Marvell: Wireless Outlook

By Guest Author Nalini Kumar Muppala

We last covered Marvell in-depth through Vijay Nagarajan’s series of articles published in the spring of 2008. It’s time to take a fresh look at recent developments at the company.

Fiscal 2009 Financials

Even though revenue increased only slightly in FY 2009, Marvell started to deliver on its promise to reign in costs. It showed a net profit of 5%, a big improvement over a net loss of 4% the previous year. Gross margin improved and is back in the lower 50s, in line with its long term operating model. Reduced R&D costs and operating expenses resulted in an increased cash flow. As Sramana pointed out earlier, all debt was paid off and cash on hand is about $950 million. Further cost savings are planned by reducing the workforce by 15% by end of CY 2009. Storage products still account for over 50% of revenue.

Design Losses

Perhaps the most popular of the design losses for Marvell in recent times concerns the iPhone 3GS. As Vijay predicted, Marvell lost the WLAN slot, first in the iPod Touch then in the iPhone 3GS to a BT/FM/WLAN combo chip from Broadcom. A similar product from Marvell, 88W8688, was available long before the Broadcom part. Thus, performance and/or pricing must have helped Broadcom win. There were other design losses, and we will look at what some of these might mean for the company’s business outlook. It must be noted that the same part that was dropped from the iPhone 3GS was designed into the Palm Pre, for which the design cycle presumably started well before that of the iPhone 3GS. Will Broadcom steal this slot in an updated Palm Pre?

Application Processor

Marvell needs to be commended for its application processor (AP) market share. According to a Forward Conceptsreport, Marvell was the third-largest supplier of standalone AP in 2008. It took 17.1% of the revenue. The report also notes that Marvell shipped more units than Samsung; however, Samsung took 17.4% of the $1.3 billion market by virtue of higher ASP. TI led the pack by a big margin. These numbers reflect Marvell’s strategy to target low-cost devices and leverage economies of scale. This bodes well from a cost perspective.

But this standing could change soon. As discussed below, Samsung is not sourcing from Marvell as much as in the past. TI’s OMAP has been designed into some of the most popular smart phones released this year.

Apart from mobile devices, Marvel targets these application processors for connected digital photo frames, where it competes with Samsung and several small Taiwanese players. Marvell also targets these application processors for home entertainment devices, which are currently dominated by MIPS and Power-PC parts.

Perhaps the significance of application processor to Marvell will be clear if we consider that over 1,000 employees, about 18% of the entire workforce, are involved in this effort. Marvell does not license the latest RISC cores from ARM. In fact, it only licenses ARM9 RISC core. While TI, Samsung and several other application processor vendors have licensed the latest RISC cores from ARM, Marvell is sticking to the homegrown, ARM instruction set-compliant, Sheeva processor. Such processors have long been embedded into other solutions such as storage control, WLAN, and Ethernet chips. In fact, CEO Sehat Sutardja claims that Marvell used its first gigahertz ARM processor several years ago in printer engines. To put this in perspective, Samsung and TI recently made big announcements about plans for gigahertz application processors.

AP or AP+BB: Let’s play both

Marvell is placing its bets on both a dedicated application processor and a combined communication processor (application processor + baseband (BB) on a single chip). This could turn out to be a smart move. The AP parts target high-end smart phones and non-cellular consumer devices requiring high performance, while the AP+BB parts target some smart phones, low-end feature phones vying to add more smart phone features, and other low-power consumer devices.

A lot of hope rides on Marvell’s “Tavor” PXA930 communication processor. Tavor is a quad band GSM/GPRS, UMTS/HSDPA BB + AP on a single chip. So far, it has not delivered the deluge of design wins that Marvell initially hoped for. In fact, Tavor faced a setback at RIM. Tavor is designed into a few Asus and Kogan devices. The CEO has promised that more devices with Tavor are in the making.

The RIM Saga

>Tavor is present in the BlackBerry Bold and an initial version of the BlackBerry Storm. Qualcomm’s MSM7600 (also AP+BB) is used in later versions of Storm and the BlackBerry Tour. Thus two of the most recent RIM phones use Qualcomm for a slot that used to be Marvell’s. Marvell has assured investors that relations with RIM are not heading south and that they have a working road map for the next several years. Coincidentally, the BlackBerry Storm and Tour do not have Wi-Fi. This probably tilted the scales towards MSM7600.

Qualcomm now provides about 25% of RIM’s processor needs. I wonder if RIM is taking Nokia’s route of multi-sourcing to reduce risk and pitting competitors to deliver better solutions. RIM is one of the few phone vendors gaining market share this year, and Marvell will likely benefit from riding this trend. However, Marvell will not reap all the benefits of a strong RIM showing; Qualcomm will take a part of it. In the long run, it would be best if Marvell started selling to other Tier 1 customers as well.

The Samsung Saga

Marvell is out of the recent devices from Samsung. The last device from Samsung with a Marvell processor was a phone from the Omnia line, released in November 2008. Omnia HD, the latest hot-rod phone in the Omnia line, sports a TI OMAP part. Lately, Samsung has been using one of their homegrown processors unless there is a compelling performance reason to go elsewhere, and when they do, it is mostly to TI, Qualcomm. Samsung’s multiple supplier strategy is nothing new. It is quite possible that Marvell is currently shipping to Samsung, and we might see such phones in the future. However, anecdotal evidence shows that Marvell is losing ground at this Tier 1 customer.

Light Elsewhere

Marvell has been investing effort and resources into providing products for the HD video market. It faced a setback when it backed Toshiba’s HD-DVD, which turned out to be the loser of the HD video format battle. However, the company claims to have made significant progress recently and is working with several HDTV manufacturers. Marvell could benefit tremendously if HDTVs start adding storage and web connectivity. Connectivity solutions such as Wi-Fi, Bluetooth and Marvell’s Qdeo video processing solution could be sold alongside its market leading storage technology solutions.

Holes in the Portfolio

While Marvell has one of the most diverse portfolios of silicon for mobile devices, it lacks a RF transceiver. All other major wireless semiconductor suppliers have one. Freescale provided RF solutions to RIM phones alongside Marvel baseband parts. Acquiring or developing one in-house would round off the company’s mobile platform offering.

Still No GPS

A year after entering into a GPS software-licensing agreement (SLA) with Trimble, Marvell does not seem to have anything to offer on this front. Little information is provided other than the company’s continuing to emphasize that its strategy is to provide GPS functionality at a small incremental cost. This would make it an easy decision to add GPS functionality to low-cost devices. It is likely that the path taken would be to sell GPS software optimized to run on its application processors for a small royalty. Marvell claims to be offering sample solutions to its customers for evaluation.

The verdict is still out on the reliability of a software-based GPS. Ongoing gains in processing power could support complex software that matches a hardware solution. However, it is likely that a silicon solution would be cheaper. Case in point: CSR did not gain much traction as long as it was providing GPS as software from its Nordnav acquisition. CSR gave up and acquired SiRF earlier this year. It remains to be seen if Marvell’s bets will pay off.

A compelling GPS solution could be bundled with other offerings. The prospect of selling more application processors is enticing. However, Marvell would have to compete with the likes of Broadcom’s BCM4760, which is a PND on a chip combining GPS and AP. BCM4760, which is sampling now, could replace Marvell application processor in some PND designs.

Gaining some momentum in the PND market was cited as one of the benefits of Marvell’s entering into an SLA with Trimble. While Marvell is still working out its solution, the competition is forging ahead. CSR acquiring SiRF is a blow to Marvell’s PND strategy. SiRF is the market leader for GPS in PNDs, and CSR will now push connectivity + navigation solutions to this market with a new-found aggression. ASSP for GPS seem to be gaining momentum, and Marvell is not well positioned for it. With the smart phones’ incursion into its turf, the market for dedicated PNDs is projected to go into a steady decline. Marvell will need to hurry up to get a slice of what is left in the next few years.

MID Play

This is a big opportunity for Marvell. Given its stagnant market share in the baseband market, Marvell is betting on mobile internet devices. Marvell demonstrated a complete reference design for MIDs mostly based on its own silicon. This is an opportunity to provide complete system solutions including AP or AP+BB, connectivity solutions such as BT, WLAN, power management, SSD control silicon, etc.

Marvell is betting heavily on low-cost consumer devices. The targeted price of $100 to $200 is thoughtful. A price of $100 makes for a conviction buy.

Conclusion

With costs under control and at comfortable levels, Marvell can now concentrate on growing the business. Increasing revenue by way of more design wins should be the top priority.

Marvell’s wireless business needs to get some big design wins at Tier 1 customers. This will provide much-needed volume and revenue ramps. The emphasis needs to be on providing design platforms.

A stagnant or declining baseband market share might prove insufficient to support continuing development. There are lessons to be learnt from the precedent set by major chip vendors that are leaving baseband space.

Marvell has in the past delivered revolutionary chips which took their target markets by storm. Will breaking into home entertainment be the next big thing for Marvell?

This segment is a part in the series : Marvell