While Nvidia gets all the attention among companies providing chips and equipment to AI data centers, there are dozens of others benefiting from the soaring spending from the industry's hyperscalers. Advanced Micro Devices (AMD 0.05%) and Arista Networks (ANET -2.17%) have both seen their revenue climb thanks to ongoing AI spending. But if you can buy only one of them, AMD is the stock to own right now.
Both companies are executing well with strong demand for their products and a huge runway as AI spending takes off. But AMD's stock looks more attractive for multiple reasons. Here's what investors need to know.
Image source: Getty Images.
AMD: the also-ran GPU maker is still a big business
AMD's management sees the AI accelerator market, which includes GPUs and custom-built silicon, exceeding $500 billion by 2028. That's more than 60% annual growth from 2025 to 2028.
For reference, the company's data center revenue, which includes GPUs designed for AI training and inference as well as CPUs, totaled just $12.6 billion last year. So, even if management's estimates turn out to be too high and it captures just a tiny sliver of that market, there's still a lot of growth ahead for the business.
And AMD is positioned to capture a good portion of that market. It introduced several new products at its Advancing AI event earlier this month, including the Instinct MI350 series of GPUs.
It also said the rack-scale MI400 will come out next year, which will compete with Nvidia's Rubin line, scheduled to come out in the second half of 2026. AMD says the MI400 series will be 10 times more powerful than the 300 series, while Nvidia expects a 3.3 improvement multiple in its next generation.
With improved relative performance, AMD's position as an alternative to Nvidia for hyperscalers is strengthening. Its unlikely to overtake the market leader, but it's important for companies building out new data centers to have an alternative if there are supply shortages and to prevent overdependence on one supplier.
AMD's strong position in data center CPUs could also make it a staple for years to come. It's consistently taking share in the market, which provides a steady and growing base for its data center business.
But that also means that this revenue probably won't outpace Nvidia's growth until GPUs become the main source of revenue for the segment. For example, AMD's data center revenue increased 57% last quarter versus Nvidia's 73% increase.
Shares of AMD currently trade for about 37 times forward earnings estimates. That looks expensive at first blush, but earnings will be weighed down by charges related to Chinese export controls.
Without those headwinds, analysts see earnings growing 47% in 2026, and they could rise more than 20% per year from there with strong data center revenue and overall margin expansion. So, at just 24.5 times 2026 earnings expectations, it looks like an interesting investment opportunity.
Arista: the market leader in data center networking
Arista makes network switches that can scale up to huge AI accelerator clusters while maintaining high data-transfer speeds.
When it comes to AI training, ensuring that data can get from one server to another is of the utmost importance, as any level of latency is magnified exponentially as clusters and AI models grow bigger. The company's equipment helps ensure limited downtime for the expensive AI accelerator chips lining the server racks of AI data centers.
The combination of high-end hardware and its extensible operating system (EOS) gives Arista a key competitive advantage. EOS makes it easier for data center operators to get the most out of their equipment with AI-focused features like the newly released cluster load balancing -- in which a cluster of servers maximize performance by distributing incoming traffic among themselves to avoid overwhelming any one server.
Nvidia is starting to encroach on Arista's market, however, with the introduction of its Spectrum-X networking platform. The system is designed to work with its GPU clusters, taking advantage of Nvidia's dominant position in the chip market.
But Arista's modular and programmable system is much more flexible, and the switching costs of overhauling entire data centers are significant. As such, it's hard to see the company losing its top position in networking equipment, especially as AI accelerator purchases evolve.
As spending on AI accelerators climbs toward that $500 billion-plus estimate from AMD's management, Arista should benefit as the top provider of network switch equipment. With most of that growth coming from its biggest customers, investors shouldn't expect strong gross margin expansion.
Meanwhile, the company may have to spend heavily on marketing, along with research and development, to fend off competition from the likes of Nvidia, which may limit operating leverage. That said, Arista's profit margins are already relatively high.
With the stock trading for about 37 times forward earnings expectations, that makes it just as expensive as AMD. But analysts don't see earnings rising at the same rate as AMD's will, with an average increase in earnings per share of just 18% over the next three years. As such, the price looks much more expensive for what you get compared to AMD, and it's worth waiting for the stock to come down before investing.
So, investors looking to invest in one of the top AI stocks not named Nvidia could do well with one of its closest competitors, AMD.
