SEHAT SUTARDJA, HIS

wife, and his brother started

Marvell Technology Group

Texas Instruments

STMicroelectronics

National Semiconductor

"We were engineers, very focused, very optimistic about coming out with the best product, so we never thought for a minute about the challenges in the market," says Chief Executive Sutardja. "The thought never crossed our minds."

As it turned out, Marvell's disk-drive chips leapfrogged rivals', and the Santa Clara, Calif.-based outfit grew past annual revenues of $1 billion in just seven years. The company added chips for networking, imaging and wireless. By early 2006, the stock market valued Marvell at better than $20 billion, and its founding family members were multibillionaires.

Then the fast-moving chip maker stumbled, in classic Silicon Valley fashion. Acquisitions weighed on profit margins. The company falsely backdated stock-option awards. After an internal investigation concluded that Sutardja's wife, Weili Dai, had participated in the backdating, she stepped aside from her key executive posts. In the last two years, Marvell shares have fallen by two-thirds, to a recent $11.

But Marvell's prospects remain pretty darn good. It is addressing its management problems: The company is working hard to fill a vacant chief-financial-officer slot, and it hopes that a deal with regulators will allow Dai to resume her executive role.

Expenses have leveled off, so profits should grow as Marvell chips ride successful products like disk drives from Western Digital, and the BlackBerry handset from Research in Motion.

Meanwhile, Marvell has chips for large new markets like DVD players and LCD televisions. In a couple of years' time, these developments could cause Marvell's earnings and its shares to double.

famous for helping start firms like SiRF and S3). They knew they were getting into the semiconductor business late in the game, says Sutardja, "so we were focused on developing technology that was three or four years ahead of the competition."

Fortunately, Sutardja was a gifted designer of processors, as well as "mixed-signal" chips, which combine analog and digital circuitry to handle wavy analog signals as well as the zeros and ones of digital logic. Marvell developed a chip for hard-disk drives that was small, cheap and very fast at reading the densely-packed bits of data on the drive platter. It took a couple of years of sales calls, but they landed disk-giant Seagate Technology Holdings as a customer. Marvell sold $21 million worth of chips in its first year of shipping.

That was 1999. Marvell went public the next year amid the crash of the tech-stock market. But the company's growing revenues helped it buck the ebb tide, as nearly all the hard-drive makers adopted Marvell chips. The company had a knack for entering established markets with products that were a lap ahead of rivals. Chips for data networking and Wi-Fi hardware won business at Intel and Cisco Systems. In five years, Marvell revenues rose nearly six fold, to more than $1.6 billion for the fiscal year ended January '06.

Despite its fast growth, Marvell in some ways still resembled a family business. Sutardja's brother Pantas oversaw chip design, while his wife Dai was in charge of Marvell's communications products. The business was the focus of their lives. Sehat Sutardja says that he sleeps and breaths electronics, dreaming up patentable ideas in his free time. Dai worked just as hard. "Sometimes I'd even complain to her," says the CEO. "Why are you still at 11 o'clock at night talking to customers in Taiwan?" In Marvell's early days as a public company, the three founders took no stock options themselves. But they were plenty rich, owning 30% of the company.

With Marvell shares at their peak levels in 2006, the company arranged the brave purchase of Intel's cell-phone chip business. Intel had spent billions to establish its XScale chips in a wireless market dominated by TI,

Qualcomm

Motorola

That's when the bomb fell on Silicon Valley and on Marvell, which was full of options-backdating tinder. With the Securities and Exchange Commission and a federal grand jury knocking on the door, the company conducted an internal investigation. Its inquiry found that three-quarters of Marvell employee stock options had been backdated or otherwise falsely booked, to obtain cheaper exercise prices.

Backdating had understated Marvell's expenses by more than $360 million over 10 years. Earnings for the fiscal year ended January 2006, for example, had to be restated from $331 million to $200 million, or from $1.05 a share to 32 cents (although pro forma were $397 million, or 63 cents a share, without option expenses). When numbers for the January '07 year finally were ready, they showed a sales increase of 34% to $2.2 billion, but option accounting led to a loss of $12 million, or 2 cents a share. Most tech investors still ignored option expenses of a company like Marvell, so in the eyes of Wall Street, Marvell had January '07 pro forma earnings of about $359 million, or 56 cents.

More problematic for Marvell's future, the investigators found that Dai had been involved in selecting falsely recorded dates. Marvell's investigators recommended that she be sacked with the other executives responsible for the backdating. In May 2007, Dai gave up her executive titles, but she didn't leave Marvell. The investigators said that neither she nor the Sutardja brothers had received any of the options that she helped backdate. Marvell's family leaders did get other options that were backdated; those had been awarded by a separate compensation committee. Chief executive Sutardja and Dai gave back $8.8 million in backdating gains. Last month, they began foregoing their salaries.

The SEC has warned Dai that the agency was considering an enforcement suit against her. Sutardja says Marvell's working hard to resolve its backdating mess, and he hopes Dai will be able to resume her full duties at the company. "It's hard," says Sutardja of his wife's predicament. "She has always been focusing on the company 99.9% of the time....For her not to be involved in running the day-to-day business is not for the benefit of the shareholders."

Sutardja is also spending a lot of time finding a permanent replacement for Marvell's chief financial officer, who resigned during the scandal. When the company hires a new CFO and settles the backdating issues, shareholders will surely feel more optimistic about Marvell's stock.

Marvell's business merits optimism. Its core products for data storage are becoming ever more popular, and it will soon be shipping new offerings for consumer electronics and third-generation smartphones that combine Marvell's technologies for power-thrifty processing and wireless communications.

nvesting in new products has weighed on Marvell's operating margins. Gross margins have also slipped in recent years, from 53% to 48%. But help is on the way. As a "fabless" semiconductor company, Marvell has always had its chips fabricated at others' semiconductor plants. An Intel foundry continued to make Marvell's cellular chips even after Marvell bought Intel's cell phone-chip business in 2006. But Intel's costs were relatively high. Starting about a month ago, Marvell began having all its new cellular chips made at Taiwan Semiconductor Manufacturing at far lower cost.

The XScale cellular chips power a lot of high-end cell phones, like the Palm Trio, the Samsung BlackJack and the RIM BlackBerry. Workaholics have been impatiently waiting for a BlackBerry that can run on high speed 3G cellular networks, and investors have fretted at every rumor that Marvell is having problems with its 3G chips.

When a 3G BlackBerry finally appears with Marvell chips inside, investors will rejoice. Smartphones like the BlackBerry have comprised a relatively small number of the billion cell phones sold annually, but the segment's sales are doubling every year. Marvell hopes to bring fancy smartphone features down-market, where annual unit sales number in the hundreds of millions. At unit-sale prices of 10 bucks per chip, Marvell could add billions of dollars in new sales.

Meanwhile, the company's having continued success with the hard-drive chips that still bring in almost half of its revenues. In the 10 years since it began selling chips to read disk drives, Marvell has integrated other drive circuits in what the industry calls a "system on a chip." Such products have been adopted in just about every hard-drive product, including those of Samsung Electronics, Toshiba and Western Digital the last of which has been on a tear, with drive sales jumping nearly 50% in the December '07 quarter.

The company's disk-drive chips will use an important new higher-speed signal-processing scheme called iterative decoding this year, says data-storage boss Alan Armstrong. The new technology will allow hard-drive makers to continue increasing the storage capacity of a disk drive well beyond today's upper limit of a trillion data bits. Marvell will also see revenues from the first products that adapt its magnetic hard-drive technologies to optical-disk devices like Blu-ray DVD players.

Another Marvell division supplies the brains for nearly every printer on the planet. And the company owns the market for Wi-Fi chips in portable consumer-electronics gadgets like cameras and game consoles.

If Sutardja's company can get its corporate-governance problems behind it and deliver products for 3G cellular and consumer electronics, future-year earnings could leap from the washed-out level of around 35 cents a share that analysts expect for the just-ended January 2008 fiscal year.

The Bottom Line

Marvell shares could double in two years if it fixes its options mess, controls costs and gets a 3G-capable chip to market.

Some buy-side investors are betting that Marvell can earn $1.00 to $1.50 if it executes well in the fiscal year ended January 2010 (again, not counting options). Sangeeth Peruri, of J&W Seligman, thinks such earnings would return the forward-earnings multiple on the stock from today's 16 times earnings back to the level of over 20 times that Marvell enjoyed in years past. If so, Marvell's shares could more than double from 11 today.

"Revenues are growing, driven by a lot of product cycles," says Peruri. "You could get 30-50% annual earnings growth for the next three-to-five years."