Vendors & EquipmentArista NetworksR4 SeriesAI InferencingEdge Computing
Arista pushes R4 switches to the edge for distributed AI inferencing
Arista ships 7282R4 and 7280R4 routing platforms targeting distributed AI inferencing at the edge, after posting $3.7 billion revenue on AI scale-out demand.
Vendors & EquipmentWhy it matters
- Arista unveiled the 7282R4 and 7280R4 at SC26 in Chicago; both are available now.
- The 7280R4 supports 400G and 800G ZR/ZR+ coherent pluggable optics; a new optical line system carries a 25.6 Tb/s amplifier.
- The 7020R4 is now shipping, with the 7020HR4 due in early 2027.
- Arista posted $3.7 billion in revenue in its most recent earnings on strong AI scale-out demand.
- Temperature-hardened R4 platforms are pitched as compute nodes near 5G and AI-RAN cell sites.
The story
Arista Networks has shipped two new R4-series routing platforms, the 7282R4 and the 7280R4, aimed at moving AI inferencing from centralized clusters to the network edge. The vendor announced the additions at SC26 in Chicago, extending a product line originally built for AI cluster interconnects into front-end and edge-routing territory.
The moves matter for operators and enterprises planning distributed AI workloads. Arista argues its new platforms deliver predictable, low-latency connections for workloads spread across sites — a growing pattern as power constraints force operators to split compute rather than concentrate it.
What did Arista announce at SC26?
The headline products are concrete:
- 7282R4: adds dual supervisor engines for redundant control planes.
- 7280R4: a coherent transport platform that takes 400G and 800G ZR/ZR+ pluggable optics.
- Pluggable optical line system: features a 25.6 Tb/s optical amplifier for extended-reach transmission.
Both the 7282R4 and 7280R4 are available now. Arista says they provide the high-performance routing needed to bring inference workloads closer to end users.
The 7020R4, unveiled last year, is now shipping as well. Its sibling, the 7020HR4, is due in early 2027. The 7020HR4 and 7020R4 compact routing platforms support both IPsec and MACsec encryption. The 7280R4 adds integrated MACsec for its coherent optics.
The platform updates also include ruggedness enhancements such as temperature hardening. Arista is pitching the new R4 lines for use as compute nodes in outside plants near 5G and AI radio access network cell sites — a positioning that puts the vendor into telco and edge-compute territory beyond its data center stronghold.
What does the edge push change for practitioners?
With support for deep packet buffering and workload steering, Arista claims the new platform line can offer lossless transport "across any distance." That claim sits squarely in vendor marketing territory and awaits independent validation in production deployments.
The commercial logic is clearer. Brendan Gibbs, Arista's associate VP for AI, routing, and switching platforms, framed the use case in a statement: "Distributing AI inferencing across front-end networks brings compute closer to end users, dramatically cutting round-trip latency and ensuring seamless real-time interaction for agentic AI and complex workflows."
For network teams, the pitch translates into routing platforms that can sit between centralized AI clusters and the access edge, steering inference traffic with buffering deep enough to absorb bursts. The encryption support across IPsec and MACsec addresses the security requirements of traffic leaving the data center perimeter.
How big is the business behind it?
The numbers anchor the strategy. Arista posted $3.7 billion in revenue in its most recent earnings, driven by strong demand for AI scale-out offerings.
That scale-out business remains the core. But the company led by CEO Jayshree Ullal is deliberately looking beyond scale-up and scale-out deployments toward what it calls campus and scale-across demand.
In August, Ullal said the distributed market continues to be a major area for Arista. The vendor is targeting connectivity for sites forced to spread workloads across locations because of power constraints — a constraint increasingly shaping AI data center design as single-site power budgets cap how much compute can be concentrated in one place.
Separating the roadmap from the marketing
Not everything announced is shipping today. The 7282R4 and 7280R4 are available now, and the 7020R4 has moved from announcement to shipment. The 7020HR4 remains a roadmap item until early 2027.
The RAN-adjacent positioning — temperature-hardened platforms serving as compute nodes near 5G and AI-RAN cell sites — marks the most speculative extension. It depends on edge inferencing demand materializing at telco sites, a market that has repeatedly underperformed vendor forecasts over the past decade.
The near-term revenue story rests on AI scale-out, which is already delivering. The edge story is the growth option layered on top. With the 7020HR4 set to complete the compact routing family in early 2027, Arista has given itself a two-year window to prove that distributed inferencing demand justifies the expanded portfolio.
Also reported
Source: sdxcentral.com
More from Daniel Okafor
Show full bio
Senior reporter covering regulation, policy and their impact on Carrier business, 5G and 6G, fiber and broadband at Telecom Gazette.
559 articles