Ookla recently published an article that looked at speed test and coverage characteristics of FWA cellular broadband provided by AT&T, T-Mobile, and Verizon. FWA is always an interesting topic since the three companies have collectively attracted over 17 million broadband customers to the new business line. There were some interesting findings from Ookla worth discussing.

Ookla noted that 70% of FWA speed tests come from urban areas. That doesn’t seem surprising to me since that is where most people live. While FWA was originally touted as a rural technology, the fact that it’s marketed at prices below other broadband technologies makes it very attractive in urban areas.

Ookla notes that T-Mobile has the fastest download speeds, with a recent median speed of 222.7 Mbps, which is 38.4% faster than AT&T and 76% higher than Verizon speeds. I think Ookla missed the likely reason for this, which is that AT&T and Verizon both have legacy telco bases and have legacy customers. T-Mobile has no legacy business and likely sells FWA only to customers within a few miles of cell towers. The other two companies are using cellular technology as the alternative that allows them to tear down copper lines. They will offer this technology far away from towers, even if the speeds are exceeding slow, as long as a connection will support a voice signal – which is the regulatory requirement they are supposed to meet to justify discontinuing copper. Both companies also have a significant base of legacy cellular hotspot customers who can be far away from a tower getting much slower speeds. These hotspots are functionally equivalent to FWA, but are marketed differently, with small monthly data caps. When I’ve examined the speed tests for the three carriers for customers within a few miles of a cell site, they all deliver similar speeds.

Ookla also notes that T-Mobile’s Q2 2026 median upload speed of 18.1 Mbps was 80.1% higher than AT&T and 48.5% higher than Verizon – for many of the same reasons. The really interesting finding was that less than 40% of FWS speed tests in 48 states are exceeding the FCC’s definition of broadband at 100/20 Mbps. I assume that most of those that don’t meet that standard are doing so because of upload speed. This goes a long way towards explaining why the current FCC is not willing to consider increasing the definition of broadband above 100/20 Mbps, since that might declare that FWA, satellite service, and some WISPs are not really broadband per the FCC definition. The current FCC seems hellbent on being able to claim that satellite broadband has solved the rural broadband gap.

All three companies saw slower speeds in the second quarter of 2026 compared to the first quarter. Ookla likely correctly pinned this on foliage. It’s easy to forget when we discuss cellular coverage that it is weakened by foliage just like every other wireless technology. It raises the interesting question of whether a service is really broadband if it only meets the FCC definition of broadband in the winter?

Ookla recognized a huge 60% increase in AT&T broadband speeds comparing the second quarter of this year to the third quarter of 2025. AT&T has deployed 50 MHz of additional spectrum during that time that it purchased from EchoStar. The AT&T cellular speeds at my house more than doubled this past spring. This is a good reminder that the carriers are constantly taking steps to increase performance. They are engaged in a fierce marketing war, and none of them wants to look like they have an inferior network.

Finally, Ookla noted that rural FWA broadband speeds and latencies are lower than in urban areas for all three carriers. There are two explanations for this. First, not all rural cell towers are equipped to the same standard as urban towers. But this also comes back to the distance issue. In rural areas, customers are willing to try FWA even when it doesn’t have superfast speeds – because a 25 Mbps download on FWA beats the alternatives available. A household can often add FWA for $35 per month to an existing cellphone plan, which is considerably cheaper than Starlink at $130.