecture, which integrates the Rubin GPU with a suite of other units, including the Vera CPU, the Groq 3 LPX inference accelerator, and complementary racks for storage and networking.
In recent discussions with Nvidia personnel, I’ve gained surprising insights into the functions of these systems. Much like the Rubin GPU itself, they are highly specialized; however, their role isn’t to process tokens, but rather to ensure that every component external to the GPU operates with maximum efficiency. If the GPU serves as the engine, these other systems constitute the rest of the vehicle.
The Vera CPU, in particular, is designed to tackle the complexities of data orchestration. Jason Hardy, Nvidia’s VP of storage technology, explained, “Vera is important because there’s only so much memory that you can put in a single server or any sort of compute platform.”
As data centers have expanded their computing power, memory capacity has likewise increased, enriching companies like Micron during this second phase of the infrastructure boom. Yet, delivering that data to the GPU precisely when needed is far from simple. As companies strive to reduce tokens-per-watt, the critical importance of this intelligent data traffic management becomes clear.
Hardy noted, “We saw upwards of 3x improvement in these operations, where the Vera CPU is allowing for acceleration. So now we can use our flash to its fullest potential, because we can get all that performance out of it without bottlenecking.”
Similar challenges are being addressed outside of Nvidia. OpenAI, for instance, focused heavily on preempting these issues by minimizing data movement with its Jalapeño chip.
“We designed Jalapeño to minimize data movement and communication delays,” the company stated in a blog post earlier this month. “Its large domain allows the entire workload to remain within one connected system, minimizing data movement and helping the complete request stay fast and efficient from beginning to end.”
While this represents a different strategy—avoiding data movement by conducting a workload within a single, integrated chip—the underlying principle remains consistent: boosting efficiency through sophisticated traffic control rather than merely increasing processor cycles. This shift, in turn, unveils an entirely new layer of infrastructure ripe for competition among companies.
This emerging emphasis on data orchestration isn’t an automatic victory for Nvidia. The company will still need to contend with rival chipmakers and hyperscalers, just as it does with GPUs. However, the battleground has moved to a new domain where the ability to ensure the entire system operates efficiently holds more weight than simply building a competing GPU.
And, at least in these initial stages, Nvidia appears to possess a commanding advantage.