Why This Company May Be the Next to Bail on Smartphone Chips

After Broadcom announced that it will bow out of the race for cellular apps processors and modems, it's probably time to start thinking about which company is going to be next. Even as the industry has consolidated over the years with the exit of Texas Instruments, Freescale Semiconductor, ST-Ericsson, and now Broadcom, the consolidation isn't over yet. So who's going to be next on the chopping block?

The characteristics of the next player to exit
Thanks to the ease of licensing an ARM Holdings (ARMH) processor core and GPU IP, the players with staying power will be those that can afford to stay on the leading edge of modem development, have robust connectivity solutions, and can deliver competitive products within a very compressed timeframe. More to the point, though, this is a very R&D-intensive business, and only companies that can afford the multiple billions of dollars per year required to play in this market will ultimately survive.

Keep in mind that market-leader Qualcomm (QCOM -2.36%) has achieved the sheer dominance that it has today thanks to a very high R&D budget and a very smart management team that has effectively utilized that budget. On top of that, Qualcomm's principal business -- wireless IP licensing -- generates nearly $7 billion a year in operating income and grows roughly in line with the growth of cellular-enabled devices. Qualcomm can afford to play hardball in chips because of this massive cash cow.

So the winners will have to have the financial wherewithal to withstand a very aggressive assault -- both technically and in the market competing for designs -- from Qualcomm.

Marvell seems to be next on the chopping block
Marvell Technology Group (MRVL -4.77%) is a semiconductor company that plays in three principal markets: storage controllers (44% of revenues), networking (18% of revenues), and mobile and wireless (33% of revenues). Until recently, Marvell's baseband business has been fairly weak, as the company's 3G solutions came under pricing and competitive pressure. However, with the company's introduction of 4G LTE solutions, particularly in China, the tables have begun to turn.

In the most recent quarter, the company reported 30% sequential and 139% year-over-year growth in mobile and wireless, driven pretty hard by LTE growth. As part of this virtuous cycle, the attach rates for connectivity solutions is essentially 100%, as the low-cost handset markets tend to want to go with the same vendor for the entire platform if possible. Marvell's mobile business is looking great right now.

An early lead on 4G is driving these results
Right now, Marvell has an early lead with low-cost 4G LTE solutions in the market. Its leading product, the PXA 1088, integrates four 1.3 GHz ARM Cortex A7 CPUs and a five-mode LTE category 4 modem integrated onto the die. As part of the platform, Marvell is offering its own connectivity solution, power management IC, and audio CODEC. This isn't anything revolutionary, but it is designed to be cheap (and is apparently built on Taiwan Semiconductor's very mature 40-nanometer process) and is readily available today.

Right now, Marvell is one of the few with an LTE modem, which allows it to compete with Qualcomm for design wins. However, over the next six to 12 months, there will be a flurry of other "me too" players such as MediaTek rolling out LTE-capable integrated parts for the low-cost markets Marvell is targeting. While Marvell will certainly be in the running, it stands to lose share (or risks seeing margins erode) at the hands of a player like MediaTek. If Marvell had a structural cost structure advantage or if it had a major technological edge over its competition, then it'd be much easier to be long-term bullish on the company's staying power in mobile.

Is Marvell spending enough to remain long-term viable?
Marvell's entire R&D budget for the year ending in February 2014 was $1.156 billion and SG&A worked out to an additional $270 million. Broadcom's operating expenses for cellular alone worked out to $700 million, let alone the connectivity portion of its mobile and wireless business (which Broadcom had previously indicated cost about $400 million in R&D per year). To put this into perspective, both Qualcomm and Intel incur north of $3 billion on mobile operating expenses, and MediaTek spends about $1.1 billion a year on R&D, the vast majority on mobile-related technologies.

In the long run, this inability to support a larger R&D budget for mobile apps processors and modems will prove a significant headwind to the competitiveness of this business and could relegate the company to very slim margin niches should the company choose to continue in this business in the long run. While Marvell may actually grow this business enough to support dramatically increased R&D from current levels, such a scenario just seems difficult -- but not impossible -- to imagine.