We think there may a crisis brewing the mobile phone industry driven by changes in the chips that power, or fail to power, our smartphones. This is a post about Qualcomm, Mediatek and, of course, Apple.
We are framing this as part of our regular survey of the mobile baseband market, which we have been publishing (mostly) annually for over a decade. This industry has changed a lot in recent years. And as we reviewed the industry, we realized that some of these changes have been so gradual that they may have slipped beneath the radar, but are nonetheless very important.
Smartphones contain two key semiconductors – the Baseband (which is sometimes called a modem) and the Applications Processor (AP). The baseband is the chip that makes a phone mobile. It connects the phone to the cellular networks (3G, 4G, 5G, etc.), it decodes wireless signals and translates them into 1’s and 0’s with meaningful signal. The AP is analogous to the CPU in a laptop, and does pretty much what the name says it does – it runs all the applications on the phone – the Operating System (OS), the camera, all the apps, etc. There is considerable interaction between these chips, and so often they are bundled together either as two chips in a single package, or a single chip with discrete blocks for the different systems.
Crucially, the choice of a baseband drives the design of the rest of the phone. This gives the modem vendor important leverage in the ecosystem, driving many follow-on chip purchase decisions.
The largest vendor of both products is Qualcomm. They typically sell a bundled baseband/AP combination under the Snapdragon brand. Generally speaking, they prefer to sell the bundle, but for certain large customers (e.g. Apple, Samsung) they just sell a baseband. Very rarely do they ever sell a standalone AP, although they are trying to do this now to enter the PC market.
The second largest vendor is Taiwan’s Mediatek. Mediatek sells its bundled modem/AP under the Helio brand name. Mediatek also still sells chips for a lot of non-smart phones (i.e. feature phones) which do not require an AP.
Depending on who you ask and which time period you scrutinize, Mediatek has arguably overtaken Qualcomm in market share. Teasing out the data is difficult because so often modems and APs are bundled together that it is hard to pinpoint share, regardless the two companies are very close. For a long time, the two companies were hyper-competitive. That abated a decade or so ago, Mediatek effectively stopped trying to penetrate high-end, flagship phones and Qualcomm abandoned their efforts to break into low-range China phones. The two companies then largely divided the market between them, with the only real competition taking place in the mid-range, $300-$500 segment of the market. Qualcomm could not really support the sales engineering work it would take to penetrate below that, and Mediatek just had better ways to spend its time that chasing Apple and Samsung Galaxy designs.
Sitting below Mediatek is the Chinese company Unisoc – part of State-owned National Chip Champion Tsinghua Unigroup, and consisting largely of the companies once known as Spreadtrum and RDA Micro. Unisoc mostly competes at the very low end of the market and still does decent feature phone business, keeping Mediatek on its toes to some degree.
Not coincidentally, at the same time Mediatek and Qualcomm called a truce, which solidified pricing, the large phone makers started to make their own silicon. We will come back to this below. And then a few years later, Mediatek acquired a team of modem engineers that had once been the kernel of Nokia’s semis team. Through this, the company has gradually been able to improve its offerings even notching a few flagship wins at Samsung. More importantly, this very seasoned team helped Mediatek accelerate its pace of development. For years, Mediatek was content to build chips for new standards two or three years behind Qualcomm. Starting with 4G, and then really with 5G, Mediatek meaningfully narrowed that gap. At this point, Mediatek is, roughly speaking, less than a year behind Qualcomm in terms of productizing the new standard.
Historically, Qualcomm built its model around being the first to provide basebands for the new standard. For almost two years they had a virtual monopoly on 4G modems, giving them jealousy-inducing pricing. That does not seem to be happening for 5G. The two companies do not exactly compete on price. Qualcomm’s strategy is to diversify its portfolio, adding chips that target different segments of the market. And while this puts a little pressure on their margins, it has the added benefit of expanding the market. Combining all this, we are seeing a rapider than usual proliferation of 5G modems with both Mediatek and Qualcomm offering chips priced for the mid-range not just the high end, as we would expect in the early days of a new standard.
Competing with Customers
And this brings us to Apple, where all conversations about smartphones end up. Recognizing the importance of smartphone software, from almost the beginning of the iPhone in 2007 Apple designed its own AP. This became formal when Apple unveiled its A4 in 2010, essentially the first chip built by Apple’s now famous Apple Silicon team. And the advantages of this approach eventually became glaring. Apple controls both the hardware and software on its phones. This gives it a big advantage as it can tightly couple the two together and drive important performance and energy advances. Samsung had been building its own APs for longer, but Apple showed how big an improvement this could bring. By building its own chips, Apple was also able to control the roadmap of features in the AP, rather than have to negotiate with the merchant vendors against all their own competitors. In particular, Apple has continually added image processing features to its AP, helping them build a strong lead in video and camera apps.
The gap in performance prompted other handset makers to follow this path as well. Huawei in particular built some very solid chips, but their efforts were cut short, a casualty of the US-China Trade War. At this point, all the leading smartphone vendors are designing their own APs. Apple only uses its own AP, the A Series. Samsung uses its own Exynos AP for 40%-50% of its phones, and splits the rest between Qualcomm at the high end and Mediatek for lower priced phones. The Chinese vendors – Xiaomi and BBK (Vivo, Oppo, OnePlus, etc.) still rely largely on the merchant vendors but are all at various stages of building their own APs.
Smartphone vendors are struggling to differentiate their products. Apple does this with its hardware/software combination, but none of the other vendors control the Operating System they use. Instead, they are dependent on Google who is seemingly ambivalent about Android. This leaves the AP as one of the few ways these other vendors can differentiate. This point has been hammered home lately as Apple seems to have picked up much (maybe most) of the share opened up by Huawei’s exit. At the same time, pricing for Android phones has come under pressure, see above about the rapid proliferation of 5G phones.
This puts the merchant chip vendors in a dangerous position, especially Qualcomm. Apple’s AP is not only better than all the other internally-designed APs, it is better than Qualcomm’s and Mediatek’s. Depending on how you measure it, Apple’s A14 is one to two years ahead of Qualcomm’s flagship Snapdragon product. Squeezed betweeh Apple’s leading offering and the merchant vendors’ lagging solutions what can Apple’s competitors do?
- Design their own chip
- Rely on Qualcomm and Mediatek to catch up
- Find some alternative
Designing a chip turns out to be harder than it looks. As we have detailed, designing a chip is expensive. Over the years, the pace of the smartphone business has accelerated greatly. One of the big advantages that the Chinese phone makers brought to the table was their ability to iterate quickly and turn around phone designs really quickly. This approach works in a somewhat fashion-oriented business like phones, but does not sit well with a multi-year planning cycle required to roll out a chip roadmap. There is a management philosophy making the rounds among many China Internet companies lately that every company should re-organize itself every six months. This idea has merits for an Internet company (maybe), but we have seen it wreak havoc on one handset vendor’s chip teams, as new management teams come in with new priorities, causing a complete re-design of the chip at the 11th hour before tape-out. Samsung is another important example, they have been in the AP business longer than Apple has been making phones, but they still rely heavily on others’ chips. Building a sustainable chip business is hard and may require expertise beyond the reach of many of the vendors.
On the other hand, waiting for Qualcomm and Mediatek is fraught with peril. For some reason, Qualcomm does not seem to be sounding alarm bells internally about its growing performance gap against Apple, nor the narrowing gap with Mediatek. Mediatek has made a lot of gains but still lags in some critical areas, like AI for image processing. Qualcomm does have a great track record, and their chip design team is a machine at executing, but right now it seems to be an engine with the gears not fully engaged.
Lastly, is there any hope someone else will come along? No, there isn’t. Google just unveiled a phone AP. That deserves a post in its own right, but that post will conclude – you can’t rely on Google to stay in this business. There are a number of companies in China designing things that could someday be an AP, but these are years from viability. The major chip companies (including Intel, TI and Nvidia) have all been in this business and exited long ago. There really is no knight in shining armor over the horizon.
From our view, the most likely outcome at this stage is that Apple will continue to take the cream of the market. Recall our analysis from 2009, where we pointed out that Apple’s focus has always been on profit share not market share. That took a pause for a while, for a variety of reasons, but the underlying forces remain at work. Everyone else will face continued pricing pressure resulting from largely undifferentiated phones.
This is not good for either of the AP vendors, but it is much worse for Qualcomm. Qualcomm’s model is based on winning the high end of the market. This arena allows them to sell higher-featured, higher-priced chips, as we said. And with share at the high end shifting to Apple, who does not exclusively use their own AP, Qualcomm’s addressable market is shrinking. This is compounded by the handset vendors who do end up rolling their own APs. Qualcomm will likely still be competitive against those, but the sale becomes harder. Compounding the problem, recall that Qualcomm’s much-contested, highly-profitable licensing business charges royalties based on the price of phones sold. If prices continue their recent decline, that profit pool shrinks too. Note that Apple still has to pay a royalty here, but at what everyone believes in a fairly low rate. All of this leaves Qualcomm with a shrinking market. True, the advent of 5G mmWave can stave this off, hence Qualcomm’s many proclamations that mmWave is imminent. For our part, we remain highly skeptical about the timing of mmWave.
And just to complete the pile-on. Recall that Apple is also building its own baseband. Whenever this becomes available (best guess 2024), that will both hit Qualcomm’s modem revenue and apply further pressure to all the other handset makers. Our guess is that Apple will build proprietary communications features into its new modems (think blue bubbles but for phone and video calls), features that no one else will be able to copy.
Qualcomm still has many levers it can pull. As noted, they have incredible talent and all the assets they need to regain technical parity for their AP. They can also try once more to dig deep into China and build out their capabilities there. Finally, we think they are one of the few organizations with the resources, capabilities and incentives to take Android off Google’s hands. A truly-open source Android, driven by Qualcomm behind the scenes could solve many of the problems faced by Qualcomm’s customers.