Arista Shifts AI Clients to AMD, Pressuring Nvidia
AI hardware allocation is shifting. Arista Networks CEO reports increased AMD adoption, challenging Nvidia's dominance. This diversification impacts the competitive landscape for AI acceleration, with implications for major tech players.

The competitive landscape for AI acceleration is showing signs of a significant shift, with Arista Networks CEO Jayshree Ullal indicating a notable move towards Advanced Micro Devices (AMD) in their client deployments. This development was evident in recent market activity, as Nvidia's stock saw a nearly 3% decline on Friday, while AMD experienced a gain of close to 1%.
A Changing AI Hardware Allocation
Ullal's commentary during Arista's earnings call revealed a substantial evolution in their customer hardware choices. "A year ago, it was pretty much 99% Nvidia," Ullal stated, contrasting it with the present situation. She elaborated that currently, "20% to 25% of our deployments" are favoring AMD as the preferred accelerator. This represents a dramatic swing from a near-monopoly to a more diversified adoption.
Nvidia's Dominance Under Pressure
While the stock movements were not seismic, Ullal's remarks carry considerable weight. Nvidia has held a commanding position in the AI graphics processing unit (GPU) market since the AI boom gained momentum with the launch of OpenAI's ChatGPT over three years ago. Nvidia currently commands an impressive approximately 90% market share in AI chips. However, challenges are emerging not only from AMD but also from Google, which is experiencing growing adoption of its tensor processing units (TPUs). Nvidia stands as the most valuable U.S. company, boasting a market capitalization exceeding $4.5 trillion. AMD, in comparison, is valued at roughly $335 billion, a figure bolstered by an 85% surge in its stock price over the past twelve months.
Arista's Role and Strategic Adjustments
Arista Networks plays a crucial role in the AI infrastructure ecosystem by supplying the essential Ethernet switching technology that interconnects high-performance chips. Late last year, AMD announced a strategic partnership with Arista to develop bespoke AI clusters designed for both training and inference workloads.
However, this diversification in Arista's deployments is partly driven by strategic necessity. Nvidia has been reducing its reliance on third-party networking solutions by developing its own integrated technology to link its GPUs. This move was highlighted in October when Nvidia announced that Meta and Oracle would be integrating Nvidia's proprietary networking technology into their respective infrastructures, a development that contributed to a significant 10% drop in Arista's stock over two days.
Strategic Departures and Market Reactions
According to Ben Bajarin, a chip analyst at Creative Strategies, the situation reflects a strategic shift. He commented that while "it's not the end of the world," being "designed in with Nvidia helps, and they are being designed out, if not fully, to a degree."
Nvidia's introduction of its Spectrum-X Ethernet platform in 2023 appears to have had a notable impact. The year following its launch saw Arista's stock value decrease by more than half. Although the stock has since recovered some of its losses, climbing 19% last year, and is currently up 6% in 2026, the competitive pressures are evident.
An Nvidia spokesperson declined to offer a comment on the matter, and AMD had not immediately responded to requests for comment at the time of reporting.





