According to a report by Ookla, India’s 5G networks are facing challenges in supporting real-time artificial intelligence (AI) applications due to relatively slow upload speeds and high multi-server latency.
While India ranks ninth globally in headline 5G download speeds, the report found that the country’s multi-server latency stands at 51.6 milliseconds (ms), above the sub-50 ms threshold considered optimal for AI applications such as large language models (LLMs) and agentic AI.
As per Ookla, India is one of only four markets, alongside South Korea, the United States and Spain, that remain above this latency threshold.
The report said multi-server latency is the most appropriate benchmark for evaluating AI readiness because it reflects network responsiveness under typical operating conditions, matching the performance requirements of AI workloads.
The report noted that the AI era demands a shift from traditional download-centric network performance metrics, as emerging AI applications require significantly lower latency and improved uplink capabilities.
Further, the report highlighted that text-based LLM interactions currently account for the bulk of AI traffic on mobile networks. These workloads involve a two-way communication pattern in which user prompts are transmitted upstream to cloud servers before responses are streamed back.
More advanced AI applications, including multimodal AI and augmented reality (AR), place even greater demands on network infrastructure. Unlike conventional mobile usage, which typically follows a 90:10 downlink-to-uplink traffic ratio, AR applications with cloud-based processing can increase uplink traffic to around 40 per cent or more, depending on application complexity.
Ookla found that India allocates only 7.53 per cent of its 5G throughput to uploads, resulting in a median upload speed of 15.75 Mbps. Although the country’s upload share has increased by 1.53 percentage points between 2023 and 2025, it remains below the 20 Mbps benchmark considered necessary for advanced AI applications such as AR and multimodal vision.
The report also highlighted cloud infrastructure latency as another major bottleneck. While mobile networks can optimise connectivity to their own network edge, latency between telecom networks and cloud data centres continues to affect AI performance.
According to Ookla, median cloud latency in India stands at 108 ms for Microsoft Azure, 114 ms for Amazon Web Services, 115 ms for Google Cloud and 158 ms for Oracle Cloud Infrastructure.
