In our post from last week out of CES, we touched on the subject of 5G Modems, and we want to expand on that here. The market for smartphone modems for 5G looks a lot like the market for 4G modems at the same point in the launch of that standard, which looks a lot like the market for 3G modems. However, there are some big questions about where the market will go from here.
A quick recap. Modems, (aka basebands) are the crucial chip that allow mobile phones to connect to the cellular networks operated by the telecom carriers. Without these chips, devices have to rely on local Wi-Fi networks to connect.
The early days of a new standard always face the same 3-way problem (hen-rooster-egg?). The operators cannot launch network service until devices (i.e. phones) are ready. But the device makers cannot release devices until they have modems. And the modem makers cannot release their chips until they get approved by operators. This was a big problem with 3G as the networks were ready a year before the devices came on the market. The industry was in much better shape with 4G, with everything coming together pretty much on time. In this respect, 5G is looking pretty good.
The carriers are starting to roll out “5G” networks this year, and since these are really just enhanced 4G networks, the rest of the ecosystem is somewhat in place already. Qualcomm demonstrated its first 5G modem over two years ago, and we believe it is already commercially available (As a reminder, we have no confidential information, this is just our guess based on public comments.)
The roll-out for 5G will take many years, which greatly smooths out the problems above. However, there are some wrinkles.
First, there is only one commercial, merchant 5G modem vendor today – Qualcomm. Again, this is pretty standard, as they invest a lot in ‘helping’ to develop the standards and have historically been the first to market. In December, Mediatek announced that it is sampling its 5G modem, with commercial shipments “in 2019”. Setting aside the squishiness in that timing, this a big advance for Mediatek. Historically, they usually lag the new standard by a a year or two. Samsung has announced its 5G modem and Huawei looks likely to have its solution available soon.
Which of course begs the perennial question in mobile – “What about Apple?”. Apple cannot use Huawei’s or Samsung’s for competitive reasons (to greatly oversimplify it). Mediatek has been trying to break into Apple for years, but we have no reason to think that has changed. And of course, Apple and Qualcomm are engaged in lawsuits in seemingly every jurisdiction from here to Djbouti. Apple has been relying on Intel as a bulwark against Qualcomm. And Intel has announced that its 5G modem will be ready “in 2019”.
As we head into Mobile World Congress in February, the big question will be exactly where all these modems stand in actual readiness. This part of the industry is notorious for announcements that… let’s call it… put expectation ahead of reality. A key test will be to visit the vendors’ booth and see who is actually displaying chips and who is deploying a rack-sized demo board that just happens to not be working at this moment. From what we have heard, this year is no different with a lot of questions about availability of already press-released modems from a few of the names on that list above.
Again, all of this is a bit muddied by a true definition of 5G. We will not dig into all the features here, but another good test at MWC will be to ask what 5G features are supported on each chip and which are not quite available yet.
There will be 5G phones shipping this year. All the major handset handset vendors should have 5G (or at least 5G-ish) phones on the market, with the exception of Apple. As we noted in our CES post, Apple is likely in no hurry to launch 5G as it will have only limited benefit to its customers, with early 5G features heavily favoring improvements that for the most part only impact the operators. Intel seems to be pulling in the schedule for its 5G modem, so it is possible that Apple will have something ready, but at this stage the main significance of a “5G Phone” is the marketing value of that label.
However, going forward, it is important to notice that the leading handset brands, except Apple, have their own modems. Depending on how you count, Samsung and Huawei lead the world in market share, and they will use some combination of their own modems (where they can) and Qualcomm’s (where they have to, e.g. the US).
In our post last week about Apple’s new campus in San Diego, we said that it looks like Apple may finally, actually (not kidding this time) be building its own modem. Certainly, looking at the competition, you would think that they have to do this. However, it is important to note that both Huawei and to a lesser degree Samsung have network equipment businesses. This means they already have large teams of wireless protocol engineers who can (theoretically) help in the development of a chip for phones. Apple does not build base stations (ever), so they will have to dig a bit deeper.
As we highlighted, a lot of the work in building a modem requires supporting the full features of the standard, including all the edge cases and legacy protocols. Building a 5G-only modem is a fairly straightforward exercise, but as one contact told us: “Where they [i.e. Qualcomm] get you is when you have to fall back to an older standard.” Every 5G modem will have to support 4G and 3G and 2G, until someone builds a 5G-only network, which is years away (if ever). It is unlikely that Apple has this recipe book yet. Further, another big hurdle for modem vendors is getting those modems certified to operate on the operators’ networks. This is a cumbersome, somewhat bureaucratic process that can take a lot of time. This is part of the reason Mediatek has never gained much traction in the US. Apple, with its massive cash pile, could get it done, but the question is do they really want to do it themselves.
After our last post, we had a few conversations with industry observers and participants, which has led us to a view of how things could play out. Apple clearly wants to move as much software to their own chip teams as possible. And the industry is moving towards a higher degree of vertical integration. But Apple lacks support for the full modem feature set and probably do not want to handle the back-end work like carrier certification. So it appears that Apple is moving towards a model where they divide up the work with Intel. In this scenario, Intel licenses Apple a bunch of modem software that Apple then burns into its own silicon. Intel either sells Apple a very stripped down chip or conceivably “half” a chip with some features on Apple silicon and some on Intel silicon. Intel handles the carrier certification and other messy bits, and Apple focuses on integration between its phones and the key pieces of modem communications (think of ways to save battery life).
It turns out there is precedent for this. This model looks a lot like how one-time industry leader Nokia built its 2G and 3G modems with chip supplier Texas Instruments. Not so long ago TI held the lead position in modem market share almost largely through its relationship with Nokia. (Remember OMAP? Bonus points if you also remember Locosto.) As the market changed and Nokia exited the handset business, it became clear that much of TI’s modem design was actually done inside Nokia, TI largely just did the integration and back-end operations. That is why there are lots of great modem and RF engineers for hire in Finland today and not so many in Texas. (No disrespect to the State of Texas, they have a lot of great base station engineers.)
And then there is the whole issue of World War Lawsuit between Apple and Qualcomm. Eventually, the two companies will settle, but there is a pretty wide spectrum of potential outcomes. If the two can reach a semi-amicable settlement, then Apple could once again become one of Qualcomm’s biggest customers. This is exactly what happened after Nokia and Qualcomm went through their own outbreak of legal wrangling ten years ago. Post-settlement, the two became very close until Nokia fell apart. This outcome would be very bad for Intel. Practically, it would mean the end of Intel in mobile. Theoretically, it could lead Intel to buy Qualcomm. But file that under highly, highly speculative.
At the other end of the spectrum, if Apple prevails over Qualcomm, especially if they manage to invalidate a Qualcomm patent or two, then the industry could change pretty dramatically. That being said, it seems highly unlikely that Apple can score such a one-sided win. And while we have absolutely zero legal training, Total Victory almost never happens in commercial Intellectual Property litigation.
So the middle ground looks something a lot like where the industry is today. Qualcomm would remain the leading merchant vendor with Apple at unfriendly arms length. Intel dragging along in Apple’s wake, and everyone else using some combination of merchant chips and internal solutions.
To sum up, despite all the negative headlines and ongoing lawsuits, Qualcomm is in a pretty strong position going into the (multi-year) 5G roll out. They cannot ‘dominate’ like they did with 4G, but they are still positioned very strongly. Mediatek is a bit of a wild card, maybe they can break into the US and Tier 1 vendors with 5G, but maybe not. Intel remains in a tight spot, dependent on a single, domineering customer. The one big difference between 4G and 5G is that growing vertical integration at Samsung and Huawei (and maybe others?). In the past these internal solutions have lagged the merchant options, but they have been steadily improving. Going into MWC, this is where we would focus our investigation.