5G is a multi-faceted technology that cannot be described easily. While there are many misconceptions about 5G, which I covered last year and debunked at length, now we will dive into the technology itself: what is possible, what is necessary and what is reasonable to expect from the technology. In the last year, we have seen a lot of companies make bold claims. And while the industry has moved forward, in many ways, the perceptions around the technology have not. This article will be split into a two part series—first, we’ll take a look at the current state of 5G networks and modems.
2020 ‘Quick’ 5G Review
2020 was an interesting, big year for 5G. AT&T launched its small footprint, low-band network in November 2019, followed a month later with the launch of T-Mobile’s 5G Nationwide 600 MHz 5G network. T-Mobile followed that up with the world’s first Standalone 5G network in August of 2020, and Verizon announced a mmWave footprint increase the following month. 2020 also saw the finalization of T-Mobile’s acquisition of Sprint and the subsequent merging of T-Mobile’s 600 MHz spectrum with Sprint’s 2.5 GHz. Towards the end of the year, in conjunction with the launch of Apple’s first 5G phone, the iPhone 12, Verizon launched its nationwide 5G network leveraging DSS (dynamic spectrum sharing) to improve 5G coverage. While this deployment was initially met with a lot of criticism for its slow speeds, Verizon quickly pulled its average 5G speeds down much closer to AT&T and T-Mobile’s.
Current State of Networks
Currently, in the US, all three major operators have some form of low-band and high-band (mmWave) spectrum. The only carrier with ample mid-band spectrum is T-Mobile, with an average of 150 MHz of spectrum in most markets. Here in San Diego, T-Mobile has almost 200 MHz of spectrum. This is crucial because 60 MHz has already been shown to yield almost 1 Gbps of throughput when combined with LTE in NYC. It’s these speeds and the capacity of mid-band holdings that AT&T and Verizon are going after in the C-Band auction.
The merger of T-Mobile and Sprint resulted in T-Mobile’s ownership of the largest 5G spectrum holdings of any carrier. This sparked a frenzied C-Band Auction. To date, we know that AT&T and Verizon, along with Dish, Comcast, Spectrum, T-Mobile and others, have already spent upwards of $81 billion on spectrum. The auction winners, listed in order of most spent to least spent are:
- Verizon—$45.455 billion
- AT&T—$23.407 billion
- T-Mobile—$9.336 billion
- US Cellular— $1.283 billion
Light Reading’s Mike Dano wrote a very good break down of spectrum costs per MHz and total spectrum holdings after the auction. Additionally, my colleague Will Townsend also shared some helpful insights about the auction on RCR Wireless. Ultimately, after clearing costs are settled, Verizon will pay $53 billion for its huge chunk of A-Block spectrum, which should become available by the end of this year. T-Mobile, since it already had lots of 2.5 GHz mid-band, only bid on less prime B and C-Block spectrum, which won’t clear until 2023. Verizon has clearly indicated it believes that mmWave and low-band DSS alone won’t cut it—as such, it is willing to spend itself out of its current network. While it remains unclear how long it will take Verizon and AT&T to roll out their newly acquired C-Band spectrum, it will likely cost tens of billions due to the propagation characteristics of the 3.7 GHz to 4 GHz bands (C-Band). Verizon and AT&T will need to further densify their networks to ensure acceptable coverage, and they will likely need to bring new sites online—a potentially difficult task given the way some municipalities view 5G.
A complete 5G network needs to be standalone, implementing low, mid, and high-band spectrum together as one. Right now, there is much confusion about bands, spectrum, and such, but in the long term, most consumers will be completely unaware of what flavor of 5G they are using because they will likely be using all three of them. 4G mostly relied on low-band spectrum, spanning 700 MHz to 2100 MHz. Meanwhile, 5G spans from 600 MHz to 47 GHz—a much broader array of spectrum with different characteristics.
Mid-band is the goldilocks band of 5G—it can carry much further than mmWave and has far more capacity and higher speeds than low-band spectrum. Right now, around the world, most operators either have low-band and lack mid-band or they have mid-band and lack low-band due to limited spectrum availability. Realistically, you want both to have a comprehensive 5G network with the proper coverage and speeds that utilizes multiple bands. Additionally, aggregating low-band with mid-band spectrum improves the coverage of mid-band spectrum since the uplink/upload is the weakest link in a connection—if you can offload the uplink to low-band you can improve higher band coverage. Eventually, carriers will need mmWave (high-band) to alleviate congestion in dense areas like stadiums, malls and other places where people are packed in tightly. That said, mmWave is less of a concern right now since many high-density areas are at limited capacity and many people are staying home due to Covid-19. Additionally, mmWave could help deliver affordable gigabit broadband to the home or small business via fixed broadband (though this will likely take time for operators to deploy, and they probably won’t tackle it right away). It is not nor will it ever be realistic to expect mmWave to have the same coverage as current 4G bands. It literally defies the laws of physics.
As of right now, I see mmWave as a line-of-sight technology. I have personally tested Verizon’s mmWave 5G network and while earlier deployments and devices were not stellar, I have seen Verizon’s mmWave effective to at least 500 feet while delivering speeds upwards of 1 Gbps. I also regularly saw speeds of 2 Gbps at half that distance—surprisingly, I even saw speeds of 2 Gbps inside of vehicles. In my experience, mmWave 5G drops off very quickly and the experience varies from device to device based on antenna configuration. However, modems and antennas are constantly changing, and I believe that the mmWave experience will get better with every generation. I am extremely excited to see what this technology will mean for venues, especially concert venues and sports stadiums. The amount of capacity and bandwidth that mmWave affords will be absolutely paradigm shifting for entertainment. With concerts by and large on hold, now is a good time for venue owners to upgrade their infrastructure to 5G mmWave if they can afford it. An upgrade will do much to prepare venues for the return of concert goers, while enabling new revenue opportunities that require mmWave’s reliable low-latency, high-bandwidth connectivity.
The model for other operators to emulate is T-Mobile’s aggressive rollout of 2.5 GHz and 600 MHz with carrier aggregation and Standalone 5G. Right now, T-Mobile’s 600 MHz low-band network serves as the coverage layer for its network while the 2.5 GHz band will serve as the capacity and bandwidth layer. I have already seen in my testing that T-Mobile’s mid-band is capable of download speeds of 500 to 600 Mbps. However, T-Mobile has not fully rolled out this band in my area. I expect that performance numbers could increase further with the appropriate devices capable of carrier aggregation between FDD and TDD bands. In fact, when combined with 4G and LAA (license assisted access) via ENDC (E-UTRAN New Radio – Dual Connectivity) T-Mobile’s mid and low band together were capable of delivering over 1 Gbps in the company’s strongest market, New York City.
Like Verizon, AT&T’s 5G rollout started with mmWave. The company even initially pitched 5G mmWave as a business-only application. However, the company quickly pivoted towards low-band 5G and hasn’t said much about mmWave outside of entertainment use cases in venues. AT&T was the second largest bidder in the C-Band auction which will give it the mid-band it needs to be competitive with T-Mobile. I found AT&T’s 5G network to be consistently quite fast, around 200 to 300 Mbps, but nowhere near as fast as Verizon’s mmWave or T-Mobile’s 2.5 GHz mid-band. That said, AT&T’s floor is higher than Verizon and T-Mobile, and as a result, the average speeds tend to be quite good.
These differences between AT&T, T-Mobile and Verizon are why all three of these companies claim 5G leadership in one way or another. Now these companies are shifting into the second phase of their rollouts—building capacity, improving speeds or improving coverage. As a result, we will see Verizon’s speeds drop as it improves coverage, and T-Mobile’s speeds increase as it rolls out 2.5 GHz in 2021. With more spectrum, AT&T’s low-band has proven to be faster than T-Mobile’s. This allowed it to start out with faster 5G than T-Mobile, at the cost of a smaller coverage area. In the next part in this series, we will go into more depth about what to expect in the future.
The state of 5G modems and RF
5G modems are constantly evolving, and the modems ultimately drive what the devices can do and which network features improve speeds, coverage, etc. While Huawei, MediaTek, Qualcomm and Samsung offer competing modem solutions, only Huawei and Samsung supply modems for their own devices. This leaves Qualcomm and MediaTek as the primary suppliers of 3rd party modems and even applies for companies like Intel who previously had its own modem division that it eventually sold to Apple. Apple was previously Intel’s sole modem customer, but left Intel for Qualcomm when the company could not deliver a competitive 5G solution on time. While Apple does not have its own modem in market and is currently a Qualcomm customer (X55), it has a major effort underway to build its own modems (starting with 5G).
Huawei is also in a complicated position regarding 5G modems as the company is unable to manufacture any of its leading-edge processors available through its chip subsidiary, HiSilicon. While the company is currently looking for a supplier of SoCs and modems that is not restricted by the US, the company still has a competitive modem in the Balong 5000. Huawei has been using the Balong 5000 for quite a while now, but it does support virtually everything you would need a modem to support at this time. HiSilicon claims 2x CA in Sub-6, with channels of up to 100 MHz and a maximum downlink of 4.6 Gbps but will need more advanced features in the future. For mmWave, HiSilicon claims a peak downlink of 6.5 Gbps and a combined 5G + 4G of 7.5 Gbps. The Balong 5000 also supports Standalone (SA) and Non-Standalone (NSA) 5G. The latest version of HiSilicon’s Balong 5000 is available integrated into the 5nm HiSilicon Kirin 9000, the latest SoC for Huawei’s flagship phones (like the Huawei Mate 40 Pro).
Samsung’s latest modem is the Exynos 5123, a discrete modem for Samsung’s own devices. Samsung generally pairs this modem with its own SoCs, as it did with this year’s Exynos 2100 SoC. What makes this year’s modem different is its integration inside of the SoC, which can save on power and overall board space. Currently, the Exynos 2100 and 5123 modem can be found in the Global version of the Galaxy S21 which is available outside of China, Hong Kong, Japan and the US. The Exynos 5123 supports both SA and NSA 5G, along with Sub-6 and mmWave and up to eight carrier aggregations in 5G and LTE. I believe that 5G will likely be on mmWave as most Sub-6 CA involves two carriers. Samsung claims a peak downlink of 7.35 Gbps over mmWave and 5.1 Gbps over Sub-6.
MediaTek has a combination of both discrete and integrated modems. The new 6nm Dimensity 1200 and 1100 5G smartphone SoCs integrate last year’s 7nm Helio M70 modem, which offers a peak throughput of 4.7 Gbps downlink, a 2.5 Gbps uplink and support for both SA, NSA 5G and carrier aggregation. The M70 will be offered as a discrete part as well, especially for certain large customers like Intel, who has partnered with MediaTek to use its 5G modems in its platforms this year. In addition to the M70 as a discrete and integrated part, MediaTek just announced the new Helio M80 modem. The new M80 brings support for mmWave in MediaTek’s 5G modems for the first time (something virtually all of MediaTek’s competitors have had for a while) and 3GPP Rel.16 support. The M80 also brings carrier aggregation between Sub-6 and mmWave bands, giving it peak throughput up to 7.67 Gbps in the downlink and 3.76 Gbps in the uplink. MediaTek says the M80 is expected to sample to customers later this year, though there is no concrete date for when we’ll see it in devices. MediaTek also recently reported that it surpassed Qualcomm in Q3 of 2020 as the number one supplier of modems in the world—however, the difference between MediaTek and Qualcomm is very close. The rankings will likely change in Q4 once those numbers roll in. Regardless, MediaTek is clearly growing and catching up with Qualcomm.
And what of Qualcomm’s offerings? Qualcomm has successfully led the 4G and 5G charge, and its lineup offers one of the most diverse selection of modems available. Qualcomm is a systems solutions company, so it rarely makes just one chip for an application. The company’s approach to 5G is to call the entire wireless part of a smartphone the “modem-RF-system”—implying that its 5G wireless solutions are more than just modems. This is a departure from the past where Qualcomm was mostly focused on modems. Now, due to the complexity of 5G and 4G bands, the company has adopted an even more integrated approach.
Qualcomm’s first 5G modem, the X50, was a very short-lived offering. While it enabled some of the first 5G phone designs, it was not really designed to be a long-term chipset. Qualcomm’s current lineup starts with the flagship X65, the company’s first 4nm modem. Available in commercial devices by the end of the year, this discrete 3GPP Rel.16 modem is also the world’s first 5G modem capable of 10 Gbps. The X62, announced alongside the X65, shares the 4nm process and the 3GPP Rel. 16 capability. It also supports mmWave and sub-6 aggregation but is only capable of speeds up to 4.4 Gbps. I expect that the X62 will be the mainstream or entry-level modem to support all of the latest 3GPP Rel.16 features and will be popular for IoT applications
Last year’s Snapdragon X60 will likely be in most flagship Android 5G phones this year. Shipping inside Qualcomm’s flagship Snapdragon 888 smartphone chipset, the X60 is 3GPP Rel. 15 compliant and was the company’s first 5nm modem. The X60 has a peak download of 7.5 Gbps and an upload of up to 3 Gbps. For 4G, the X60 features download speeds of up to 2.6 Gbps and 4G LTE upload speeds of 316 Mbps.
Another modem in the lineup, the X55, appears to have the same level of performance as the X60 but lacks carrier aggregation across different TDD, FDD and mmWave bands of 5G. The X60 is an example of the new multi-band approach to improving speeds and the early 5G user experience in general. The X55 pre-dates the Snapdragon X60 by a year, the X55 is an immensely popular modem and was found in most 5G flagship phones launched in 2020. The X55 is found inside many different platforms, including PCs and 5G hotspots. One of the things that makes the X55 so prolific is that it is also the modem that Apple uses to deliver 5G connectivity to all versions of its iPhone 12. Additionally, thanks to the X55, all iterations of the iPhone 12 in the US have Apple-designed mmWave antennas.
In addition to the X65, X60 and X55 high-end 5G modems, Qualcomm also has the X52 and X51 5G modems, which it ships inside of its mid-tier and entry-level SoCs. These modems have reduced throughput compared to the X60 and X55, but both still support mmWave. The mid-tier, X52, has a peak download of 3.7 Gbps and peak upload of 1.6 Gbps while the entry-level X51 has a peak download of 2.5 Gbps and upload of 900 Gbps.
All of this to say, Qualcomm’s lineup is the most comprehensive on the market, offering both integrated and discrete 5G modems for all price tiers. Qualcomm’s efforts are not limited to modems—its RF Front-end (RFFE) business operates at over $1 billion a quarter. Qualcomm couples its RFFE offerings tightly with its modems to maximize performance and battery life. Qualcomm’s RFFE business includes a series of antenna and filter modules built for smartphones and other devices. When Qualcomm announced the Snapdragon X60 last year, it also announced the QTM535, the company’s third generation mmWave antenna module, which is even more compact than the second-generation module. Qualcomm also makes 5G filters and duplexers for different 5G bands and for applications including industrial and automotive (in addition to smartphones).
Qorvo is one of the world’s leading suppliers of RF components, with dozens of 5G products between its infrastructure and mobile businesses. While Qorvo has new products to address the new 3.3 to 4.2 GHz bands being used for 5G today, the company does not have any mmWave mobile modules or antennas—Qorvo’s only mmWave 5G offerings are on the infrastructure side. That said, the company does offer a front-end module for bands n41, n77, n78 and n79.
Skyworks also offers many 5G front-end modules and switches that combine support for 3G, 4G and 5G. The company has Sky5, Sky5 Ultra and Sky5 LiTE offerings which combine front-end modules, power amplifiers and switches for different 5G applications. Like Qorvo, Skyworks does not offer any mobile mmWave 5G options yet, but either company could reach that point eventually. That said, both Qorvo and Skyworks are still supplying many of the front-end components for many of the non-mmWave 5G phones out in the market today.
Murata is a large component supplier in the mobile industry, but also has a multitude of RF front-end components in market for 5G—12 different product categories within the smartphone business alone. Murata also has 14 different product categories for cellular RF, 70 different SAW Duplexers and 175 different SAW filters. While Murata does not currently have any mmWave products shipping, it did demonstrate a capability at MWC LA 2019 in partnership with its subsidiary, pSemi. Murata co-designed an mmWave 5G antenna-integrated module that supports 28 GHz. Purportedly, the company is working on IC front-ends, packages and antennas, as well as 39 GHz mmWave ICs.
Apple’s position in this space is interesting and compelling. After Apple left it 5G modem partnership with Intel to join one with Qualcomm, Intel’s only choice was to shutter or sell off the division. Intel picked the latter, selling the division to Apple for a meager $1 billion, including the IP—Apple’s actual reason for the purchase (I covered this in July of 2019 when it happened). We can probably expect Apple to have its own modem around 2025, given license and supply agreements with Qualcomm. Apple could integrate this modem into its SoCs to further reduce cost and power consumption, as other players like Huawei, MediaTek, Qualcomm and Samsung have already done. I believe that one of Apple’s biggest motivations to to go this route is the potential to amortize the costs of modem development and drive increased cloud and services usage. I said as much in more detail in my blog about Apple’s moves in 2020.
While Apple is currently a Qualcomm customer and licensee, I expect that will end once it has its own modem products. The two companies will likely re-negotiate their licensing agreement shortly thereafter. All of this noted, Apple is not necessarily waiting to develop and use 5G components until it has its own modems. In fact, multiple tear-downs confirm that the mmWave 5G antenna modules in the new US iPhone 12 models are actually an Apple in-house design manufactured by USI. The fact that Apple is taking on one of the hardest problems in 5G head-on shows that the company is willing to commit the resources to push the technology forward. Apple effectively still used Qualcomm for the modem, transceivers and envelope tracking in first 5G phone, the iPhone 12, along with Skyworks and Murata for many of the FEMs (front-end modules).
In Part 2 of this series, we will cover the state of 5G devices and standards and what to expect in 2021 and beyond. Stay tuned.
Disclosure: Moor Insights & Strategy, like all research and analyst firms, provides or has provided paid research, analysis, advising, or consulting to many high-tech companies in the industry. The author holds no investment positions with any of the companies named in this article.