Your browser is not supported. Please update it.

17 July 2015

Will Marvell be the latest casualty of the modem meltdown?

The LTE modem business continues to consolidate in the hands of a few players with the scale and resources to play in this unforgiving game. Broadcom, Nvidia, Texas Instruments, Renesas and others have thrown in the towel, and Marvell, whose Chinese successes have kept it in the game, is cutting back too.

According to Will Strauss of research firm Forward Concepts, Marvell is closing its design facility in Austin, Texas, and its wireless operations in Israel as part of a major reorganization of its LTE modem business. This will involve the loss of about 150 engineers, with most of their work presumably being transferred to the firm’s design center in Shanghai, in its most important market, China.

This is not the case of a company failing to bring a viable product to market at all. Marvell’s PXA1908 LTE modem is shipping in volume (and is in some of Samsung’s lower end Galaxy models) and has not suffered any major problems, but it seems that the company is struggling to achieve the volumes that are needed to stay afloat in a business which is, nowadays, all about scale. Companies have to perform the difficult double act of investing huge sums in ensuring they are at the cutting edge of modem design – keeping pace with constant steps forward in LTE-Advanced and now in M2M variants – while also ramping up to high volume and low cost very quickly.

Qualcomm, for all its own challenges, is a master of this two-step, and has the market share to sustain the constant parallel lines of advanced R&D with rapid time-to-market and time-to-volume. MediaTek, of course, is getting more sophisticated at this game all the time and, having moved steadily upmarket from its 2G roots and white label roots, is now taking LTE market share off Qualcomm. A recent client note by Randy Abrams of Credit Suisse predicted that MediaTek would double its LTE chip and SoC market share this year, from 20% at the end of 2014 to 40% or even more. That would mean exceeding the Taiwanese firm’s own targets and shipping between 160m and 165m LTE units this year.

Strategy Analytics analyst Sravan Kundojjala concurs, predicting in a recent report that “MediaTek will continue to gain share in LTE basebands in the next few quarters”, mainly because of its strong links with Chinese handset makers, which are themselves moving up the value chain from 2G into affordable 3G and 4G smartphones.

Because Qualcomm and other western suppliers have tended to steer clear of the lowest reaches of the handset market, they lack the entrenched relationships which MediaTek has built up over the past decade – with firms that are now becoming more attractive customers, especially as smartphone adoption veers to China and to other markets targeted by Chinese OEMs, such as India. India will overtake the US by 2017 to become the world’s second largest smartphone market after China, believes Strategy Analytics.

MediaTek has its own market timing challenges however. It has already been overtaken by Spreadtrum in 3G modem sales, which will only be a problem if it fails to compensate quickly enough by growing its 4G business – and pushing it upwards in terms of margins. Its LTE sales are still led by entry level products, where it is very exposed to Spreadtrum, but its higher end Helio family is starting to make an impact. In the first quarter, between 10% and 15% of its sales were of its top end octacore chipset and it expects that to have increased by about five percentage points in Q215.

Strategy Analytics ranks Qualcomm, MediaTek, Spreadtrum, Samsung, Marvell and Intel (via its Infineon Wireless acquisition) as the market leaders in LTE chips, as of the first quarter of this year. However, all players outside the big two are squeezed – Spreadtrum is in third place, but still has only 7% – and some of the increased share MediaTek will seize this year will be at the expense of smaller rivals rather than Qualcomm. Samsung has the advantage of its own smartphone company as a customer, but Intel and Marvell are looking vulnerable in LTE – the former, in particular, if its SoFia LTE processor/modem, essential for the Chinese market, has been delayed until next year, as DigiTimes reports (see separate item).

The other up-and-coming player which will surely break into these rankings soon is Huawei’s HiSilicon. The company, like Samsung, has a captive customer in its parent’s handset unit, and it has been investing heavily in engineers and design facilities in the past year. It has opened a center in the UK and hired people laid off by Nvidia’s closure of its modem arm, the former Icera; and Strauss reports that it is taking on many of the former Marvell engineers in Israel (some are also going to Intel).

HiSilicon is already a major player in the multimode LTE modem market, supporting different flavors of 4G (including China’s TD-LTE alongside FDD, plus 2G and/or 3G). This is the main growth area for smartphone modems, and when this is separated from the wider LTE baseband space, Strauss says the players with multimode offerings are HiSilicon, Intel, Leadcore, Marvell, MediaTek, Qualcomm, Samsung and Spreadtrum.

One of Marvell’s challenges is that it has stayed in the modem game, when so many have fallen by the wayside, largely because of early success in China, but that is now the market where new competitors are emerging most strongly. There were rumors earlier this year that it might merge its modem activities with those of a Chinese player, which would create a headache for the big two, combining Marvell’s mature architecture and strong design with some real scale.

At the start of this year, Marvell held the number two position in LTE modems in China, with 30% share, having been building handset products and alliances in the country since it acquired Intel’s ARM-based XScale business. That position has been diluted now, but Marvell has some key products for the Chinese space, such as a 64-bit processor implemented on the same die with the LTE/TD-SCDMA modem. That combination puts the company ahead of MediaTek in the vital area of SoC integration, since the Taiwanese giant has been using a separate TD-LTE chip.

Marvell also puts GPS, Bluetooth and WiFi on the same die and has its own LTE RF transceiver chip, plus a power management chip with integrated codec and charger. This combination of elements into one SoC is matched only by Qualcomm, which has also invested heavily in incorporating its own connectivity and RF elements alongside its more established processor/modem offerings. In addition, the processor has a dedicated security processor and an integrated sensor hub. The former support trusted mobile payments and the latter provides sensor fusion capabilities, allowing the device to use readings from many different sensors for virtual reality or smart home applications, among others.

Marvell, then, has strong design expertise, but it is over-reliant on one increasingly competitive market – hence concentrating its engineers in Shanghai – and, with single-digit global share, needs bigger economies of scale to attack growth markets like Indian handset makers effectively. As we argued about the possible takeover bid by Tshinghua for Micron, in an era where the chip industry is mature (only 2% growth across the board this year) and globalized, cross-border mergers may be the best way forward. US-Chinese deals are politically sensitive, but a Chinese partner for Marvell’s wireless chip business could give its excellent products the resources and scale they need to survive.