{"id":46805,"date":"2026-06-17T17:13:22","date_gmt":"2026-06-17T16:13:22","guid":{"rendered":"https:\/\/www.gsma.com\/solutions-and-impact\/technologies\/networks\/?post_type=gsma_theme_resources&p=46805"},"modified":"2026-06-18T10:31:35","modified_gmt":"2026-06-18T09:31:35","slug":"computing-on-the-critical-path-network-compute-determinism-for-ai-native-services","status":"publish","type":"gsma_theme_resources","link":"https:\/\/www.gsma.com\/solutions-and-impact\/technologies\/networks\/gsma_resources\/computing-on-the-critical-path-network-compute-determinism-for-ai-native-services\/","title":{"rendered":"Computing on the Critical Path: Network-Compute Determinism for AI-Native Services"},"content":{"rendered":"\n
\n
Download report<\/a><\/div>\n<\/div>\n\n\n\n
<\/div>\n\n\n\n

The evolution of mobile networks toward 5G and 5G-Advanced has established a strong foundation for deterministic communication. Mechanisms such as QoS differentiation, network slicing, and URLLC enable prioritised handling of latency- and reliability-sensitive traffic across the radio and transport domains. According to 3GPP specifications, 5G systems target user-plane latency down to the millisecond level with extremely high reliability for selected industrial and critical communication scenarios. 3GPP Releases 18 and 19 introduce further enhancements such as AI\/ML-assisted RAN operations, improved support for XR services, and more advanced mobility management mechanisms, strengthening the network\u2019s ability to adapt to dynamic traffic patterns and heterogeneous service requirements.<\/p>\n\n\n\n

At the same time, the rapid expansion of AI-enabled services is fundamentally changing the nature of end-to-end service behaviour in mobile systems. An increasing number of services\u2014such as embodied intelligence, large-scale safety monitoring, and interactive AI devices\u2014no longer depend solely on communication performance. Instead, service outcomes are jointly determined by data transmission, computing execution, and decision-making latency across terminals, edge platforms, and central clouds. In these services, computation is no longer an internal platform concern but an integral part of the service delivery path, directly affecting responsiveness, safety, and reliability.<\/p>\n\n\n\n

This shift exposes a structural limitation of existing approaches. While current mobile networks provide mechanisms to prioritise and isolate traffic, they do not directly account for the execution time, scheduling behaviour, or resource contention of AI workloads once data reaches the computing domain. As a result, even when network-level latency targets are met, end-to-end service behaviour may still be dominated by compute-side variability. This gap becomes increasingly visible as AI inference and analytics move from background optimisation to the critical path of real-time services.<\/p>\n\n\n\n

Looking toward 6G, this challenge is gaining broader recognition. 3GPP studies beginning with Release 20, including work in TR 22.870 and TR 23.801-01, identify computing and AI as native capabilities of future systems and highlight the need for tighter coordination between communication and computation. In parallel, industry discussions around AI-Native Networks [8][9] point toward architectures where network and compute resources are jointly managed to support service-level objectives. These developments indicate a clear trajectory: deterministic communication alone is no longer sufficient, and deterministic computing must be considered as part of end-to-end service assurance.<\/p>\n\n\n\n

The growing inseparability of AI execution and mobile connectivity presents a shared industry challenge. Computation is now part of the end-to-end service path, so the challenge is to establish a common understanding of deterministic behaviour and the technical consistency needed to achieve it. Addressing this challenge calls for a common problem framing, a clear articulation of existing enablers and their boundaries, and a shared view on architectural directions that can guide future alignment across standards, platforms, and deployments. This white paper supports that objective by examining representative AI-enabled services with deterministic demands on both communication and computation, analysing current capabilities and limitations, and outlining architectural directions that can progressively enable bounded and predictable end-to-end behaviour over 5G, 5G-Advanced, and future networks, providing a common reference for discussion and collaboration toward end-to-end determinism.<\/p>\n\n\n\n

<\/div>\n\n\n\n

Complete the form below and download the whitepaper<\/h2>\n\n\n