Arista Networks continued its strong performance into the second quarter of 2026, with revenues surpassing the $3 billion mark for the first time.
The vendor posted $3.036 billion in Q2 revenues, a 12.1 percent jump from the first quarter of 2026 and a 37.7 percent increase Year-on-Year (YoY). To put the networking vendor’s earnings into perspective, Arista netted $2.32 billion in revenue for the whole of 2020.
Arista CEO Jayshree Ullal told investors that Arista was experiencing “significant demand” and that helping drive that momentum was its Ethernet switch line, which has since surpassed 100 cumulative customers from mere single digits in 2024.
While supply chain challenges and rising component costs continue to bite, Ullal was bullish about Arista’s position, saying it was making “solid progress” addressing its “tight” supply chain
“While the industry-wide supply tightness and rising component costs persist, Arista has taken individual and aggressive proactive steps,” Ullal said.
To get ahead further, Ullal highlighted to investors the recent hire of former Nefeli Networks chief executive Eugenia Corrales to spearhead Arista’s global operations while almost tripling its multi-year purchase commitments to $9.7 billion, compared to just $3.6 billion a year ago.
“We’ve secured multi-year agreements with leading vendors of strategic components, qualified new suppliers in key areas to limit risk, and built out supply chains for next-gen AI technologies,” Arista COO Todd Nightingale added. “Our memory supply has been secured for 2026, and we have extended visibility well into 2027 across DDR4 (fourth-generation double data rate), DDR5, and NAND (non-volatile flash) memory. Importantly, we’ve increased our resiliency through optionality and expanded vendor qualification.”
Following its strong outing, Arista raised its revenue expectations for the third quarter to approximately $3.3 billion. The vendor also pushed up its guidance for full-year revenue to $12.6 billion, representing a 40 percent growth.
Following the debut of its 1.6-terabit switches in June, which are slated to ship sometime in early 2027, analysts at William Blair project “a path for sustained high growth” for the vendor over the next year.
Arista bets big on distributed compute
With its supply chain largely secured, Arista’s earnings call saw its leadership team tout a potential growth area in scale-across, or distributed compute networking.
Ullal told investors that its supply chain maneuvering has positioned Arista as a key player in this emerging data center market segment, as well as more traditional scale-out (cluster-to-cluster interconnectivity).
While scale-across as a concept has gained traction of late, Ullal said that the segment continues to be a big thing owing to “a constant state of scarcity for power, space, and compute,” with distributed computing providing a means to utilize already online, and yet disparate facilities.
“Getting all those three to lock in, whether you’re a cloud titan, an AI titan, or a neocloud, it’s going to be very, very difficult,” Ullal said. “Currently, our scale of cost is dominated by the cloud and AI titans, but I see no reason why it wouldn’t apply to, and we’re already seeing it apply to several neoclouds. I would say less in the enterprise and definitely more in the neoclouds and titans.”
Addressing analyst questions around Nvidia’s networking dominance in the scale-up domain, Ullal said Arista “does better” in scale-out and across domains when pitted against NVlink, but has eyes on the second stage for scale-up when it comes to non-Nvidia accelerators, noting: “We’re excited about the Google TPUs (tensor processing units). Increasingly, we see that as a formidable training processor.
“We’re very excited about the range of inference accelerators as well, and we have a number of partners who are working [on] that,” Ullal added. “You can imagine many of our customers are building their own in-house. … Where it’s non-Nvidia, Arista will be excited and will be working more closely both in the scale-up and scale-out, in some cases to build custom racks with their custom processors so that we can better tune our network with the behavior for their inference or training engines.”
That doesn’t mean the vendor is ignoring scale-up in the long term, reminding that only a few years ago InfiniBand held the protocol crown: “Now we don’t mention InfiniBand, but it took two, three years to move to Ethernet.”
“It will take time to go from a proprietary scale-up that’s been around a long time with NVLink to these other alternatives, even if Ethernet’s really good. I expect us to do much better there,” Ullal said.
More in Investment / M&A / Financing
More in Networking
