Cellular Broadband Speeds – 2019

Opensignal recently released their latest report on worldwide cellular data speeds. The company examined over 139 billion cellphone connections in 87 countries in creating this latest report.

South Korea continues to have the fastest cellular coverage in the world with an average download speed of 52.4 Mbps. Norway is second at 48.2 Mbps and Canada third at 42.5 Mbps. The US was far down the list in 30th place with an average download speed of 21.3 Mbps. Our other neighbor Mexico had an average download speed of 14.9 Mbps. At the bottom of the list are Iraq (1.6 Mbps), Algeria (2.1 Mbps) and Nepal (4.4 Mbps). Note that these average speeds represent all types of cellular data connections including 2G and 3G.

Cellular broadband speeds have been improving raoidly in most countries. For instance, in the 2017 report, Opensignal showed South Korea at 37.5 Mbps and Norway at 34.8 Mbps. The US in 2017 was in 36th place at only 12.5 Mbps.

Earlier this year Opensignal released their detailed report about the state of mobile broadband in the United States. This report looks at speeds by carrier and also by major metropolitan area. The US cellular carriers have made big strides just since 2017. The following table compares download speeds for 4G LTE by US carrier for 2017 and 2019.

2019 2017
Download Latency Download Latency
AT&T 17.8 Mbps 57.8 ms 12.9 Mbps 63.8 ms
Sprint 13.9 Mbps 70.0 ms 9.8 Mbps 70.1 ms
T-Mobile 21.1 Mbps 60.6 ms 17.5 Mbps 62.8 ms
Verizon 20.9 Mbps 62.6 ms 14.9 Mbps 67.3 ms

Speeds are up across the board. Sprint increased speeds over the two years by 40%. Latency for 4G is still relatively high. For comparison, fiber-to-the-home networks have latency in the range of 10 ms and coaxial cable networks have latency between 25 – 40 ms. The poor latency in cellular networks is one of the reasons why browsing the web on a cellphone seems so slow. (the other reason is that cellphone browsers focus on graphics rather than speed).

Cellular upload speeds are still slow. In the 2019 tests, the average upload speeds were AT&T (4.6 Mbps), Sprint (2.4 Mbps), T-Mobile (6.7 Mbps) and Verizon (7.0 Mbps).

Speeds vary widely by carrier and city. The fastest cellular broadband market identified in the 2019 tests was T-Mobile in Grand Rapids, Michigan with an average 4G speed of 38.3 Mbps. The fastest upload speed was provided by Verizon in New York City at 12.5 Mbps. Speeds vary by market for several reasons. First, the carriers don’t deploy the same spectrum everywhere in the US, so some markets have less spectrum than others. Markets vary in speed due to the state of upgrades – at any given time cell sites are at different levels of software and hardware upgrades. Finally, markets also vary by cell tower density and markets that serve more customers for each tower are likely to be slower.

Many people routinely take speed tests for their home landline broadband connection. If you’ve not taken a cellular speed test it’s an interesting experience. I’ve always found that speeds vary significantly with each speed test, even when run back-to-back As I was writing this blog I took several speed tests that varied in download speeds between 12 Mbps and 23 Mbps (I use AT&T). My upload speeds also varied with a top speed of 3 Mbps, and one test that couldn’t maintain the upload connection and measured 0.1 Mbps on the test. While landlines broadband connections maintain a steady connection to an ISP, a cellphone establishes a new connection every time you try to download and speeds can vary depending upon the cell site and the channel your phone connects to and the overall traffic at the cell site at the time of connection. Cellular speeds can also be affected by temperature, precipitation and all of those factors that make wireless coverage a bit squirrelly.

It’s going to be a few years until we see any impact on the speed test results from 5G. As you can see by comparing to other countries, the US still has a long way to go to bring 4G networks up to snuff. One of the most interesting aspects of 5G is that speed tests might lose some of their importance. With frequency slicing, a cell site will size a data channel to meet a specific customer need. Somebody downloading a large software update should be assigned a bigger data channel with 5G than somebody who’s just keeping up with sports scores. It will be interesting to see how Opensignal accounts for data slicing.