With both Verizon and AT&T planning major 5G announcements for the 2019 CES, I wanted to speak directly with key technologists at both of the top U.S. carriers to really dig into their 5G network and device plans for the year. After a great chat with Verizon’s VP of network engineering, I enjoyed an equally illuminating interview today with AT&T VP of RAN and device design¬†Gordon Mansfield, who provided legitimate insights into the carrier’s 5G plans — and explanations for some of its recently controversial marketing decisions.

The executive takeaways from our interview include a few surprises: There’s more to AT&T’s confusing 5G Evolution branding than was initially thought; those two upcoming Samsung phones might not be as different in cellular performance as some people assume; and the company is apparently using its steep initial 5G data pricing to see what the market will bear. We also discussed 5G VR, AR, and other topics.

Here’s a lightly edited transcript of our discussion.

Gordon Mansfield:¬†Just for background and context… I’m sure you’ve already heard that we launched our initial 3GPP 5G mobile networks at the end of last year, on the 21st. We’ll be continuing to expand those throughout 2019.

We’ve deployed a lot of our spectrum in the past year, so we announced 400 markets with what we call 5G Evolution, which is a lot of our spectrum in use, and that will continue to be built out through this year. A lot of the work we’re doing with FirstNet, getting that coverage expanded, and adding the new radios for that LTE build, is that those radios are all 5G NR standards ready. So when the software is ready to start upgrading those, then they also have an ability called dynamic coexisting, so I can simultaneously on the 5G radio serve a 5G or LTE customer.

VentureBeat: The announcement of 5G Evolution has caused quite a stir over the past couple of weeks. You’re introducing some actual 5G hardware across the country, but also lighting up non-5G device screens with a “5G E” badge for 5G Evolution. For the customers who right now are seeing 5G Evolution appear on their phones, is what they’re seeing an indication of a tower that’s going to flip (via software) from saying “5G E” to “5G”?

Mansfield: Over time, all of those towers that are 5G E will ultimately have 5G capabilities. A lot of that, the timing at which that flip happens, there’s a lot of dynamics — how much capacity do you have, what software capabilities are available, but we very much will. Where we’ve deployed that infrastructure to support it, a lot of it is software upgradeable to support 5G. But you have to worry about how many (5G) handsets, how many users you have, you have to continue to serve a base of customers that have current handsets.

VentureBeat: When we see a phone today that has a 5G E indication on it, what spectrum is it actually using?

Mansfield: It’s actually using multiple bands. It uses our 700 MHz spectrum, 1900 spectrum, AWS spectrum, WCS 2300 spectrum, and some more bands — 850 is even used. When you get that indicator, it’s not one band or one frequency. We’ve deployed multiple radios at that site, along with the appropriate backhaul, along with all the capabilities that allow us to bring all of that together.

When you see that indicator, there are at least three radios being aggregated together, and more often than not, four. We don’t turn that on until the area has three or more carriers with a lot of other advanced capabilities on those radios that enable the phone to get those final speeds.

VentureBeat: You’ve already announced that your 5G network is going to be supporting sub-6 GHz frequencies. What are the actual frequencies you’re using for sub-6 GHz 5G service — not the low-band and upper low-band frequencies we just discussed?

Mansfield: Yes, it is — those are all sub-6 Ghz frequencies.

VentureBeat: So you’re referring to sub-6 Ghz as anything from 700 to 2100 or 2300 MHz?

Mansfield: Our current holdings for sub-6 range from 700 MHz up to 2300 MHz or 2.3 GHz. Certainly with CBRS coming around, we’ve started to do some work with CBRS at 3.5 GHz for fixed wireless. At some point that spectrum could also be used for 5G. All of the hardware we’ve deployed from day one is 5G NR capable.

VentureBeat: Is the plan to deploy this up-to-2.3 GHz spectrum as the sub-6 “5G” service, and then millimeter wave will only hit relatively high traffic locations as “5G+”?

Mansfield: Yeah. The way you can think about that is that millimeter wave, as has been talked about, is short distance but high bandwidth with multiple gigabit speeds. Today, millimeter wave is coupled with the sub-6 frequencies on LTE. Over time, you can pivot those same radios that are sub-6 on LTE into an 5G NR waveform and can combine those. You can aggregate the very high bandwidth of millimeter wave with those sub-6 frequencies, allowing you a very seamless mobility, because sub-6 gives you a very nice blanket. When you’re in those zones with millimeter wave radios, it adds that and just throws your throughput way up.

VentureBeat: What are you expecting to see for throughput for sub-6 5G?

Mansfield: It always comes down to how much spectrum you have. The reason the mmWave throws off those multi-gig speeds is that you can take 400-800MHz chunks of spectrum. When you get down to sub-6, no one holds more than a couple hundred megahertz. With our current holdings, you’ll get a lift on those frequencies over what you can do on LTE, but it’s not a significant lift.

Today we’re deploying all of this bandwidth on LTE and getting those high speeds, they’re not going to magically have higher speeds on 5G. It all comes down to the bandwidth. So if someone only applies a small channel, it could be lower speed than what you now get.

VentureBeat: What speeds should consumers actually expect from AT&T’s 5G service, now that you’ve had a chance to test it with Netgear’s 5G Nighthawk?

Mansfield: A lot of reporting says that the Nighthawk isn’t initially performing better than regular LTE. The fact of the matter is, go back a few generations, and the same thing is true — there’s a starting point from which you evolve. We stated very clearly and didn’t hide from the fact that we were starting with a 100 MHz carrier. There’s a lot of reasons, but there’s a lot of complexity with everything we’re doing with mmWave.

That will change very rapidly. I can tell you that month over month, over the next couple of months, we will see a step. We’ve got things laid out of new functionality; we expect to software upgrade each of device and infrastructure on a fairly regular cadence. We’ll have step functions of speed improvements, I expect, month over month improvements just in the first quarter alone. Certainly by the end of the first quarter, we’ll be peaking well into over a gig. I’ll leave it at that for now.

VentureBeat: Any news on devices?

Mansfield: It’s been widely reported that we intend to launch a Samsung smartphone in the first half of 2019…

VentureBeat: Two Samsung smartphones

Mansfield: The first one, just to be very clear, is a millimeter wave. And the second one includes sub-6 (5G).

VentureBeat: The knee-jerk reaction from the media was, “why should anybody consider buying the first Samsung smartphone if there’s a second one coming out only months later that has broader network support?” What’s AT&T’s response to that?

Mansfield: The fact is, by that logic, you’d never buy a new phone. We have new phones that come out every year with advanced capabilities. Fact. Every year.

VentureBeat: This is a big jump, though. To say to someone, listen, we’re going to have 5G in two different flavors — 5G and 5G+ — but you’re only going to be able to connect to our “stadium 5G” with phone number one, and your “walkaround 5G” is going to be the one coming later…

Mansfield: I’d say that it’s not. The reason it’s not is, your biggest gain in throughput performance is going to be with millimeter wave. So the fact is, with millimeter wave, and combining it with LTE technology on sub-6, and being able to take those sub-6 carriers and simultaneously, you get the benefit immediately of those sub-6 frequencies using an LTE waveform, adding onto it a mmWave 5G NR waveform. You get that experience today.

Now, at the end of the year when you get NR on that (second phone), quite frankly, most consumers… will that device have longer legs forward? Absolutely. But will you get a very similar experience on that device as you do on the device earlier in the year? Yes.

From an evolution of technology, nothing happens flip of the switch. Everything builds on top of everything else. So by us pushing the core to the edge, a lot of the latency benefits that you get, guess what — our existing network is getting the benefit of that right now.

VentureBeat: What latencies are you seeing with 4G LTE on optimized equipment?

Mansfield: We’re typically getting high 20s, low 30s (milliseconds).

VentureBeat: What would you expect with 5G, and millimeter wave specifically?

Mansfield: Initially it’s very similar, a little less, like a couple of milliseconds less. But over time, I expect it to come down sub-20.

VentureBeat: Sub-20 milliseconds is OK, just enough to maybe do VR and AR, but the whole promise of the 5G standard was supposed to be under 10, maybe under 5 milliseconds?

Mansfield: The numbers I’m quoting are wide area. If you look at places where you’d want to do a lot of AR/VR — people aren’t going to want to do AR/VR going 70MPH down the highway. I hope not…

VentureBeat: Give it a little bit of time.

Mansfield: But if you go into stadiums, like AT&T Stadium, there’s things I can do using edge compute capabilities, where the traffic you want to use from an AR/VR perspective never leaves the stadium. That, I can put those latencies at sub-5 (milliseconds). There will be pockets where you’ve absolutely optimized the network for a specific experience, that will be single-digit latencies. But when you think about the wide area, to start, sub-20 (milliseconds), and it comes down to how far out have you pushed the (network’s) edge. There has to be a business case, right?

VentureBeat: An average consumer might consider buying a wireless VR headset they can put on at home. I don’t know that that person is going to want to go to a stadium and put on a VR headset. The whole reason you’ve gone to that stadium is to experience a live event. Granted, that person might put on a pair of glasses temporarily to see action closer, or use AR overlays to track stats, but that seems years away.

Mansfield: With 5G, you can have different applications with different latency requirements. Like using AR in a manufacturing environment… You can go and build scenarios that are very special purpose. To get to real single-digit latencies, that’s important to understand. That will come first. Over time you’ll build the infrastructure further and further out that will allow it to come down on a broader basis for consumers. I think it’s important for people to realize that you can’t make an investment where everything happens everywhere at one time. It’s just not practical.

VentureBeat: Industrial AR has been going on for the last couple of years, and it seems like everyone who enters AR with the pipe dream of a consumer product winds up eventually doing enterprise because nobody’s willing to pay $2,000 for a headset to wear at home to interact with aliens through their walls.

Mansfield: A lot of those industrial AR applications, though, are wired, if they’re truly really good with granular abilities…

VentureBeat: Even HoloLens?

Mansfield: HoloLens is very limited in what it can do. What we’re talking about here is complete 3D you can look around. It’s very different than what HoloLens does. With 5G and edge compute, you now can do that wireless. That’s the big deal in enterprise with 5G, you can take things like robotics, manufacturing line, that can take days to reconfigure. If you can do all of that wirelessly, and then also have the corresponding ancillary processes and capabilities that come with it, it starts to open up a tremendous opportunity.

VentureBeat: Why would a company thinking of doing AR headsets en masse across the workforce go with 5G rather than just wait for Wi-Fi 6 or some other high-bandwidth solution that doesn’t require paying a per-headset subscription fee to AT&T or somebody else?

Mansfield: Some people have been trying to do Wi-Fi-based systems for the better part of a decade or more. Sure, they can get bandwidth and low latency, but where they struggle and fail every time is jitter. Jitter is not their friend. The latest and greatest Wi-Fi systems, with better throughput and with lower latency, take a cellular system — even on LTE, with localized edge compute, and slightly higher latency, the throughput’s pretty similar. But much better jitter.

Think about drones. Jitter matters. On Wi-Fi, that drone crashes into the wall 9 out of 10 times. On cellular, it never did. The difference is jitter. It’s network instability. When you get into totally unlicensed, uncontrolled areas, it is very hard to maintain a consistent jitter factor. And it’s also hard in a noisy environment. Even on the same spectrum, we use LAA on 5GHz spectrum, as Wi-Fi uses, from a performance perspective, LAA outperforms Wi-Fi every time. Why? You’re not dropping as many packets. The 3GPP waveforms deal with those changes in environments because it’s a mobile system.

VentureBeat: On 5G pricing, currently AT&T is offering 15GB of 5G data for $70 a month. Every company I have spoken with has said that unlimited is the way 5G needs to go over time, as consumers can’t even fathom the amount of data that will be consumed by VR, AR, automotive, and other applications. What is the long-term pricing strategy here for AT&T’s mobile 5G service, and why go with a 15GB data cap?

Mansfield: I’m the technologist, and I don’t set the prices. But I can say, we will — the industry will set the price, consumers will set the price. We set initial pricing, and our marketing partners took data they look at and set that pricing. We’ll see what happens over time.