AMD has officially announced the general availability and shipping of its innovative Pensando Pollara 400 AI NIC, targeting the accelerating demands of Artificial Intelligence, large language models, and agentic Artificial Intelligence applications. The surge in these advanced workloads is prompting data centers and enterprises to seek parallel computing infrastructures that not only offer superior performance but are also adaptable to evolving Artificial Intelligence and machine learning needs. A critical technological challenge in this context is scaling intra-node GPU-GPU communication networks for maximum efficiency.
Embedding its commitment to open ecosystems and customer choice, AMD designed the Pensando Pollara 400 AI NIC as a fully programmable network interface controller aligned with the nascent standards defined by the Ultra Ethernet Consortium (UEC). This new offering brings the promise of reducing total cost of ownership by enabling flexible, future-proof data center architectures without compromising on performance. The Pollara 400 is engineered to provide the scalable, high-throughput connectivity required for Artificial Intelligence and machine learning clusters, making it easier for organizations to build and expand robust Artificial Intelligence infrastructure.
The launch signifies AMD’s strategic focus on both innovation and open standards across Artificial Intelligence networking. By supporting high-speed, programmable, and UEC-compliant network capabilities, the Pensando Pollara 400 is positioned as a pivotal component for next-generation Artificial Intelligence data centers. With the product now available and shipping to customers, AMD is reinforcing its role in accelerating Artificial Intelligence deployment at scale, ensuring organizations can meet both current and future infrastructure demands for increasingly complex and resource-intensive Artificial Intelligence workloads.
