The 5G network on your phone is good enough to run ChatGPT and similar AI chatbots. But the next wave of AI applications, things like augmented reality glasses, real-time AI vision, and immersive AI-powered environments, needs speeds and response times that no 5G network in the world can currently deliver. Not even close.
That is the finding from a new analysis by Ookla, the global network intelligence company best known for its Speedtest tool. The report assessed 5G AI readiness across 22 markets worldwide and found a clear gap between where networks are today and where they need to be for tomorrow’s AI use cases.
Ookla focused on something called latency, which is the time it takes for data to travel from your device to a server and back. In everyday terms, it is the gap between tapping something on your screen and the result actually appearing.
For AI applications, latency matters enormously. The lower it is, the more responsive and usable an AI-powered service feels. When latency is too high, you notice it, with delays, stuttering, and degraded quality.
Ookla set out two thresholds for each AI use case. The minimum threshold is the point below which users start noticing things going wrong. The target threshold is where genuinely smooth, consistent performance becomes possible.
Don't Miss This Article
For text-based AI tools like ChatGPT, the requirement is a latency of under 50 milliseconds. Ookla found that 18 out of 22 markets it assessed already meet this minimum threshold. So if you are using a 5G connection to chat with an AI assistant, your network is almost certainly fast enough for that task today.
For conversational voice AI, where you speak to an AI and it responds in real time, the requirement tightens to under 40 milliseconds. Thirteen of the 22 markets assessed currently meet that bar.
So far, so encouraging. But here is where the picture changes sharply.
Future AI applications go far beyond text and voice. Augmented reality vision, where an AI overlay sits on top of what you see through smart glasses or a headset in real time, requires a minimum latency of under 30 milliseconds. The ideal target for smooth, consistent AR performance is under 10 milliseconds.
Of all 22 markets Ookla assessed globally, only Singapore currently meets even the minimum threshold for this use case, at 24.6 milliseconds. And even Singapore is still far from the sub-10 millisecond target needed for consistently good performance.
No other market in the world currently meets the minimum requirement for AR vision over 5G. Not the United States. Not South Korea. Not the United Kingdom. And certainly not Nigeria.
One of the more revealing findings in the report is what happens to latency when networks get busy. Under normal, low-traffic conditions, many 5G networks perform reasonably well. But under load, performance degrades sharply and unevenly.
Ookla found that degradation ratios, measuring how much worse latency gets under pressure compared to normal conditions, run from 3.7 times worse to 11.4 times worse across different markets. That is a wide and unpredictable range.
The gap between operators within a single market can be just as wide as the gap between different countries entirely. This means that even within the same city, two people on different networks could have very different AI experiences under the same conditions.
Don't Miss This Article
Nigeria launched commercial 5G services in August 2022 and has been gradually expanding coverage since, primarily in major urban centres. As of mid-2026, 5G penetration in Nigeria sits at 4.34 percent of mobile connections, still a small fraction of the overall subscriber base, with 4G dominating at 54.41 percent.
Nigeria was not among the 22 markets included in Ookla’s assessment, which focused on more mature 5G markets. But the findings carry a direct implication for Nigeria’s 5G story. If the world’s most advanced 5G networks cannot yet deliver the latency needed for AR vision and next-generation AI, Nigeria’s relatively early-stage 5G rollout is even further from that target.
This is not a reason for alarm. Text-based AI tools, which are what most Nigerians using AI today are actually interacting with, work just fine on 4G and early 5G networks. The gap only becomes significant when the conversation turns to what AI will be able to do in three to five years, not what it can do right now.
Ookla argued that the industry needs to rethink how it measures network performance for the AI era. Download speed, the traditional measure that most people associate with network quality, is becoming less relevant for AI readiness. What matters more now is upload capacity, the time it takes data to reach cloud servers, and the path that data takes to get there.
This shift in what counts as a “good” network has implications for how operators invest in their infrastructure and how regulators think about spectrum allocation and network quality standards going forward.
The headline takeaway from Ookla’s report is straightforward. The 5G era was built for streaming, video calls, and the internet as we knew it when 5G was being designed. The AI applications being built right now, and especially the ones coming in the next few years, are pushing well beyond what current 5G infrastructure was designed to handle. Closing that gap will require a new round of serious network investment, smarter architecture, and in some cases, entirely new approaches to how mobile networks are built and measured.
