Mobile failures drive chip giants to new alliances, while Qualcomm survives

The failure of many chipmakers to challenge Qualcomm in the smartphone chip market is still having ripple effects on many players today, continually shifting the landscape for mobile semiconductors. Even MediaTek, once seen as a danger to the US giant in its core business, now gets only 20% of its revenues from smartphones, while mobile disappointments have driven companies like Intel and Marvell to seek new strategies and new acquisitions.

One of many reasons why Broadcom should want to acquire Qualcomm is to make up for its own failure in the mobile modem/processor market, which remains highly strategic – cellular chips should generate both unit volumes, thanks to the wireless IoT, and a revival in margins, thanks to 5G, over the next few years.

But Broadcom, and its briefly-owned Renesas Mobile adventure, are hardly the only casualties of a mobile market in which few challengers managed to keep the dominant Qualcomm awake at night. Indeed, Qualcomm’s issues stem from customer and government frustration with its stranglehold on cellular technologies, not from any real weakening of its technical leadership.

That customers like Apple turn to the law courts, not merely to other chip suppliers, is a sign of how slowly rivals have been able to provide alternatives to Qualcomm in key product areas – and that will likely be repeated in 5G, given the San Diego firm’s near-monopoly on the devices in 5G trials. Apple is seeking to work with Intel, MediaTek and others to reduce its reliance on Qualcomm (or Samsung), but could find itself disadvantaged in 5G if its new friends fail to keep pace.

To be fair, Intel has a 5G modem in the market and is giving Qualcomm a rare run for its money in an emerging cellular segment. But if Intel succeeds in snapping up major commercial contracts, it will be the first time it (courtesy of its Infineon Wireless acquisition) puts real pressure on Qualcomm in an advanced technology.

Others are already focusing elsewhere rather than hurl large piles of dollars at the 5G race. MediaTek, although a top three modem player, has been mainly successful in lower end products, and now looks more focused on the IoT, where its low cost, efficient  products come into play, than advanced smartphones. The same is likely to apply to Chinese players like Spreadtrum, at least in the early, premium phase of 5G.

MediaTek expects that chips for IoT, game consoles and ASICs will contribute about a third of its total revenue this year, reducing its reliance on handsets and opening up new customers such as Amazon, Google, Sony and Chinese internet players. CFO David Ku, on the third quarter results call, said the company’s IoT strengths included sensors, processors, communications and ultra-low power technologies. The firm is seeing demand from Chinese bike share fleets, voice assistants (its quad-core processor is in the Amazon Echo), game console ASICs and set-top boxes.

Not that it will move away from smartphones altogether, especially if rumors that it is negotiating with Apple to supply future iDevices prove true.

But MediaTek’s Q3 sales fell by over 18% to NT$63.7bn ($2.1bn) because of strong price competition in the smartphone market, especially in China. The smartphone business in 2016 and 2017 has been “flattish”, said the company, especially as companies like Samsung and Huawei increasingly use their own designs.

In 2017, MediaTek expects to ship about 440m chips for smartphones and tablets, about the same as last year. It has about 20% share of the smartphone market by revenue, down from 40% during the Chinese LTE boom of 2015, and about half its sales in this industry still come from China. The high end Helio smartphone SoC line accounts for 15% of total sales, down from 20% last year, but should make some gains with the launch of the Helio P23 and the MT6739 for high end devices this quarter, as well as the Helio P70 and P40 for the midrange, due in the first half of 2018.

“Mediatek lost almost all the share in this category in the past year to Qualcomm’s Snapdragon 630/660, though MediaTek should gain some back with the new modem,” said Credit Suisse analysts in a client note.

“With our new product portfolio coming online, we are more confident about gaining market share back next year,” Ku said.

The succession of mobile chip failures has had many knock-on effects including some of the consolidation which has occurred in recent years. Former market incumbents like Freescale and even Renesas Mobile first exited the handset modem space, then focused on the IoT, and then were themselves acquired.

Marvell and Cavium may be the latest marriage of convenience to target the growth areas of the mobile industry. Marvell also tried and failed to challenge Qualcomm in the handset SoC space though it has interesting WiFi and Cloud-RAN offerings. By combining with Cavium, which is biting at Intel’s heels in the Cloud-RAN and data center sectors, it would create the most credible player to date in the emerging category of ARM-based server chips – and provide a headache for Qualcomm’s own new server portfolio.

Last week, the companies were reported to be in “advanced talks” about an acquisition of Cavium which would create a chipmaker worth about $14bn, according to the Wall Street Journal. At the end of last week, Cavium had a market capitalization of about $4.5bn while Marvell’s was over $9bn.

Meanwhile Intel, while its modem business is finally flourishing, has exited the modem/processor SoC space after repeated product flops and is now heavily focused on the IoT, especially connected cars; and on mobile infrastructure, taking advantage of the operators’ shift towards virtualized networks running on servers, rather than specialized networking gear running on exotic ASICs.

It has not engaged in the wave of mega-purchases but has made some strategic acquisitions, namely of FPGA leader Altera, plus some ARM-based telecom network assets like LSI Networks (from Avago) and Mindspeed Wireless (formerly Picochip, from MACOM). Moving into FPGAs has given it valuable assets in programmable chips, which are increasingly essential to complement microprocessors in cloud and telco network infrastructure (supporting offload of intensive processes and flexibility to adapt to technology or use case changes)’

But Intel still faces losing its global leadership position to Samsung, a blow which would be partly down to its years of mobile failure, which have prevented Intel replacing its core PC chip business, as that declined, with revenues from smartphones and other devices. The IoT now urgently needs to fill the void, but remains a nascent market, and many of its products will be low cost and low margin – hence the concentration on high value areas like cars, drones, gateways, robots and wearables.

Samsung almost trebled its third quarter profit this year, largely because of chip sales which broke company records and hit $17.8bn, ahead of Intel’s Q3 revenue figure of $16.1bn. Growth in semiconductor profit offset quarter-on-quarter declines in the handset and display businesses, with the chip division accounting for US$8.9bn of Samsung’s $12.9bn in profit, particularly driven by high end memory chips.

All this makes Samsung highly likely to be the number one chip vendor in 2017. Overall, its group sales in the third quarter were up 30% year-on-year to KRW62.05 trillion ($55.4bn) and profits were up 261% to KRW14.53 trillion ($12.9bn). The firm said its handset unit saw strong shipments because of the launch of Galaxy Note 8 and solid sales of the new Galaxy J series, but its earnings declined sequentially as shipments shifted towards the mass market models. Samsung Display posted an earnings decline despite higher sales of flexible OLED screens for premium smartphones.

Intel will bite back of course, and has multiple strategies to snatch back its lead, many focused on leveraging its greatest advantage, its installed base of server processors, to expand into higher growth segments like network infrastructure, AI systems and (more problematically) webscale cloud platforms.

This has even driven the giant to form some strange alliances, including a recent one with its perennial challenger and the only significant x86 rival, AMD. This focuses on the last remaining growth area left in the PC market, the ultra-mobile laptop (often a hybrid tablet/PC, sometimes even a Chromebook).

The two companies have partnered on a product which integrates an Intel Core processor, a semi-custom Radeon graphics chip from AMD, and second generation High Bandwidth Memory (HBM) into a single package. Intel has failed successively to make its own premium  dedicated GPU.

The combination will help the companies prolong the life of the PC, and fend off mutual foe Nvidia in key areas such as artificial intelligence (AI) acceleration. The hand of Apple may be seen here – as the kingmaker prepares to weaken Qualcomm and possibly bolster MediaTek and Intel, it may also have pushed Intel for a processor to support an ultra-thin, graphics-intensive MacBook (and AMD is its usual GPU supplier).

Christopher Walker, general manager of Intel’s mobility client platform, wrote in a blog post that the partnership was the result of identifying an opportunity for thinner, lighter and more powerful laptops which also support high end graphics and gaming (the latter usually add weight and size). Slimline laptops usually have embedded, shared memory to reduce size but high performance graphics require dedicated memory with a wide memory bus. Intel’s Embedded Multi-die Interconnect Bridge (EMIB) technology allows  three different devices to be incorporated into a small package at low power.

Though AMD is about to launch a new Ryzen Mobile processor, the Intel multi-chip solution with HBM2 memory and a discrete graphics chip will be more powerful and will help AMD to address the very high end of the laptop space.

Jon Peddie, principal of Jon Peddie Research, told EETimes: “Intel has demonstrated extraordinary enlightenment and shed the ‘not invented here’ attitude that has crippled so many companies.”

“The growing threat from Nvidia in both graphics and AI acceleration has turned the long running battle between Intel and AMD into a situation of strange bedfellows,” added Rob Lineback, a senior analyst at IC Insights.

Intel chose to partner with AMD rather than Nvidia in GPUs because AMD had allayed its new partner’s concerns by separating the graphics unit into a separate entity, Radeon Technology Group, said analysts.