From our sister service Wireless Watch
There were several mobile chip-related news announcements during our summer break which would have been unthinkable just a few years ago. Samsung’s semiconductor arm overtakes Intel in revenue and profit. Intel wins a modem deal with Apple, ahead of Qualcomm. Samsung pips Qualcomm to the post to be first with a six-carrier LTE modem. Is the mobile chip market changing forever or are these temporary blips?
The chip industry landscape is shifting rapidly now after a decade of gradual but steady change, much of it driven by the rising importance of mobile devices and the eclipse of the PC. That change shifted power and profits from Intel, with its successive failures in the mobile processor market, to Qualcomm, the apparently invincible king of that sector. But the plates have kept shifting, and now the smartphone is being eclipsed, in its turn, as a source of growth, raising the potential for a new power player to emerge, driven by the rise of ‘connected everything’ and the Internet of Things (IoT).
Intel, Qualcomm, Samsung and other players from many segments are chasing the IoT opportunity, but meanwhile, they need to keep traditional businesses healthy too. That is where the problems are currently arising for Intel – managing a difficult transition from a PC-oriented device business, and floundering in some areas of the IoT – and for Qualcomm, which is facing the possible destruction of its unique business model at the hands of Apple and various antitrust agencies.
Samsung emerged the winner of the recent round of quarterly results and news headlines, especially as other challengers for Qualcomm’s crown, notably MediaTek, are currently lagging. The Korean firm, which faces troubles of its own in its smartphone division, has offset those problems with strong performance in its semiconductor and display units.
This enabled it to overtake Intel – by some calculations at least – in the second quarter, largely on the back of the booming memory market, which Samsung dominates. Samsung’s semiconductor business reported $7.2bn in operating income on revenue of $15.8bn in Q2, while Intel said it had earned $2.8bn on sales of $14.8bn. This is a mighty blow for Intel, which has been the world’s biggest chipmaker for 25 years. Most expect Samsung to widen the gap during the rest of 2017, though by the end of this year, Intel’s own memory chip production will have expanded and it could even catch up Samsung in the vital flash memory sector.
One analyst firm, IHS Markit, actually thinks Samsung’s lead is less clear than it appears from the top line figures. It excludes the sale of components to third parties (since they would also claim the sale, leading to double counting), so that would take foundry revenues and other indirect sales out of the equation. That would leave both companies on revenues of about $14.4bn.
Samsung will be well aware that its achievement is heavily reliant on memory, a volatile market and one where its lead is not unassailable. Even though it had almost 37% of the flash memory segment at the start of this year, it could be squeezed by new investments at Intel and at second-placed Toshiba, though the latter is being hamstrung by delays to its deal to sell its memory business to a consortium led by Bain Capital.
But the Korean company is reducing its reliance on memory chips by making its own investments, in other areas of semiconductor technology including mobile processors and modems. Its own handset division is the biggest customer for these products, including the Exynos system-on-chip family, enabling Samsung to reduce its own dependence on Qualcomm silicon (though it still uses Qualcomm chips in a large number of devices). But Samsung wants to be a major merchant vendor of mobile chips too, and that means taking on Qualcomm head-to-head.
Over the past few years, the Korean firm has made progress in this area, consistently narrowing the timing and innovation gap between its own processors and modems, and those of the market leader. Now it has announced a rare instance of getting there first in an advanced cellular technology, unveiling the first LTE modem to support six-way carrier aggregation.
The product can enable devices with download speeds of 1.2Gbps, a 20% increase on peak speeds from five-carrier aggregation, which is starting to be trialed or deployed by a few operators. The six-way functionality is part of the Category 18 modem specifications for LTE, which also include 4×4 MIMO and 256 QAM modulation, as well as well as support for LTE-LAA (licensed assisted access), so that 5 GHz unlicensed spectrum can be aggregated with a licensed band.
Woonhaing Hur, VP of system LSI protocol development at Samsung Electronics, said: “With the increase of high quality online content services, the demand for high performance LTE modems continue to rise as well. The 1.2Gbps maximum downlink speed with 6CA support highlights Samsung’s leading design capabilities and well positions Samsung for the upcoming 5G era.”
Samsung expects a mobile SoC using the new modem to be in mass production by the end of this year, with its own Exynos presumably being the first target – highlighting the corporate dilemma between keeping these advances inhouse in order to give Samsung Mobile an advantage, or using them to win wider business before Qualcomm and others catch up.
Of course, its rivals in the handset market may feel disinclined to rely on Samsung for essential chips (though they are sometimes forced to in flash memories, especially at times of supply constriction). Samsung’s role in the iPhone has been eroded over the years as its handset division’s competition with Apple intensified. Apple took processor design inhouse at an early stage and has repeatedly cultivated alternative sources for foundry services (TSMC), memory (Toshiba and others) and displays (Sharp, LG and others) – though Samsung is still an Apple supplier in all these categories.
Now it is Qualcomm’s turn to feel the chill when Apple feels it wants to reduce a supplier’s power (along with Imagination Technologies on the GPU front of course). Apple has relied exclusively on Qualcomm for modems since the third iPhone generation, ousting its original partner Infineon. Now Infineon’s mobile unit belongs to Intel, and Intel has been introduced into the supply chain alongside Qualcomm, providing modems for some iPhone models.
Its most recent win, at Qualcomm’s expense, is reportedly in the latest Apple Watch. Bloomberg sources said that Apple has picked Intel to supply the LTE modems for a rumored cellular smartwatch, which could function without being tethered to an iPhone.
The new Watch is reportedly due to be released before this year’s holiday season, and would allow it to stream music, download map directions, and send messages untethered. The Bloomberg sources say that AT&T, T-Mobile, Sprint, and Verizon plan to sell the device, and discussions are ongoing with European carriers. The WatchOS v4 is also due in these new devices, providing the additional software stacks that will let the device communicate with those LTE networks.
Intel’s LTE modems are part of the ongoing Apple-Qualcomm legal dispute (see separate item). After Apple selected Intel chips for some iPhone 7s, Qualcomm filed its first patent infringement suit, claiming that Apple’s implementation of the Intel-based device violated Qualcomm’s IPR. This prompted Apple to make its own antitrust claims and start withholding royalty payments, and probes are now underway at the US International Trade Commission (of Apple) and Federal Trade Commission (of Qualcomm).
We are still a long way from knowing whether the rumored Watch will use LTE-M or NB-IoT, the two low-power, IoT-focused LTE variants that were introduced in 3GPP Release 13. It seems unlikely that NB-IoT will make an appearance, given its lack of voice communication and data, but LTE-M might well be used for the more power-sensitive features.
However, the Watch has been designed for daily or nightly charging, meaning that low power consumption is not as pressing a concern in the smartwatch as it is in other battery-powered devices that can’t rely on being recharged. As such, Apple could probably quite easily use something like LTE Cat 1 or Cat 4 in the Watch, without having to stress about power constraints.
An Apple Watch deal will be more valuable to Intel for brownie points and credibility than actual revenues, given the small size of the market. Where the real power, and potential revenues, will lie will be in early 5G devices of all kinds, from smartphones to fixed routers to cars.
To date, Intel has not had modems at the cutting edge where Qualcomm lives, and it has been assumed that it would be confined to the lower end of the iPhone family. But that may change as it upgrades its offering, and if Apple’s tensions with Qualcomm are not resolved.
Intel and Samsung are starting to go head-to-head with Qualcomm in the area where its mobile engineering prowess has made it unbeatable – products to support emerging standards and early operator trials. So far, Qualcomm has had pre-standard 5G trials to itself, supplying modems for the prototype devices, but Intel and Samsung now have 5G modems too. Intel initially had an FPGA-based 5G product but followed that, in January this year, with Gold Ridge, the firm’s first 14nm baseband, which took on Qualcomm’s X50 5G offering.