PRESS RELEASE: Telecom Infra Project's Ethernet-Based Networks for AI/ML Infrastructures gains momentum in 2026
- Telecom Infra Project
- 3 days ago
- 3 min read
As announced in FYUZ 2025, Dublin, TIP established the Ethernet-Based Networks for AI/ML Infrastructures (ENAI) as a new Subgroup within OOPT, with Co-Leads from AT&T and KDDI, and broader community support from Edgecore, IP Infusion and DriveNets.
ENAI targets to define open, interoperable architectures for AI infrastructure, including Ethernet‑based scale‑up and scale‑out within servers and AI clusters, and open IP‑based transport for scale‑across connectivity across distributed data centers.
Ethernet-Based Networks for AI/ML Infrastructures (ENAI), a new Subgroup within the OOPT Project Group, kicked off last December. This initiative, led by AT&T and KDDI, aims to define open, vendor-neutral Ethernet-based network architecture capable of supporting large-scale, distributed AI training and inference workloads and evaluating the impact of east-west traffic patterns in telcos’ networks.
The scale of artificial intelligence (AI) clusters is overflooding the capacity of a single data center in terms of power consumption and space. This is motivating the industry to seek new solutions to “scale across” multiple datacenters, the AI infrastructure. On the one hand, distributed AI inference workloads across datacenters are increasing and impacting, notably, on network traffic patterns telecom operators must adapt to. On the other hand, distributed deployments of AI clusters which can accommodate inference and training workloads across multiple, smaller datacenter locations is a new trend which can solve the abovementioned challenges in terms of power and space, specially in high dense urban areas.
The ENAI subgroup aims to address these scenarios, focus on solutions based on Ethernet technology, by developing reference architectures based on open and interoperable solutions which can broaden and diversify the supply chain meeting the performance requirements of AI.
2026 Objectives
The ENAI Subgroup is focused on:
Building an open ecosystem around the above-mentioned AI infrastructure connectivity use cases, based on vendor neutral IP transport over Ethernet, to accommodate AI workloads in intra- and inter-data center scenarios, across metro and even long-haul transport networks.
Define a reference Ethernet-based scale up and scale out open network architecture for AI infrastructure, as an alternative to InfiniBand, within servers and AI clusters.
Define a reference open, IP‑based scale‑across transport architecture for distributed inter‑datacenter AI use cases, based on workload requirements such as lossless connectivity, latency, and jitter tolerances.
Documenting low-level AI workloads (training and inference) requirements such as lossless connectivity, latency, and jitter tolerances, for distributed “scale across” inter-datacenter use case and define the target transport architecture requirements based on these requirements.
Developing technical product requirements for TOR switches, fabrics, NICs, and optical components conforming the backend network elements.
Establishing benchmarking, testing, and validation criteria aligned with TIP’s test and validation badging program.
This initiative follows a structured 2026 roadmap, beginning with ecosystem expansion and architectural definition, followed by technical requirement development and vendor evaluation.
TIP’s Role as an Industry Accelerator
“ENAI operates within TIP’s globally recognized, operator-driven framework for open collaboration. By leveraging TIP’s governance model, community platform, and established test and validation programs, the subgroup ensures that outcomes will be open, standards-based, interoperable, and aligned with real-world deployment requirements...The ENAI subgroup has already attracted participation and interest from multiple vendors and global operators, reflecting strong ecosystem momentum around AI-ready transport networks. Organizations and other communities interested in participating in the ENAI initiative can join through the TIP community platform.” - Olli Andersson, CTO, TIP
“As AI workloads scale beyond single data centers and are increasingly applied across society, telecom networks are required to meet a new set of transport requirements. Through ENAI, KDDI aims to promote open, Ethernet-based AI infrastructure architectures that are aligned with the broader open-technology ecosystem, while leveraging IP transport for connectivity within and between data centers, as well as across metro and long-distance networks. By working closely with TIP’s global community, we seek to contribute to building the network foundations for the distributed AI era”. - Kenji Kumaki, Chief Architect at KDDI
###
About the Telecom Infra Project
The Telecom Infra Project (TIP) is a global community of companies and organizations that are driving infrastructure solutions to advance global connectivity. Founded in 2016, TIP is a community of diverse participants that includes hundreds of companies – from telecom operators and technology partners including OEMs, chipset manufacturers, vendors, hyperscalers, system integrators and other connectivity stakeholders. The community collectively is redefining network technology by developing, testing and promoting standards based, open, disaggregated, autonomous, cloud-native, AI-driven solutions. These solutions when deployed will deliver the high-quality connectivity that the world needs – now and in the decades to come. Find out more: www.telecominfraproject.com
Media Contacts:
Telecom Infra Project, Inc.
Global Corporate Communications Department
