If Eli Harari had followed his natural inclination, he might still be in the lab inventing the next great memory chip technology. The technical achievements of this Princeton-trained materials scientist include more than 100 U.S. and foreign patents as well as the invention of electrically erasable programmable read-only memory (EEPROM) technology during the 1970s.
Yet for the past 18 years, as founder and CEO of flash memory chip maker SanDisk, Harari has devoted most of his time to an arguably more challenging task: ensuring the survival and continued expansion of one of the technology industry's fastest-growing, most successful companies.
"I still love inventing," he says, "but a lot of what I do as CEO has nothing to do with that."
Indeed, although Harari enjoys visiting with SanDisk engineers and reviewing their projects and patent applications, most of his attention these days is devoted to a wide array of other matters. In the space of a few hours, his activities may range from overseeing recent acquisitions and the company's $5 billion manufacturing expansion plan to dealing with Sarbanes-Oxley regulatory compliance or even commenting on what fruit is served in the cafeteria at the company's Milpitas, Calif., headquarters.
"Any day, I have to deal with 30 to 40 things going on, more or less at the same time. It's truly a multidimensional chess game," he says. "If I didn't have the right people here, I'd go crazy." Fortunately for his sanity, and the company's health, Harari is able to delegate many decisions to a dedicated team of senior and midlevel managers who, he says, deserve much of the credit for the company's success.
The challenges probably are similar for chief executives at other successful, multibillion-dollar companies. But the fact that Harari is still running the company he cofounded nearly two decades ago puts him in select company, along with the likes of Apple Computer's Steve Jobs and Oracle's Larry Ellison. Relatively few companies achieve the kind of success SanDisk has experienced, and fewer still manage to do it with their original founder at the helm.
Harari says that managing SanDisk's growth has required constant learning and adaptation. "When we were a $500 million company, I had never run a $500 million company. When we were a $1 billion company, I had never run that," he admits. "So basically it's an adventure."
And what an adventure it's been. During the past five years, SanDisk's annual revenues have multiplied almost tenfold, from $366 million in 2001 to nearly $3 billion expected in 2006. Since it weathered the tech industry's 2001 downturn, the company has remained solidly and consistently profitable, despite the need for heavy manufacturing investments and the pressures of a hypercompetitive market in which flash memory chip prices invariably decline by 50 percent or more each year.
SanDisk also had one of the technology industry's best-performing stocks during 2005, when its share price tripled to nearly $80. (Granted, it lost half that value during mid-2006, before rebounding back into the $50 range.) Institutional Shareholder Services rates its corporate governance practices as better than those of 95 percent of all other technology hardware and equipment companies, and its 2005 return on capital employed was an impressive 17.3%.
In addition, SanDisk has exerted significant influence on the electronics business. A flash memory technology pioneer, it either invented or codeveloped most of the industry's standard memory card formats—including CompactFlash, MultiMediaCard, SD, miniSD, Memory Stick Pro and TransFlash—which, in turn, have enabled the development of a variety of new portable electronics products, from digital cameras to multimedia cell phones.
Although SanDisk is now one of the semiconductor industry's rising stars, it was far from an overnight success. In fact, it took years of persistence and often unconventional strategies to sustain the company through its early lean years.
Harari recalls being laughed at initially when he and cofounders Sanjay Mehrotra and Jack Yuan approached potential investors in 1988 with a plan to use memory chips as a rugged, low-power alternative to hard disk drives. Venture capitalists were interested in the concept of solid-state data storage, but they balked at Harari's insistence on developing a new kind of flash memory chip rather than using existing chips from Intel or other suppliers.
Rejecting several early offers that would have compromised that vision, Harari sank $15,000 of his own money into the startup, which originally was called SunDisk. Venture capitalist Irwin Federman eventually provided $250,000 in seed money, and in late 1998, Federman, now SanDisk's vice chairman, led a group of investors—including Oak Ventures, Western Digital and AT&T—in an $8.5 million first round of financing.
Within two years, SunDisk had developed its first 4-megabit flash chips and outbid both Intel and Texas Instruments to land IBM as its initial customer. IBM ordered 10,000 of the company's Flashdisk drives, which crammed hundreds of chips totaling 20 megabytes of storage capacity into a device that emulated a 2.5-inch disk drive. At $1,000 each, however, the drives proved as unsuccessful as the IBM handwriting-based portable computers in which they were used.
By 1995 SunDisk was working with the likes of Kodak, Canon and Polaroid to develop specifications for a succession of removable flash card standards. The company also went public on the NASDAQ exchange that year, after changing its name to SanDisk to avoid confusion with Sun Microsystems.
But the digital camera market Harari had been counting on proved stubbornly slow to arrive. For several years, SanDisk stayed afloat by developing custom memory systems for the likes of NASA's Space Shuttle program. "A quarter of a million dollars here, half a million there—this kept us alive until eventually we got lucky," says Harari.
SanDisk's early-2000 decision to form a joint venture with Japan's Toshiba was a crucial turning point. The partnership, called FlashVision, combined SanDisk's multilevel cell (MLC) technology—which allows multiple data bits to be stored in each flash memory cell—with the Japanese chip giant's expertise in NAND flash, which cost less and recorded data faster than the then-dominant NOR technology. The companies initially agreed to jointly invest $700 million to equip an existing 200-millimeter Toshiba fabrication plant in Virginia to make NAND flash chips, which would be equally shared between the companies.
Previously, SanDisk had been strictly fabless, relying on UMC, Tower Semiconductor and other foundries to build its chip designs. But after enduring severe flash chip shortages, company officials realized that they needed to develop their own chip-making capabilities. "It became very clear that the winners in this business ultimately would be those that owned the R&D and the flash manufacturing," says SanDisk's president and cofounder, Mehrotra, "because that's what would allow them to produce memory at the lowest cost."
By the time FlashVision started production in mid-2001, however, the tech industry was in a downward spiral and SanDisk's sales were plunging. The company lost nearly $300 million that year and was forced to lay off 40 percent of its employees. But it stuck with the Toshiba joint venture, which has since grown into what Mehrotra calls "perhaps the most successful partnership in the semiconductor industry." Today the companies jointly operate two additional fabs at Toshiba's Yokkaichi, Japan, manufacturing center and are building another 300-mm facility scheduled to start production there in late 2007.
SanDisk should account for 10 percent of global NAND production in 2006, according to Merrill Lynch, and Toshiba adds another 19 percent. Yet the two companies aren't content with their current market positions. They plan to spend a combined $10.8 billion on additional NAND production capacity from 2006 through 2008, significantly more than industry leader Samsung, which, Merrill Lynch forecasts, will invest $8.3 billion during the same period.
Although SanDisk's manufacturing costs are increasing rapidly, the ability to make its own leading-edge chips is proving invaluable to the rest of its business. By producing most of its own chips, the company claims, it can build flash memory cards for less than half of what its costs competitors, most of which pay a hefty markup to buy chips from other suppliers.
Toshiba and SanDisk get further cost advantages from having converted nearly all their production to lower-cost MLC technology while the rest of the industry struggles with the technically challenging transition. Even Samsung is using MLC in only about 60 percent of its chip production thus far. "Clearly multilevel cell technology is a core competency and a real differentiator for SanDisk," says analyst Doug Freedman with American Technology Research (ATR).
Another SanDisk advantage that is sometimes overlooked, and which analyst Richard Whittington of investment bank Caris & Co. calls the company's "crown jewel," is its expertise in designing flash controller chips.
Most products that use flash memory require controllers to transfer data in and out of the chips and handle error correction. Whereas other NAND makers require customers to buy controllers from third-party suppliers, SanDisk develops most of its own, which improve its chips' performance and shorten the time to market of its customers' products.
SanDisk also manages to profit from the entire industry's growth, thanks to its formidable portfolio of flash-related patents, which it defends vigorously and licenses to many of its competitors. Its license and royalty revenues, already bringing in more than $300 million annually, are expected to reach $500 million next year.
Vertical integration
Clearly, SanDisk has become more than a typical fabless chip maker. Increasingly, it is starting to resemble a vertically integrated consumer electronics company, with control over its supply of key components and a presence in multiple end markets. "It created the flash card business," says Semico Research analyst Jim Handy, "but it's remade itself into a consumer company."
Whereas most of its chip competitors sell their output to manufacturers of digital cameras, phones, music players and USB flash drives, SanDisk gets about 70 percent of its product revenue by selling removable flash memory cards and other products under its own brand name at nearly 200,000 worldwide retail outlets. Semico estimates that SanDisk's $1.6 billion in 2005 flash memory card sales accounted for 26 percent of the world market, more than twice the 10% share of its nearest card rival, Sony.
SanDisk's latest consumer play is a line of digital music players. As of June, its Sansa players were the second-best-selling U.S. brand, according to research firm NPD Group. Although SanDisk's 10 percent market share lags far behind the dominant 76 percent of Apple, with its iPod players, the company has been taking significant share from other competitors. "SanDisk has positioned itself as the 'other than iPod' MP3 supplier," says Caris analyst Whittington.
The company's sales of USB flash drives should benefit from its July agreement to buy Israel-based msystems for $1.55 billion. msystems pioneered the USB drive market and is also a leading vendor of embedded flash for mobile phones. The deal, expected to close late this year, should help keep SanDisk's new factory space fully used and make the combined company an even stronger NAND supplier for both cell phones and USB drives.
New competition
Analyst Tristan Gerra of brokerage Robert W. Baird says that SanDisk's vertically integrated strategy boosts the company's profits and its stock price. Yet he doesn't expect the strategy to remain unique much longer. "Because this is a good business model, you're going to see other companies adopt it as well."
Micron Technology took a step in that direction in March by paying $680 million for Lexar Media, ranked by Semico Research as the third-largest flash memory card vendor, with 2005 sales of $526 million. The deal should allow Lexar to procure lower-cost flash chips while giving Micron another outlet for its rapidly expanding flash production.
In 2005 Micron and Intel formed a NAND manufacturing joint venture called IM Technologies. Although it is still a relatively small player, Merrill Lynch expects the joint venture's output to increase from its current 5 percent of worldwide output to 9 percent by 2008. By then, it estimates, the two companies will have invested more than $4 billion in new manufacturing capacity at fabs in Virginia and Utah.
Competition from the world's largest chip maker is always a concern. But Satya Chillara, an analyst with Pacific Growth Equities, sees a potentially bigger threat from Samsung, which he says could get "cranky" about the aggressive capital spending by its smaller rivals and decide to accelerate its own NAND expansion plans. The industry already has seen faster-than-usual price declines during 2006, as production outstripped market demand. But Samsung, which can shift production between NAND and dynamic random-access memory, could easily put the market into a drastic oversupply.
UBS Securities analyst Alex Gauna agrees that overproduction is the biggest short-term risk for SanDisk. Thus far, music players, cell phones and other new markets have been soaking up everything the industry can produce. But, as Gauna notes, "Even strong companies suffer when a market becomes commoditized."
Scaling challenges
Longer-term, analysts worry about how long the industry can continue scaling down its technology to produce smaller, less costly chips. SanDisk has just converted all of its production to chips using 70-nanometer circuits and plans to transition to 56 nm during 2007. Micron and Intel plan to begin making 50-nm chips next year. But beyond about 30 to 40 nm, it's unclear how much smaller the industry can make NAND chips before they no longer store data reliably.
Harari admits that San-Disk would be in trouble if it lost the ability to reduce NAND costs. But like the veteran inventor he remains at heart, he's confident that future technology advances will keep the company on a rapid growth trajectory. For starters, company engineers are working on new MLC technology to store 3 and even 4 bits per memory cell. msystems recently announced a 4-bits-per-cell technology dubbed "x4" that it claims will be ready for production in 2007.
Aside from MLC, Harari also has high hopes for the memory technology that came with SanDisk's $250 million acquisition of Matrix Semiconductor in 2005. Although Matrix's dense, vertically stacked memory cells permit only one-time data storage thus far, SanDisk is investing heavily to develop a rewritable version of the technology. "We don't have that breakthrough today," he admits, "but that could be very disruptive in another three to five years."
In the meantime, he says, SanDisk is just getting started. Harari's five-year plan is for his company to become one of the 10 largest semiconductor makers and one of the world's most influential consumer products companies. Those are lofty objectives, he admits, but no more challenging than his original company goal: "Changing the world—making a difference through better products and cheaper products." On that count at least, he may finally be able to declare "mission accomplished."
Click here for more information
