34 Weeks of OHA: #14
Company Name: Marvell Semiconductor
How the OHA site classifies them: Software Company
What the OHA site says about them: Marvell is a leader in development of storage, communications, and consumer silicon solutions with a diverse product portfolio that powers the entire communications infrastructure from enterprise solutions to mobile consumer devices.
What they do: Honestly, how many fracking semiconductor companies does one game-changing mobile OS initiative require? The OHA, apparently, requires nine of ’em. Nine. That’s ridiculous. One search company, nine semiconductor companies. You can’t do anything these days without inviting along a gaggle of solid-state geekery.
A scan of Marvell’s tech shows something of a focus on networking: Wireless LAN, VOIP, network switching, routers. Their wikipedia page claims they were the first to create â€œmerchant Gigabit Ethernet solutionsâ€ . They also do some storage stuff–serial ATA solution, NAS bits, etc. Their website tagline reads â€œThe Market leader in Switching, Transceivers, Wireless, PC Connectivity, Gateways, Communications Controllers, and Storageâ€. Great.
They seem to be more system infrastructure focused than other semiconductor companies in the OHA who concentrate on the sexy stuff like graphics and CPUs.
In 2006 they purchased the Xscale microprocessor tech from Intel. It’s Xscale, an implementation of the ARM architecture that’s used in Blackberries, Treos, iPaqs, and the iPhone, which really drops Marvell into the center of the mobile handset game.
What they bring to OHA and Android: From what I can tell, its pretty much all about Xscale. Dunno, maybe some Wireless or dedicated VOIP chip stuff. If I was an insider I’d tell you, but I’m not.
The thing with all the semiconductor companies is that the Android developers–and I’m talking the kernel folks here, not the guys building the Twitter clients–need reference hardware and documentation for as much of the variety of existing and upcoming hardware technology as possible, and a contact with the manufacturer when things go tits-up. How do we know if this thing runs on an Xscale? We try running it on an Xscale. If it doesn’t run quite right on that Xscale, we’d better talk with Marvell about what the problem might be and how to deal with it.
It is therefore in Google’s interest to get these folks involced â€“ Qualcomm, Broadcom, Texas Instruments, Marvell, etc., etc. Get ’em on board, offer ’em membership in your exclusive new club, give ’em an early-in on what may be a revolutionary product.
From the hardware manufacturer standpoint its a no-brainer. The investment is relatively small: some sample chips and a schematic or two. The payoff is potentially big: if Android goes huge your chip is attractive to a handset manufacturer because, as one of the reference hardware providers, Android on your tech is a solid bet; if Android doesn’t go huge, you haven’t lost anything. You get to associate your name with the Open Source, free-phone, blah-blah propaganda and get another excuse to issue a press release. All good.
What I’m saying is that a semiconductor company’s involvement does not necessarily mean that Marvell, for example, is working in the bowels of the Googleplex with the search giant’s top engineers developing the next gen of VOIP chips custom designed to work with Android and Google’s newly-purchased Skype (rumour!) network. We don’t need to suppose that Marvell has some cutting edge technology that Google got a whiff of and decided it needed, the semiconductor maker just has an established chip design that is widely adopted in the industry and Google needs to make sure its new OS will play nice.
This stuff is a lot more boring than I thought it was gonna be.