New Era in Tech: Formation of Ultra Ethernet Consortium and its potential Impact on AI & HPC Networking

Nine prominent technology companies, including Arista Networks, Cisco Systems, and Hewlett Packard Enterprise, have joined forces to establish the Ultra Ethernet Consortium. The primary objective of this consortium is to harness the power of existing Ethernet technologies and develop an innovative architecture specifically designed for high-performance artificial intelligence (AI) and high-performance computer (HPC) networking. This collaboration holds tremendous potential for transforming the way data centers and wider area networks operate.

Consortium’s plan and timeline

The Ultra Ethernet Consortium plans to leverage the strengths of established Ethernet technologies to create an architecture that can effectively support the escalating demands of AI and HPC networking. By capitalizing on existing infrastructure, the consortium aims to expedite progress and development in this field. The members envision a timeline that foresees the release of standards-based products by 2024, paving the way for networking companies to achieve significant revenue recognition by mid-2024.

Potential impact on Nvidia

Nvidia, a prominent player in the AI chip market, has experienced substantial growth this year driven by the increased demand for its chips in data centers. However, this new standard introduced by the Ultra Ethernet Consortium could present a formidable challenge to Nvidia’s dominance. As the largest seller of InfiniBand chips, Nvidia has enjoyed the synergies between its AI chip technology and InfiniBand sales. The emergence of the Ultra Ethernet Consortium has the potential to disrupt this landscape and create new dynamics within the industry.

Target of the Ultra Ethernet Consortium

The primary focus of the Ultra Ethernet Consortium is to develop AI networks that effectively connect data centers to wider area networks. By improving the networking capabilities in these environments, the consortium aims to enhance the performance, efficiency, and scalability of AI and HPC applications. This, in turn, can enable significant advancements in fields such as machine learning, deep learning, and scientific research.

Founding members of the consortium

The Ultra Ethernet Consortium brings together a formidable lineup of founding members. In addition to Arista Networks, Cisco Systems, and Hewlett Packard Enterprise, the consortium includes industry-leading chipmakers Advanced Micro Devices, Broadcom, and Intel. Meta Platforms, Microsoft, and Eviden, renowned players in their respective domains, are also part of this collaborative endeavor. It is noteworthy that despite Nvidia’s position as the largest seller of InfiniBand chips and its recent acquisition of chipmaker Mellanox, the company has chosen not to participate as a member of this new consortium.

Market response

In response to the news, market performance fluctuated for the founding members. Nvidia’s stock experienced a decline of 1.9%, settling near $462. Meanwhile, Cisco’s stock observed a modest rise of 0.7%, reaching $52.77, and Arista’s stock experienced a slight dip, resting at $175.50. These market movements suggest that investors are keenly observing the implications and potential competition arising from the establishment of the Ultra Ethernet Consortium.

The Ultra Ethernet Consortium marks a significant milestone in the advancement of AI and HPC networking. By harnessing the potential of Ethernet technologies and creating a specialized architecture, this consortium aims to revolutionize the performance, scalability, and connectivity of AI networks spanning data centers and wider area networks. The impact of this collaboration extends beyond the consortium’s founding members and has the potential to reshape the landscape for major technology companies like Nvidia. As the consortium progresses towards the development and release of standards-based products, the industry eagerly awaits the transformative possibilities this initiative may bring.

Explore more