Cisco : Has Internet growth peaked?

Cisco Blog >SP360: Service Provider

Will the Internet continue to grow at historical rates, or has growth peaked? This was the question was posed by Sanjay Patel of CableLabs to the panelists on the Bandwidth Consumption Trends panel during the August 2016 CableLabs Summer conference. I was pleased to have been invited to participate in the panel, alongside Rob Kenny of Communications Chambers and Damian Poltz of Shaw Communications.

The panel presented a surprising diversity of perspectives on the future of Internet growth. One audience member commented to me that listening to the panel made him feel the future of the Internet is still very much unknown. I had the same impression, and it makes me happy to think the Internet still has surprises in store for us.

While I believe all the panelists have differing perspectives on long-term growth, there were no flat-out contradictions. In fact, over the course of numerous discussion we concluded that our numbers can all be reconciled with each other because they all are measuring different aspect of the near-term future of the Internet, even if our perspectives on long-term growth may differ.

I'd like to summarize here the answer each panelist has offered to Sanjay's question.

Sanjay Patel, Principal Strategist at CableLabs. In additional to framing the question for the panelists, Sanjay also presented recent numbers from CableLabs' edge network traffic measurements, where traffic growth does appear to have leveled off in the past year. Sanjay's answer to the question was that growth may be flattening in the near-term.

Rob Kenny, Founder at Communication Chambers.Rob's answer to the question was that traffic may continue to grow at historical rates, but the need for broadband speeds will not. Rob has done extensive mathematical modeling to determine that it's fairly uncommon for multiple high-bandwidth applications such as HD video to be in use simultaneously within a household. And apart from console game and movie downloads, there aren't too many applications that require high speeds. As such, he projects that for all applications expected to be on the network with the next few years, the need for high bandwidth is not as high as many expect. For the UK his numbers show a median bandwidth demand of 15 Mbps in 2020 and 18 Mbps in 2025, up from 12 Mbps today. The top 1% of households will require 49 Mbps in 2025. Admirably, Rob has made his report and entire model available to the public.

Arielle Sumits, Senior Analyst at Cisco.My own answer to the question was that traffic growth will likely taper slightly, but the Internet is not saturated and growth will hold steady in the long-run. At the end of my presentation, I mentioned that in the 12 minutes I had been speaking, 15 million gigabytes had crossed the network, but 4 billion gigabytes of data had been created. Less than 1% of data created today is transmitted onto the Internet, partly because we do not currently have the analytical capabilities to process this much data. AI is entering a period of exponential progress, and as it does, it will make sense to move more data onto the network where it can be accessible for analysis. We are nowhere near a saturation point. That's the long-term view, but in the short-term we are also seeing persistent growth. There is a slight tapering in year-over-year growth rates, but much lower than we'd expect given the high volumes of traffic the network carries today. Even without substantial data from machines, there are substantial broadcast viewing hours and offline gaming hours that will make their way onto the Internet in the next five years. Regional and country level details are available on the Cisco VNI site.

Damian Poltz, Director of Technology Strategy at Shaw Communications.Damian's answer to the question was an enthusiastic affirmation that peak traffic growth will hold at historical rates. He pointed out that historical forecasts of Internet growth have almost always turned out to be low, and if one had simply projected constant growth over the past 15 years, their forecast would have been right on target. 'Don't bet against the Internet' was Damian's theme and rightly so. Damian added that predicting constant growth is the more conservative option for service providers, because it's a lot easier to dial down infrastructure investment than it is to dial it up. As a side note, Damian mentioned that his recent numbers show a dip in traffic due to Netflix's re-encoding of their video assets.

Each panelist measured different aspects of the network - Damian's focus was on peak traffic, mine was on average traffic, Rob's was on broadband speeds. When we compared numbers, we felt each of our forecasts could be made compatible with those of our fellow panelists. But in the end, I think we have differences of opinion regarding the long-term prospects for traffic growth. And the differences lie in the impact of applications that are only a blip on the traffic radar right now, such as virtual reality, cloud gaming, and IoT. I can't wait to see what a 'bandwidth consumption trends' panel looks like in 2025.