Intel's secret plan (pg. 2)
Intel hopes that Moorestown's flexibility will appeal to electronics makers, and the company is already in discussions with Sony, Toshiba, Canon, and health-care- and auto-device manufacturers about equipment that could be powered by the new chip system. First, though, Intel must find a way to manufacture Moorestown so that it doesn't, in Maloney parlance, tank margins.
That job falls to Gadi Singer, general manager of system-on-a-chip (yes, that's his real title), who must ensure that Moorestown can be built using as many of Intel's standard processes as possible. Singer learned the hard way how important standardization is: He was part of the team involved in Intel's failed foray into mobile devices 10 years ago when it inherited a low-power chip as part of its 1997 acquisition of a Digital Equipment Corp. unit.
That chip, called XScale, was based on Advanced RISC Machine (ARM) architecture, a design system that was incompatible with Intel's PC architecture. Originally designed for modems in cellphones, XScale chips were stupendous at operating at very low power, a necessity in mobile phones. But Intel had very little advantage in the market: Any company can license the ARM chip design and add its own unique features, which is exactly what Intel's competitors did. The chip barely got more than a 1% share of the cellphone business or about 16% of the PDA market. In 2006, well after the idea for Atom had been hatched, Intel sold the XScale business to Marvell for $600 million.
As management examined the problem, it realized that Intel's own PC-centric architecture could actually work to its advantage in the post-PC world. "By 2004 it became clear that a lot of the devices outside the PC space would soon be touched by the Internet," Singer says. "With the Internet come a variety of requirements and expectations that everything works as it does on your PC - that is where Intel architecture has something significant to offer." Singer and Maloney saw that Intel needed a chip with even stingier power needs, a sort of scaled-down Centrino. The Atom project was launched soon afterward.
Still, the XScale debacle has left Intel insiders wondering whether the company can pull off this high-stakes incursion into the mobile world. "That's certainly the question my board asks the most," says Anand Chandrasekher, senior VP and general manager of Intel's ultra-mobility group. Chandrasekher explains to the directors that this time Intel isn't offering a plain-vanilla version of some other guy's architecture; it's peddling the same Intel architecture that runs Microsoft Windows and Apple's Mac OS X, and on which most of the Internet runs - just in a low-power package.
The Atom line is part of the ×86 architecture, which has the deepest library of applications, everything from Excel and Outlook to the vast majority of device drivers on the planet - the stuff that makes "plug and play" possible. Intel contends that bringing computing power to the mobile world is a lot easier than taking a chip designed for mobility and giving it the ability to handle computing applications.
"Moorestown, from a performance standpoint," Chandrasekher says, "will clean the clocks of the competition."
The competition, not surprisingly, feels different. "There is nothing so sacred about the Intel architecture, about ×86, that the world has to follow it wherever it goes," says Dan Vivoli, Senior Vice President who heads up strategic marketing for what is perhaps Intel's most persistent competitor these days, graphics-chip maker Nvidia. "The iPhone is the best mobile Internet device out there, and it has nothing to do with Intel architecture."
Vivoli has a point. The iPhone, which already runs ARM, does a great job of rendering web pages. What it doesn't do is run full-blown versions of Microsoft Office. "But is anyone really going to want that?" Vivoli asks rhetorically.
While Intel is prepping Moorestown for the market, Nvidia (NVDA) is shopping Tegra, an ARM-based system-on-a-chip that it says offers all the performance at a much lower power cost than the current version of Atom. Qualcomm (QCOM, Fortune 500) is offering Snapdragon. Texas Instruments (TXN, Fortune 500) has announced OMAP4. All these system-on-a-chip solutions are slated to appear in a variety of gadgets by the end of the year. With the exception of Intel's Moorestown, they are based on ARM. Expect a battle royal, with companies that have cellphone experience and market share (Qualcomm and TI) taking on newcomers bringing serious computing chops to the fight (Nvidia and Intel).
"This is going to be a polarizing year," says Francis Sideco, senior wireless analyst with research firm iSuppli. "You are going to see all the partnerships lining up. If Intel had come in a year later, I might have said the ARM world had it all wrapped up. But this is Intel. They are still the biggest semiconductor maker on the planet, and you can't put anything past that R&D machine they have over there."
Intel also has a slew of customers, including Hewlett-Packard, Dell, and Microsoft (MSFT, Fortune 500), that surely are interested in following the Intel architecture wherever it goes. They, too, must figure out a future that extends beyond the PC, and Intel can help them do that.
Fred Weber, former CTO of Advanced Micro Devices (AMD, Fortune 500), has battled Intel longer and harder than most. From a technical perspective, he is impressed by what Intel has accomplished with Atom. His view is that Intel's success in low-power processors depends almost entirely on management's commitment to the transition, which could be rough: Investors trade Intel stock based almost entirely on the profitability of its main business, PC chips, where margins are in the 40% to 60% range. The fear, near term at least, is that moving into the high-volume and highly competitive wireless-chip business will erode Intel's profitability, and the stock will tumble.
But in my conversations with Intel executives, none seemed to waver from belief in a low-power, high-volume future. Indeed, Otellini suggests that Intel has little choice but to develop chips for consumer electronics and other communications gadgets. "The market is coming to us more than we are chasing after the market," he says. Put another way, Intel's direction seems clear, even if some of its future products are hidden from view.