Your browser is not supported. Please update it.

20 November 2018

Qualcomm sees its 5G opportunities squeezed by Intel and judges

The advent of 5G would, in the past, have offered Qualcomm a double opportunity – to be the only supplier of the latest modems, at least for a period of time (about a year in 4G); and to add new technologies to its licensing deals.

This time, though, Qualcomm will have to fight off Intel for deals in that coveted first mover market, and the licensing model for its latest patents looks increasingly uncertain.

On the chip front, Qualcomm has had a dominant position in providing modem chipsets and prototype devices for 5G tests and trials. However, Intel has muscled in too, denying Qualcomm its customary near-monopoly (Samsung and Huawei’s HiSilicon have also got some trials this time around too). And Intel may not be far behind its rival when it comes to pushing 5G chips into handsets and other devices. Qualcomm expects to see its baseband chips in smartphones by late 2019, while Intel said this month that it would have its silicon in commercial devices by mid-2020 – six months earlier than originally scheduled. In February, Qualcomm said “more than a dozen” handset makers around the world had selected its Snapdragon X50 5G NR modem, and claimed this gave it a lead of 12-18 months over 5G competitors, but it now seems that headstart may be under a year.

That is not a big window for Qualcomm, though it is likely to leverage its superior expertise in 5G trials and R&D – and its renowned modem engineering skills – to push its products into non-smartphone categories more effectively than most competitors, at least in the early 5G phases. Last week, for instance, CEO Steven Mollenkopf announced the company had “secured the world’s first major 5G design win with a leading automaker, and we are working with numerous other automakers and Tier 1 suppliers to bring 5G to vehicles”.

He told Qualcomm’s most recent earnings call: “Just as 5G launch dates were accelerated by one year from the original timeline, we also expect the pace of 5G adoption to meet or exceed that of 4G.”

And to allay investor fears that, this time, the firm will not have much of a honeymoon period as first into the new market, he insisted (with some credibility) that the complexity of 5G plays to Qualcomm’s strengths even more than previous cellular technologies have done.

He said: “5G brings an order of magnitude increase of complexity over 4G across spectrum, network dark architecture, RF front end, device form factors, especially the millimeter wave frequencies, power consumption and application and AI processors. This complexity plays directly to Qualcomm’s strength as an innovator and systems provider.”

Over at Intel, the company will have high hopes that it can replace Qualcomm in a future 5G iPhone (widely tipped to launch in 2020), given the ongoing feud between Apple and its long term modem supplier. Many reports speculate that Apple, which has already started using Intel modems alongside Qualcomm ones in the iDevice portfolio, will take a single-supplier approach for its first 5G smartphones – though other reports claim it will go with MediaTek too, or even develop its own chip. In July, for instance, the CTech by Calcalist website claimed to have seen internal communications from Intel which indicated the company would not find a slot in the first  5G iPhone, and it had therefore

halted development of a 5G chip, codenamed Sunny Peak, in favour of prioritizing efforts to get into Apple’s 2022 5G models.

Intel denied it had changed its plans. “Intel’s 5G customer engagements and roadmap have not changed for 2018 through 2020. We remain committed to our 5G plans and projects,” the company said in a statement at the time.

The company said its first 5G modem, the XMM 8160, will support device speeds of up to 6Gbps and is essentially a commercial version of the previously announced 8060, which was used in 5G tests. The new chip supports 26 GHz, 28 GHz and 39 GHz millimeter wave bands as a wide range of bands between 600 MHz and 6 GHz.

It combines LTE and 5G on the same baseband – a more integrated, multimode approach than Qualcomm’s Snapdragon X50, which separates 4G and 5G capabilities into two separate elements on the same chip. It can also support 2G and/or 3G, and either the standalone or non-standalone variants of 5G New Radio.

Qualcomm wins over Intel in its ability to integrate the modem and the processor (and other components like the GPU) in a single system-on-chip (SoC). Intel failed in its long attempt to deliver an all-in-one smartphone SoC and finally pulled back from its SoFIA architecture and its mobile processors. But in the Apple family, that doesn’t matter, since the iPhone maker has its own processor, GPU and other elements and is only looking for the fastest, most efficient modem.

An Apple win will be very important to Intel, to make up for its time disadvantage to Qualcomm. In effect, its first generation 5G modem, the XMM 8060, is commercially dead – not an unusual outcome in a new market, but the new product must still win a couple of big deals to generate real momentum for the business. Without the coveted Apple slot, there will be a tough battle against Qualcomm, especially with Huawei and Samsung, the world’s biggest two smartphone makers, increasingly relying on their own chips. Intel does not want to see its first fully commercial 5G modem confined to low volume markets such as first generation 5G automotive.

That outcome would not necessarily a big issue for Intel’s revenues and market share, since there will be only a few smartphones released in 2019. But the longer it delays, the more Qualcomm is able to entrench itself into the supply chains of the handset vendors as they start to plan their product roadmaps for 2020 and beyond.

“Intel faces a formidable challenge from Qualcomm which has broad manufacturer and carrier support for its 5G NR X50 modem,” said Geoff Blaber, an analyst at CCS Insight.

Intel always seeks to minimize the idea that, once again, it is going head-to-head with Qualcomm and that, once again it could fail. It portrays its 5G strategy as being a broad one, with the 5G modem just a small part of a platform that includes cloud infrastructure to support virtualized networks and edge compute; switch-chips and other silicon for transport and core systems; FPGAs and processors for specialized products like base stations, and so on.

But if the edge-to-cloud strategy were all that mattered, Intel would not need the modems at all. In fact, the modem business it originally acquired from Infineon is a reasonably important one and needs to compete on its own merits, growing or maintaining its top three position in the mobile baseband market. So there is plenty at stake, even if Intel can, unlike Qualcomm, still claim a 5G success in future even if it fails in modems.

A hidden weapon for Intel may be its close relationship with Chinese chipmaker Unisoc Communications (formerly Spreadtrum), in which it owns a stake. That alliance is significant for giving Intel a toehold in the huge Chinese mobile market, despite US-China trade tensions and China’s ambitions to grow its own semiconductor business. It also brings Intel the SoC capabilities it has failed to develop for the smartphone market inhouse. The US firm said it would be packaging its 5G modem into a smartphone reference design, combined with an applications processor from Unisoc, and that would enable it to push down into the midrange market, and have easier access to the vast ecosystem of Chinese branded and white label handset makers.

Like Qualcomm, Intel will also look for opportunities to sell its 5G chips into non-smartphone categories, given that 5G is unlikely to provide the same growth spurt for the stagnating handset market that 4G delivered. Embedded 5G for broadband access equipment and PCs are naturally on the firm’s hitlist, as well as cars and IoT gateways.

“Intel’s new XMM 8160 5G modem provides the ideal solution to support large volumes for scaling across multiple device categories to coincide with broad 5G deployments. We are seeing great demand for the advanced feature set of the XMM 8160, such that we made a strategic decision to pull in the launch of this modem by half a year to deliver a leading 5G solution,” said Dr Cormac Conroy, general manager of Intel’s Communication and Devices Group.

Qualcomm will be playing its tight integration card as usual. While Intel and others struggled to deliver an all-in-one processor/modem design for 4G, Qualcomm has gone further and integrated more and more components into tight packages, luring OEMs with the resulting reductions in cost, footprint and power consumption. The firm even offers its own RF front end modules, while Intel and others work with third party suppliers such as Qorvo, Broadcom and Skyworks.

On the patents licensing side of the Qualcomm business, Apple is also the major thorn in the company’s side. The iPhone maker has hurled a series of lawsuits against its modem supplier, and corralled some of its device manufacturers to join too. The suits are more significant than such legal battles usually are because they do not just relate to technical arguments about patent validity and violation, but also go to the heart of Qualcomm’s business practices. The war began in early 2017 when Apple sued Qualcomm over disputed licensing terms and fees, prompting the chip giant to countersue, claiming $7bn in unpaid royalties. The feud started to affect Qualcomm on both sides of its business when Apple not only encouraged its manufacturing partners to withhold royalty payments until a settlement was reached, but chose Intel to supply the modems for the 2018 iPhone models.

And recent reports indicated Apple was in no mood to settle. “There is absolutely no meaningful discussion taking place between us and Qualcomm, and there is no settlement in sight,” an unnamed source told news agency Reuters. “We are gearing up for trial.”

Apple’s actions are just one of a multi-fronted attack on the formula which has made Qualcomm successful for so long – it has been forced to compromise its practices to come to deals with some antitrust authorities, notably in China, and other disputes are still rumbling along, including at the USA’s International Trade Commission (ITC) and Federal Trade Commission (FTC).

In a preliminary ruling by the latter body, Judge Lucy Koh – most famous for presiding over the long-running Apple/Samsung patents trials – has ordered Qualcomm to license some of its modem patents to its competitors, including Intel and Samsung, under Frand (fair reasonable and non-discriminatory) terms. Frand is the baseline for licensing of patents which are central to standards, under most of the world’s licensing frameworks, and in theory Qualcomm subscribes to it, in its agreements with customers. However, Frand is often vaguely defined, and a common topic of legal test cases.

Koh’s ruling was the first in a case that began last year, when the FTC accused Qualcomm of engaging in anti-competitive tactics to sustain what the Commission called a “monopoly in the supply of baseband chips”.

Media reports said the FTC and Qualcomm asked Koh to delay the ruling for up to 30 days while they engaged in settlement talks, but Koh denied the request, presumably to establish an important principle in court. The full trial is scheduled to begin in San Jose, California in January.

The key principle at issue here is that Qualcomm has to license to rival wireless modem companies like Intel on precisely the same terms as everyone else. One of the unusual aspects of Qualcomm’s model is that customers have to sign an across-the-board licence for ALL of its patents, not pick and choose. Koh’s point is that this general licence cannot include standard-essential patents (SEPs) for 4G and 5G, which must have their own separate and clear terms based on Frand norms.

The FTC has argued that Qualcomm is committing antitrust violations by not letting companies like Intel license the SEPs separately.

There is some overlap with Apple’s claims, though in other ways, the cases are following different lines. For instance, there is a lot of difference between licensing a technique that is described in a patent, and helping a company work out how to go about executing on that patented technique. The standard may be quite straightforward, but putting it into silicon might be virtually impossible unless you are Qualcomm. This has always been the firm’s ‘ace in the hole’ – and our take is that it can comply with this ruling and still pursue its claim that Apple gave Intel some of Qualcomm’s trade secrets, which related to ‘how’ to implement some patents.

Furthermore, Qualcomm retains further patents on wireless and chip design which are not contained in SEPs, and which Intel has no automatic right to license. Curiously it has been Apple that has pioneered the idea that a patent is not something to license, but something to use to prevent your competitors from keeping up. Of course, that cannot apply to SEPs and ostensibly goes directly against the entire principle of patent law.

Qualcomm once famously, and successfully, defended a licensing scheme which appeared to engage in ‘triple dipping’ (illegal under most licensing regimes). It asked for royalties for the same patents from chipmakers, equipment vendors and operators and justified it because companies (most notably Ericsson) foundered when they tried to build chips using those patents without Qualcomm’s help.

This FTC case is still heading for trial in early 2019, but Koh was simply clarifying that Qualcomm cannot exclude Intel from receiving such patent licences, which is an avenue Qualcomm had been pursuing.

One of the most significant of the accusations Apple filed against Qualcomm last year was that the royalty scheme is illegal, because it is based on a percentage of the total end price of the devices built with the patents, not the component cost. This is not a rare practice, but given that north of 200 patents are likely to be in play in 5G, without counting those in iOS and other parts of the device, Apple argues that its license terms must be changed to a flat fee per device. Apple is clearly using these patent wars to try to retain its iPhone margins, and has made similar moves excluding other chip vendors – such as Imagination Technologies – from its devices, partly to retain that margin.

Qualcomm recently announced its fiscal Q4 figures, with net income damaged severely by the failure of the bid to acquire NXP, which cost $2bn in termination clauses. Without that one-off item, it showed revenues which were 25% down on last year at $5.8bn, and an operating loss of $700m. There were further charges – $1.2bn from the European Union, $687m in restructuring fees and a $676m gain related to the settlement of the Taiwan Fair Trade Commission investigation. Full year revenues were down 2% to $22.7bn, with net income down 21% to $5.6bn. In Q4 it shipped 232m wireless modems, up 17% for the quarter and 6% year-on-year.