For the sake of accuracy, perhaps Samsung should rename its chip division Samsung Memory. It's the world's second biggest chipmaker behind Intel. But with each passing year, the Seoul-based company seems to consolidate its position in every type of memory IC under the sun. In fact, memory chips—including DRAM, SRAM, and flash—account for some 85 percent of Samsung's $10.5 billion in semiconductor sales.

Chang-gyu Hwang, CEO, Samsung Semiconductor

Samsung has rocketed from zero market share in NAND flash a few years ago to No. 1 today, controlling 65 percent of the world market. In 2004, Quinn estimates that about 35 percent of all Samsung's memory revenue will be derived from flash.

The key drivers for NAND flash are memory cards (largely for digital still cameras) and USB flash drives, which are replacing the ubiquitous floppy disk in PCs and notebooks. USB flash drives have seen triple-digit growth over the last two years, and Samsung expects that to continue into 2006, Quinn says.

He adds that opportunities for NAND flash in mobile applications are opening up and believes Samsung is in lockstep with handset trends. As 3G infrastructure rolls out and mobile phones turn into feature-packed phones, data-storage needs will increase, creating stronger demand for NAND flash.

"With smart phones and media phones, we believe the architecture will migrate from NOR to NAND and open additional market opportunities for us," Quinn says.

Samsung may soon have a harder time dominating the NAND flash business, however. Several entities, such as Infineon Technologies, Micron Technology, Renesas Technology, and a duo combining Hynix Semiconductor and STMicroelectronics, are now moving into the market in a big way.

In response to the competition, Samsung will keep pushing the limits of flash technology. "From a cost-per-bit basis, competitors must have similar capability to compete," Quinn says.

And Samsung considers technology its strength. In September, the company announced the industry's first 8-Gbit NAND flash memory, built at 60 nm, for applications such as mobile storage. The leading-edge memory will enable the construction of a 16-Gbyte memory card. The new IC doubles the capacity of the 4-Gbit IC Samsung introduced in 2003.

Hwang earned an MSEE from Seoul National University in Korea and a PhD in electrical and computer engineering from the University of Massachusetts, Amherst. He once taught at the Korea Naval Academy and served as a consultant to companies such as Hewlett-Packard and Intel before joining Samsung in 1989 as director of device technology development.

Samsung Semiconductor is one room in the larger house of Samsung Electronics, which makes products as diverse as mobile phones, PCs, LCDs, DVDs, notebooks, and printers.

Nam Kim, senior analyst with iSuppli, says Samsung has made vertical integration work to its advantage. Nokia, for example, buys NOR flash memory from Samsung, even though Samsung's NOR output is tiny. That's because Samsung puts its various products in a basket and negotiates the collective price. So if Nokia buys NOR, maybe flat panels will be discounted.

Such actions require multiple layers of management to act quickly. "Samsung has an amazingly fast decision-making process, even though it is a very complicated business model," Kim says.

Kim adds that as technology leader, Samsung even stepped ahead of Intel by developing 60-nm process technology, which Intel has not yet announced, Kim adds. Honing process technology ties in with Samsung's capital-investment strategy. The company outspent everyone this year. As consumer gadgets evolve and demand higher-density flash memory, Samsung can move faster to market than rivals who are behind in the latest processes, Quinn says.

Management has found novel ways to bring efficiency to the cash-eating memory manufacturing business. One example is dual-purpose fabs. Samsung can use the same DRAM manufacturing infrastructure to make flash memory, providing a natural hedge against cyclical dips in one market segment. If the DRAM market is slow, flash can be built in the same facilities with minimal additional capex.

Yet Samsung's impressive track record is still dependent on the capricious memory market. DRAM earnings, which typically get whipsawed by price changes, account for about 50 percent of Samsung's chip revenue. The cost of memory production is also sky high. In capital equipment, Samsung this year is forecast to spend $7.8 billion, more than all the world's chipmakers.

So far, the company has managed well. Dan Hutcheson, president of VLSI Research, believes Samsung's strength in manufacturing derives from very short turnaround times and higher yields than rivals can match. "Samsung was the only DRAM maker profitable through the downturn because they're just really good at yield management," he says.

But the company's managers apparently realize that memories are too precarious to pin the business on. They have tried, without much quantifiable success, to hedge their bets by developing non-memory expertise.

Non-memory ICs, such as logic parts, generate stronger and more stable revenue than memory chips, but are far more backbreaking to master. Samsung has tried to get into microprocessors for several years, with patchy results.

"It's really difficult to get out of memories," Hutcheson says. "Once you're in, you're in. It's like the mafia."

However, Samsung plans to ramp its first 300-mm wafer fabrication facility dedicated to logic devices in late 2005, aiming at products that complement the company's efforts in NAND flash for advanced cell phones.

The fab can't come online fast enough. In 2005, the memory market is expected to slow dramatically, down from 47 percent growth in 2004 to 5 percent, according to iSuppli. The NAND flash market will see negative growth due to price pressure from additional competition entering the market.

"NAND is moving more like DRAM, it's a more volatile product than before," Kim says. "The business environment could impact Samsung's growth next year."