AMD Launched New AI Network Interface Cards
The company introduced the Pollara 400 and Vulcano 800 cards to support massive AI cluster bandwidth requirements.
Updated on Oct. 13, 2026 in Semiconductors
Live Poll
Is now a good time for your organization to upgrade network infrastructure for AI workloads?
AMD has launched the Pensando Pollara 400 and Vulcano 800, two new network interface cards designed to manage the high-speed demands of AI clusters. These platforms are engineered to overcome the limitations of conventional Ethernet transport for large-scale data environments.
Why it matters
Conventional Ethernet networks often struggle with the congestion and complex multipath requirements found in modern artificial intelligence clusters. These new cards address these bottlenecks by offloading congestion control and improving data flow for massive GPU-driven systems.
The Pollara 400 utilizes a P4-based hardware pipeline and supports RoCEv2 and Ultra Ethernet Consortium standards. A configuration employing three NICs per GPU achieves up to 2.4 Tbps of total bandwidth.
The players
AMD
This is a semiconductor company that designs computer processors and related technologies for business and consumer markets.
The details
The architecture utilizes packet spraying, selective retransmission, and path-aware congestion control to optimize traffic rerouting. AMD validated the Pollara 400 performance using its own Instinct MI350 and MI355 GPU clusters.
Timeline
October 13, 2026: AMD officially published details regarding its new AI networking hardware.
The Tech Race
The shift toward purpose-built AI networking hardware represents a move away from legacy Ethernet systems that cannot handle massive GPU-to-GPU data speeds. AMD is positioning its Pensando platform to compete directly with specialized interconnect technologies used by rival chipmakers.
These specialized components will primarily impact large-scale data centers and cloud providers by enabling faster processing times for intensive AI models. For end users, this hardware is intended to reduce latency and improve the performance of cloud-based artificial intelligence applications.
The takeaway
Reliable data transport is becoming as critical as raw processing power in the development of modern AI systems. Organizations managing large GPU clusters should monitor these networking advancements to ensure their infrastructure can support future scaling requirements.
Further reading
Learn more about the latest innovations in the Semiconductors sector.
Live Poll
Is now a good time for your organization to upgrade network infrastructure for AI workloads?
