If you were living anywhere in the United States or Canada in the late seventies or early eighties, you probably know what a Speak and Spell is. The educational toy, made by Dallas-based Texas Instruments, was a rectangular piece of brightly colored plastic that contained a keyboard and actually spoke to children: It gently corrected them when they spelled words incorrectly and praised them when they were right. With its preternaturally calm, synthetic voice, it helped teach a generation of kids how to spell. But the Speak and Spell, which was first sold in 1978, was much more than a child’s plaything. What made it work was a dazzling silicon-chip technology that was years ahead of its time. “We had created an integrated circuit that mathematically modeled the human voice,” says Gene A. Frantz, a senior fellow at Texas Instruments who helped invent it. “It was a significant breakthrough.”

It was something else too: an advance look at the technology that would later drive many of the industrial and consumer products of the Internet Age. In the late nineties it would propel TI, after more than a decade as an also-ran, to the vanguard of global digital technology. At the heart of the Speak and Spell was a type of integrated circuit that was the predecessor of what in the 1980’s would come to be known as the DSP, or digital signal processor. It had evolved from TI’s defense-related radar business. The difference between it and a microprocessor chip like Intel’s Pentium is that the Pentium performs the many tasks required to run a PC, while a DSP does one basic thing: It helps instantly translate information such as sounds and images from the analog world. DSPs run most mobile phones—TI’s chips are now in six out of every ten mobile phones sold—and are the brains behind such new-wave applications as high-speed digital modems, digital cameras, Internet audio players, and personal organizers like Palm Pilots. “DSP as a technology has now become the driver for the semiconductor industry and the whole dang world,” says Will Strauss, the president of the Arizona-based electronics market research firm Forward Concepts. “Everybody is trying to get into it.”

And it is TI’s clear mastery of DSP technology—and its dominance of the exploding DSP sector of the semiconductor market—that has put the company in a position not entirely unlike that of chipmaker Intel twenty years ago. TI introduced the DSP for commercial use in 1982. In the past four and a half years, TI’s chairman, president, and CEO, Thomas J. Engibous, has bet the entire company on the twin premises that DSP-related technology will drive the so-called post-PC era and that TI will lead it. He has sold off most of TI’s old-line businesses and in so doing dumped billions of dollars of relatively easy revenue for a chance to bet on the digital future. “We are convinced that we are providing the fundamentals for the Internet Age,” says TI’s chief financial officer, William A. Aylesworth. “We are at the center of it.”

It wasn’t always that way. TI, in fact, has looked more like a lumbering, slightly befuddled giant over the past twenty years than the hungry, bleeding-edge tech company that now believes it can dominate the next decade in the semiconductor business. Texas Instruments, of course, is the company that brought you the Computer Age. TI engineer Jack St. Clair Kilby’s groundbreaking work on the integrated circuit chip in 1958, for which he won the Nobel prize last year, was the basis of the single most important technology in the digital revolution. And in the late fifties TI, which was the first company to market an integrated circuit chip, looked a lot like a millennium-era dot-com: Its stock rose 4,600 percent on news of its breakthrough product. Like many other successful companies of that time, TI diversified, adding new businesses: defense electronics, radar and missiles, memory chips. In 1974 it stunned the world again with the introduction of the handheld calculator, and by the eighties it was in the business of building computers as well.

The problem was that by the nineties TI was doing a lot of things, but few of them well. It had led the semiconductor industry as late as the mid-eighties, only to lose its market dominance to fierce competitors like Intel and Samsung. Its defense businesses were slow growing and bureaucratic; its computers never took off. In the decade ending in 1995 the company had grown at a somnolent 7 percent a year, light-years behind the big technology high-fliers like Dell and Microsoft. In 1994 TI had about $10 billion in revenues and a deeply embarrassing market value of a mere $8 billion.

It was around then that the transformation began. TI chairman Jerry R. Junkins started it, revving up the sleepy defense-and-government-contract culture and selling off its oil-field services business (on which the company had been founded in the thirties) in 1991 and its midsized computer business in 1992. Though many of TI’s businesses were slow, stodgy, and non-tech driven, the company had somehow retained the underground mentality that had caused it to be a leading technology company in the first place. For reasons no one could quite fathom, and in the absence of any real market to speak of, TI engineers like Gene Frantz and the old Speak and Spell gang had been allowed to labor away at DSPs at TI’s Houston unit for years. “We had a skunk works here in Houston,” says TI senior vice president Michael J. Hames. “The average age was between twenty-five and thirty. It was like a start-up culture, passionate beyond common sense.”

The Houston unit became ground zero for TI’s DSP operations, which in the mid-eighties made up a mere one percent of TI’s revenues and consisted mainly of sales either to big telecommunications companies or to the military for radar and sonar applications. The first year the DSP unit broke even was 1987. Few people in the world, outside of TI’s Houston unit, believed in DSPs. Still, it was a culture that no amount of mediocre management seemed to be able to kill.

While the DSP unit pushed resolutely forward in Houston, in another far outpost of the empire—Nice, France—a quiet revolution was under way. In the late eighties and early nineties TI’s Nice operation had become a true believer in DSPs: The problem was how to sell them to the burgeoning telecommunications and cellular phone markets. In 1992 the search eventually led TI’s engineers and marketers to an extraordinary gathering in a forest in Finland that would change the world of wireless phones. During a three-day meeting, employees of Texas Instruments and the Finnish telecom giant Nokia worked together, ate together, took naked sauna baths together, and hashed out the protocols of what would become the next generation of mobile phones: digital instead of analog, a quantum smaller than the first generation of phones, and able to hold a battery charge for days. “We put together a revolution in the architecture of cell phones,” says Gilles Delfassy, TI’s director of worldwide wireless communications who was responsible for the TI-Nokia collaboration. “It was very ambitious and aggressive architecture, and it was pushing technology to the limit. It was a completely new. What we defined was copied by everybody.” That innovation not only propelled Nokia to the forefront of the world’s mobile phone makers but also proved once and for all that TI’s DSP technology was a winner and helped persuade TI’s management to shift the focus of the company’s business.

The major seismic shifts at TI started coming after Engibous became president and CEO, in 1996. Convinced that DSPs were the way of the future, he sold off notebook computers; missiles, radar, and other defense products; and TI’s huge memory-chip business. In roughly four years TI unloaded 14 businesses that had once provided the majority of its revenues and picked up 21 mostly small, hungry, tech-driven companies that provided the sort of cutting edge technology TI was looking for. In the process the entire company took on Houston’s old skunk works ethos. “The culture is much more entrepreneurial and much more fiercely competitive,” says Engibous, an engineer by training who has spent his career in TI’s semiconductor business. “The idea we promote is that it’s okay to start out and find you’ve gone in the wrong direction. Don’t worry about it if you’re wrong.”

The new TI gets fully 85 percent of its revenues from the DSP and analog-chip businesses. Analog chips usually work in tandem with DSPs: They take real-world information such as a human voice and turn it into digital signals, which DSPs then compress and decompress. Though TI still has a few other lines of business, it has focused on this particular industry like no other major semiconductor company. “There is no question that everyone understands that this is the sweet spot of the semiconductor industry,” says Engibous, “and that a large number of companies are putting a lot of their efforts there. But we’re not just putting a lot of effort there. We have dedicated our lives to it, and there are no other big companies that are doing that.”

So far the gamble has paid off. Texas Instruments has a dominant 48 percent share of the $4. 6 billion global market for DSPs and a leading 10. 8 percent share of the $26 billion analog-chip market. TI is expected to make $12 billion in revenue for the fiscal year ending in March, which just equals the revenue it made in 1996, before the sell-off of its main lines of business. More important, TI’s profits have more than tripled, to an estimated $2. 1 billion this year. TI also has a huge advantage in what is known in the business as “installed base of software.” Because the DSPs that TI makes are programmable—that is, the customer can write his own software onto the TI chip to make it compatible with a range of different products—that software then becomes an important and integrated part of the chip, making it difficult for a maker of, say, mobile phones, to switch DSP suppliers easily. TI is also helped by the fact that TI-designed software is used in more than nine hundred universities around the world to teach DSP technology to engineers. The company is also marketing aggressively to tens of thousands of smaller companies, a move it believes will give it a lock on those companies’ business when they become larger. The stock market has taken notice: From 1996 to 1999 TI’s stock outperformed that of Dell, Microsoft, and Intel.

Still, TI faces a tough market, one that promises to get only tougher as more companies—Intel included—jump into the DSP game. TI’s major competitors now are Lucent, with 25. 1 percent of the market; Motorola, with 11. 4 percent; and Analog Devices, with 10. 3 percent. All are feeding off the booming DSP market that is expanding at a rate of 40 percent a year, and all have their sights set on the exploding market in consumer products, from broadband modems to digital cameras to Internet audio players to the third generation of mobile phones, which will have wireless Internet terminals and Internet audio capabilities, among other things. They are selling DSPs as well to a whole new range of clients: makers of vacuum cleaners, refrigerators, and air conditioners, who can use DSPs’ ability to control electric motors.

Of course, not all of these markets are guaranteed to pay off. Even mobile phones, which have been a cash cow for TI so far, are no sure thing. “That fact is that 3G [third generation] cell may not be all that big,” says Tom Kurlak, who manages a technology hedge fund for New York City-based Tiger Management. “The question is, What problem are you solving? The first cell phones allowed people to be mobile. But now it’s just embellishments, and those may never appeal broadly to the masses.”

For the moment, however, TI is the top dog in a DSP market that is expected to grow from an estimated $6 billion this year to $19 billion in 2004. If that prediction holds true, it would likely mean that TI’s revenues from DSPs alone that year would be $9 billion. And that, say a growing number of analysts and industry watchers, is why Tom Engibous’ spectacular bet just might pay off.