It wasn't too long ago that Intel Corp. executives were so dismissive of personal computers priced below $1,000 that they called them Segment Zero--a dumping ground for inventory close-outs and bottom-feeding PC cloners selling yesterday's technology. Instead, Intel held to the high ground, pushing pricey chips that could run the newest, coolest software. Not coincidentally, those chips helped produce the juiciest gross profit margins, some 60% overall, one of the highest in high tech.
And why not? That bedrock strategy had turned Intel into a precision profit machine, growing almost as fast as the number of transistors on its slivers of silicon. With a 90% market share in PC processors, Intel's sales have headed skyward, growing 30% to 50% annually for the past four years. That made it the eighth-most-profitable company in the world in 1996, with earnings of $5.2 billion, not far behind Exxon Corp. and General Electric Co. GE sales, however, dwarfed Intel's $20.8 billion.
ABOUT-FACE. But that was last year--when the average PC was still a pricey $2,000. In February, Compaq Computer Corp. changed everything when it became the first top-tier PC maker to hawk powerful, dirt-cheap computers. Using the Pentium-compatible MediaGX chip from Intel rival Cyrix Corp., Compaq shipped a $999 Presario (that now sells for $799)--and the sub-$1,000 market exploded. Today, almost all big PC makers are pushing inexpensive gear, and sales in the segment have surged from just 7% of U.S. retail units in 1996 to an estimated 25% this year.
Suddenly, Segment Zero looks like priority No.1 for Intel. For the first time in its history, the world's largest chip company is launching an all-out crusade to create processors specifically for the low-cost PC market, which it now calls by the more politically correct term "basic PCs." More surprising, in the past two weeks the company has staked out even cheaper ground--chips for everything from $500 network computers to $300 digital-TV set-top boxes. And behind Intel's deep-blue glass walls, engineers are designing processors for so-called backseat PCs--computers for playing games and cruising the Internet that could be built into cars by 2000.
If the Silicon Valley giant's grand plan to move beyond PCs hits the mark, the company's signature slogan could become Intel Inside Everything. Says Intel CEO Andrew S. Grove: "For us to walk away from a market whose size is going to be measured in tens of millions of units per year, maybe bigger, is inconceivable."
But what will selling cheaper chips do to Intel's amazing profit machine? Clearly, the strategy is a marked change in the company's longstanding practice of unveiling new processors at sky-high prices and then dropping them over a period of months or years. If low-cost gizmos become a big chunk of Intel's business, the company that set the standard for spinning silicon into gold could find itself grappling with lower gross margins--and massive changes in its vaunted business model.
That could be tricky. Intel spends big bucks to stay ahead of its rivals--some $4.5 billion this year on new chipmaking plants. Capital expenditures are forecast to climb even higher in 1998, to $5.3 billion, three times that of any other chipmaker in the world. That will help catapult Intel from No. 7 in worldwide production capacity today to No. 2 by 2002, analysts figure. But for Intel's bet to pay off in an era of lower-cost chips, the company must spit out even higher volumes. "It's a risk to go out and spend billions of dollars on these manufacturing plants," concedes Intel President Craig R. Barrett. "But if we didn't, we couldn't possibly reap the benefits. We're going down the road at 150 miles per hour, and we know there's a brick wall someplace, but the worst thing we can do is stop too soon and let somebody else pass us."
Grove has no intention of leaving the fast lane. His strategy is to keep Intel ahead of the pack while keeping profits high. The plan: adopting a tactic like that of scrappy PC makers such as Compaq and Hewlett-Packard Co., which have kept their gross margins well above ground while selling bargain-basement PCs. The trick is to compensate for thin profits on the low end with higher volumes--and with pricier models aimed at the lucrative technical workstation and server markets.
MARKET JITTERS. Grove has a high-powered arsenal ready to go. He plans to counterbalance cheap chips with soaring sales of powerful Pentium IIs, such as a 400-megahertz version scheduled for release by mid-1998, a 450-Mhz model later in the year, and the highly promoted 64-bit Merced chip expected in late 1999, all of which will be used in workstations and servers and could boast 90% margins. "It's very important for us to participate at both ends of the wire," says Grove. "I think the formula is going to work out."
Still, Wall Street is jittery. Only last April, analysts predicted Intel would post sales this year of $27.5 billion and earn $8.1 billion, up 56% from 1996, putting it on a path to become the world's No.3 profit maker. But since then, stock watchers have pared back projections: They now anticipate revenues of $24.9 billion. The revised earnings estimate, $6.8 billion, is still up a strong 31% from last year but short of the 44% and 54% growth spurts of the past two years.
Moreover, analysts figure that Intel's 1998 earnings will expand even more slowly--by 11%, to $7.6 billion, according to consensus estimates from First Call Corp. "The stock won't do well with that kind of earnings growth," says analyst Charles F. Boucher of UBS Securities Inc., who pegs Intel's 1998 profit growth even lower, at just 7.6%. Intel shares have already been hammered from an all-time high of $102 in early August to $77 today.
Does Grove agree that competing at the low end could douse Intel's earnings growth? "I don't know," he says with atypical uncertainty. But his usual go-for-the-jugular style quickly returns. "We are going to be motivated by participating in each of these segments to the fullest extent of our technical and marketing capabilities," he says.
Intel may have little choice. Analysts figure sales of sub-$1,000 PCs will climb 33% next year vs. growth of 20% for the PC market as a whole. Even more telling, computers selling for $1,500 or less could mushroom from 39% of the U.S. consumer market this year to nearly half of the market by 2001, according to market researcher International Data Corp. (IDC).
At the same time, longtime rival Advanced Micro Devices Inc. (AMD) and upstart Cyrix Corp. have marked this territory as their own, rolling out processors priced well below Intel's chips and grabbing market share--together some 20% of the low end vs. 10% in 1996, according to IDC. "It's the first time in many years that there has been a viable alternative [to Intel] at the low end," says IBM Senior Vice-President Samuel J. Palmisano, who has chosen AMD's K6 chip for IBM's new line of sub-$1,000 machines.
Grove is determined to cede no further ground. The hyper-aggressive CEO is legendary for his bet-the-company turnabouts. In 1985, when the market turned sour, Intel walked away from the business that launched the company two decades earlier: memory chips. And in 1994, Intel reversed itself to replace thousands of Pentium chips containing a minor flaw, a move that cost it $475 million. Grove has turned this tactic into a management philosophy that he calls "Only the paranoid survive."
Once again, Grove is showing his stripes. On Nov. 24, Intel was reorganized into five marketing and product groups, including a consumer unit to address the no-frills market. The company's next big step is expected in February. That's when analysts predict the chip giant will cover its flank from Cyrix and AMD by slashing prices by as much as 40% on its oldest Pentium MMX chips, to as low as $70, a price not seen since the waning days of the 486 chip.
But that's just the warm-up. Intel's real assault on the low end will center, oddly enough, around the company's thoroughbred, the Pentium II. Instead of pushing older technology, Intel will throw nearly all of its weight behind the Pentium II, even if that means eating a bit of crow and slashing prices faster than planned. In the first half of the year, Intel plans to ship a stripped-down Pentium II that forgoes the "cache" memory packaged alongside the chip, which speeds performance by keeping frequently used data close to the processor. Removing the cache will slow down the Pentium II but will shave about $15 off Intel's $103 manufacturing cost, estimates consultant Micro Design Resources Inc.
Intel could sell this chip initially for around $200, half its current price and low enough to be used in PCs costing less than $1,500--but not in rock-bottom models. A $200 price would take a toll on profits anyway. Instead of the estimated 74% margin that Intel earns on its cheapest Pentium II today, a stripped-down model could gross less than 60%. Later in the year, Intel will ship a redesigned Pentium II that restores up to half of the missing cache by building it directly into the processor. This design could largely restore the Pentium II's performance without adding cost.
Will less powerful versions damage the gold-plated Pentium brand? Not at all, say Intel execs. "It's like Coke," explains Paul S. Otellini, Intel's senior vice-president for sales and marketing. "One brand, many different products."
But why use the Pentium II at all? Why not crank up the Pentium MMX, barely a year old and scheduled to be put out to pasture by the end of 1998? Intel favors the Pentium II because it gives the chipmaker an edge over rivals. It sports a new scheme for connecting to a PC's main circuit board that cannot be copied by other chipmakers. If PC makers adopt the Pentium II, from servers on down to network computers, that could lock out Cyrix and AMD.
THE HITCH. Of course, the cheaper Pentium IIs would still cost twice what competitors charge for their chips aimed at the low end. "You can't give lip service to this market," says Steve Tobak, Cyrix' vice-president for marketing. "All Intel is doing is window-dressing."
Intel has an answer for that, too. It's planning to wring costs out of other parts of the PC so that computer makers can afford Intel's chips. In mid-1998, for example, Intel will unveil a range of new products, including chip sets, the companion to the processor, that will combine many PC functions into fewer chips. These products could slice the cost of making a PC by $50. That means Intel would only have to cut its Pentium II price to $150 to match $100 rivals.
There's a hitch, however. As sub-$1,000 machines become a bigger part of the market, analysts figure that the average selling price of Intel's chips will sag--from $235 this year to $220 in 1998. Every $10 fall in Intel's overall average selling price chops $900 million off its bottom line, according to analyst Vadim Zlotnikov of Sanford C. Bernstein & Co.
Grove is banking on different arithmetic. Intel already is making a dent in the server market, although mostly at the low end. This year, 97% of servers priced below $10,000 will have Intel Inside, as will three-quarters of machines costing $10,000 to $25,000. But Intel is only a bit player in an even more profitable segment: industrial-strength models costing up to $250,000. There, it competes against the muscle of Sun Microsystems and Hewlett-Packard.
That's where Merced comes in. This powerful 64-bit processor, which was co-designed with HP, features radically different technology that speeds software by running multiple tasks simultaneously. Merced will likely cost some $1,200 or more and will be aimed initially at top-dollar computers. By 2001, Merced could help Intel grab 41% of the high-end server market, predicts IDC.
The workstation market could be yet another cash machine. The fast rise of Microsoft Corp.'s heavy-duty Windows NT software is pulling Intel processors along with it. Indeed, from around 50% of the workstation market a year ago, Intel/NT-based systems will surge to 86% by 2000, says IDC. Revenues from workstation and server chips over the next three years could total $26 billion, analysts say.
Will that be enough to make up the margin squeeze from low-end chips? A BUSINESS WEEK analysis suggests that it may. Selling just one Pentium II for $750 produces up to 10 times as much gross profit as a $150 Pentium II. If Intel sells 36 million high-end chips in the next three years, it could reduce the average price of every other chip in its portfolio by an average of $80 and still come out ahead.
Indeed, Intel execs insist that their gross margins will remain above 50%. But, says one Intel insider, "if the sub-$1,000 category grew to more than 50% of the PC market, we couldn't sustain our gross margins."
Fear of collapsing computer prices at first drove Intel into denial about the low-cost-PC phenomenon. Until November, the company stuck to its well-honed marketing message, enticing customers to buy Intel's latest chips. That included a $100 million ad campaign this quarter to promote the Pentium II, including TV ads of chip-plant workers in clean room "bunny suits" dancing on Broadway.
But behind the scenes, Intel's managers were already hashing out a response to a market shift that had clearly caught them by surprise. The PC price collapse had been stunning. In January, 1996, a $1,300 machine from Circuit City Stores Inc. wasn't enough to run Windows 95 well: It bought only a 75-Mhz Pentium PC with 8 megabytes of RAM. A year later, the same money bought a 150-Mhz Pentium with 16 MB of RAM--ample enough to run Windows 95 and cruise the Internet. "Now, there's real meaty value available for less than $1,000," says Greg Gonzales, general manager of AES Technology, a small PC maker in Austin, Tex.
"GENIE IS OUT." Intel concedes that a permanent change has occurred in low-end PC pricing--similar, perhaps, to the 40% price drop triggered by Compaq in 1992 that slashed margins for both PC makers and retailers. "Once the genie is out of the bottle, you can't put it back in," says Otellini. Now, bargain-basement PCs are catching on in other countries and in businesses. PC makers, for example, plan to roll out $800 machines for corporate buyers after the New Year. But the crucial question, Otellini says, is whether low-cost PCs are bringing new buyers into the market.
Preliminary data from Compaq and Packard-Bell/NEC Inc. suggest that 40% of consumers are picking up their first PC. The rest are split between people replacing old machines and those buying a second or third PC for kids or parents. What's unclear--and worrisome to Intel--is how much cheap models are cannibalizing sales of costlier ones. "We won't really know this for two years," says Grove. But IDC analyst Kevin Hause disagrees. "It's absolutely happening," he says.
Cyrix and AMD couldn't be happier. Cyrix, for one, has developed a chip tailor-made for this market. Cyrix' MediaGX, at the heart of Compaq's Presario 2200, is not just a processor but also contains some chip-set and multimedia functions. Putting more features onto a thumbnail-size chip lets Cyrix market a product for just $81 that matches the oomph of $130 worth of Intel and third-party chips. Some analysts and executives argue that in the low-end market, all-in-one chips are more important than top performance. "The whole system has to go onto a single chip," insists Brian L. Halla, the CEO of National Semiconductor Corp., which recently acquired Cyrix.
Nonsense, says Grove. Intel tried combo chips in the late 1980s when it developed the 386SL for notebooks. But instead of making buckets of money, the 386SL was a disappointment because PC makers shunned it as too expensive and clumsy. Today, Grove keeps a poster of the 386SL on the wall of his conference room as a reminder of what not to do.
NEW TACK. There's another reason for his disdain. Putting extra chip functions into the processor makes it bigger, taking up valuable real estate on a wafer--the disk of silicon on which chips are etched. The more chips you can produce per wafer of silicon, the more money you make. Case in point: One eight-inch wafer of Intel's tiny 233-Mhz Pentium MMX chips contains an estimated 211 chips worth $125,000. The same size wafer of larger 180-Mhz Cyrix MediaGXs is worth only $8,100, says Micro Design Resources. That's why Intel doesn't want to squander space by adding features such as multimedia and networking.
Despite Intel's aversion to chip integration, the company may be forced in that direction anyway. The target wouldn't be basic PCs, but far cheaper devices such as set-top boxes and information appliances. "They can't afford to sit on their hands," says analyst Drew Peck of Cowen & Co.
Chips for these gizmos cost a fraction of PC chips. Players such as Hitachi, ARM, and MIPS Technologies sell processors costing just $15--and dominate the new markets for handheld PCs, smart phones, and digital cameras. To compete, Intel could be forced to opt for low price over the latest technology. But for now, Grove is sticking to his more-is-better philosophy. He argues that the growing processing demands on new devices will require the power of a Pentium-class chip.
Take next-generation TV set-top boxes. In a consortium with Cisco Systems, Oracle, and Netscape Communications, Intel is proposing a design for these new devices. Intel envisions a range of set-top boxes, from $300 models that will receive TV and offer basic menus to $500 models that add Net cruising, E-mail, and PC games. Initially, Intel plans to pump out Pentium MMX chips for set-tops, moving upstream to Pentium IIs by late next year.
The giant chipmaker will try an entirely different approach for network computers (NCs)--stripped down PCs that leave most of the heavy lifting to servers. Intel's pattern is to retire old chips and then rechristen them for use in products such as printers and network switches. But in a surprise twist, Intel now says it will use rechristened versions of the Pentium for NCs.
Why even muck with such low-margin devices, given that PC sales are on a rip? Because, even with burgeoning PC sales, penetration into U.S. homes will barely hit 60% by 2000, leagues behind TVs, VCRs, and CD players. "Our business depends on expanding the market," says Ronald J. Whittier, an Intel senior vice-president. "We want to be in living rooms, cars, appliances."
Now, Grove just has to make sure his formula works for keeping the Intel profit machine in overdrive at the same time.