(FORTUNE Magazine) – As attempted billion-dollar heists go, this one was unobtrusive. There wasn't a pile of cash in the room or a trove of jewels stashed in a vault. The target was far more ethereal. Inside a banquet room at the Crystal City Stouffer Hotel outside Washington, D.C., a computer industry committee was debating elements of a new standard for memory chips. As happens in these settings, heavyweight competitors--IBM, Hewlett-Packard, Samsung, and Toshiba among them--were working together to decide the shape of the next generation of chips.

Also present at those mid-September meetings in 1992 were two representatives from a tiny technology-design company called Rambus. They listened as their industry colleagues discussed an element known, in typically impenetrable techno-gibberish, as "programmable CAS latency." A week after the meeting, one of the two Rambus staffers, Richard Crisp, met with a company attorney to talk about amending Rambus' pending patent applications. Among the new technologies that Crisp wanted to add: programmable latency. A few months before, Crisp had recommended adding patent claims for "mode registers"--right after they were debated at the same standards committee. Only a few months before that, it was "low-voltage swing." All those technologies made the same quiet journey from the standards committee agenda into Rambus patent applications.

Nearly a decade later, long after the standards had been adopted, Rambus--a Los Altos, Calif., company with 175 employees and $72 million in revenues--decided it was time to cash in. In late 1999, citing its patents, which by then had been issued by the U.S. Patent and Trademark Office, it began demanding royalties on technology that represents 80% of the $32 billion market for chips known as dynamic random-access memory (DRAM). Incredibly, Rambus--which designs its own version of DRAM technology--was attempting to claim ownership of a competing DRAM design, one that Rambus had long maligned as inferior. Rambus' design was a Ferrari, to use the company's own analogy; its competition a Volkswagen. Rambus wanted to be paid not just for the Ferraris, but also for the Volkswagens.

That audacious strategy is now backfiring spectacularly. Hailed as "the year's hottest IPO" by the Wall Street Journal after it went public in 1997, Rambus' stock was long a favorite of chat-room denizens and CNBC devotees. But Rambus is spending a hefty portion of its revenues as it battles its way through 11 separate patent-related lawsuits in the U.S. and Europe against three much larger, wealthier opponents. In the first case to go to trial, this past May, Rambus was soundly trounced; a jury in Richmond found that the company had defrauded Infineon, a spinoff of German giant Siemens. Though Rambus declined interview requests (and would answer only a handful of questions via e-mail), a review of 3,000 pages of transcripts and exhibits from the trial shows that the company plotted to gather patents on standardized technology for years, even as it participated in a standards committee that requires companies to disclose their patents. It's no surprise, then, that Rambus stock has plummeted more than 90% from its June 2000 peak.

Tales of company downfalls have become depressingly familiar in the past few years, of course. But Rambus is no flaky dot-com. It's a profitable chip-technology designer with a signature product blessed by Intel, whose elite Pentium 4 chips currently uses only Rambus technology. No, Rambus' problems have come not from the passing of an economic bubble but from its embrace of two age-old sins: duplicity and greed.

Rambus always had plenty going for it. It was founded in 1990 to solve a critical problem in computer technology. The speed of microprocessors, which crunch data, had improved so quickly that they far exceeded the rate at which memory chips could provide that data. Founders Mike Farmwald and Mark Horowitz, both Stanford Ph.D.s, invented a way to connect the two types of chips that would dramatically improve the speed at which data could be moved from memory to microprocessors. Farmwald and Horowitz had no desire to raise $1 billion to build a chip-fabrication facility, however. Instead, they set up shop as an intellectual-property company. They would sell their technology to chip manufacturers and make money on royalties and consulting fees. They called their technology RDRAM.

By all accounts, RDRAM was a giant leap forward. But selling it proved no easy task. It was so different from past incarnations of memory technology--and so much faster--that chipmakers at first doubted it could work. It was expensive to implement. And the very chipmakers to which Rambus was trying to sell its products were developing a competing product called SDRAM. Where RDRAM was seen as revolutionary, SDRAM was resolutely evolutionary. (The most significant difference between the two, according to the judge's ruling in the Infineon trial: RDRAM transmits three different types of information on a single "bus"--essentially a collection of wires; SDRAM has separate buses, each dedicated to a single type of information.)

In a June 1992 business plan, Rambus CEO Geoffrey Tate described the SDRAM as generally "inferior" and dismissed it as an "incremental improvement" on 20-year-old technology. Still, he noted, "all customers are familiar with it and understand it, so there will be a tendency to try the Sync DRAM [SDRAM] approach."

Tate laid out a four-part strategy for dealing with the SDRAM threat. The first three parts assessed ways in which Rambus could "counter" the technology in the market. Part four laid out a parallel strategy, to claim ownership of it: "We believe that Sync DRAMs infringe on some claims in our filed patents; and that there are additional claims we can file for our patents that cover features of Sync DRAMs. Then we will be in a position to request patent licensing [fees and royalties] from any manufacturer of Sync DRAMs." Rambus would follow that plan to a tee--competing with SDRAM in public while quietly working to lock up the patent rights to key parts of it.

The SDRAM standardization process was occurring at an organization with its own cumbersome acronym: JEDEC (Joint Electron Device Engineering Council), a group in which competing memory-chip makers develop standardized parts so that they can all sell, say, to Compaq or IBM without needing to redesign their products for each customer. Standards bodies are an anomaly of capitalism: In order to work, bitter rivals have to put down their swords and collaborate. The process requires the realization that self-interest sometimes demands cooperation.

JEDEC has always tried to avoid adopting standards that are already patented--or about to be--so that no one company will have the power either to charge outlandish royalties or to block a competitor from manufacturing the product. But there are so many patents out there that avoiding all of them is nearly impossible. So companies that have patents on a standard are asked to commit to license all comers at a reasonable rate. On the rare occasions where a company refuses, JEDEC changes its standard so that it doesn't use the patented design.

Accordingly, JEDEC's sign-in sheets warned members to disclose their patent rights--both issued and pending. Committee chairs announced the policy before each meeting, and pending applications were included on the list of members' patents that circulated with each meeting's minutes. Nonetheless, Rambus would later argue that until 1993 JEDEC's policy applied only to issued patents, not pending patents, and that even afterward, the organization didn't make clear that patent applications had to be disclosed. JEDEC President John Kelly says that Rambus was the only company that didn't get the message: "Our regulars will tell you it was clear."

While Rambus was monitoring the proceedings, routinely updating its patent applications with concepts such as programmable latency, one of its lawyers was growing concerned. The attorney told Crisp, the company's JEDEC representative, that even an intentionally misleading silence about patents at a standards body could lead a court to void the patents.

But on at least one occasion, silence was exactly what Crisp gave. In May 1992, when the JEDEC committee chair asked him whether Rambus had any patents relating to "two-bank designs," another component of SDRAM, Crisp simply shook his head no. (At trial, Crisp defended his muteness by noting that Rambus' patents hadn't yet issued. When Infineon's lawyer asked, "You had applications on two-bank designs, didn't you, sir?" Crisp responded, "That's right. But that's not what he asked me.")

The JEDEC chair had asked the question at the request of Infineon's corporate progenitor, Siemens, which had expressed concern about some Rambus patents. Indeed, Siemens and Rambus had been circling each other for some time. Rambus had attempted to get Siemens to license RDRAM, and though Siemens had declined--there was as yet no customer demand--its executives had been impressed. Later, during the trial, several Siemens documents were introduced that implied that the company was hoping to devise a version of RDRAM technology that didn't require paying any money to Rambus. In the meantime, however, Siemens, like other chipmakers, was focusing its efforts on SDRAM.

Meanwhile, Rambus' patents were starting to roll in: The Patent Office issued Rambus its first, on RDRAM, in 1993. (Rambus promptly disclosed the patent at JEDEC.) In 1996 Rambus' first SDRAM patent was issued. That year, at its lawyer's recommendation, Rambus bailed out of JEDEC; CEO Tate conceded at trial that the patent issue was at least one of the reasons for the company's departure. And the withdrawal letter that Rambus sent JEDEC included a list of all of Rambus' patents except its SDRAM patent. Rambus lawyers later said the omission was unintentional.

Later in 1996 came the news that would vault Rambus to prominence. Intel announced that its next generation of PC microprocessors would use Rambus technology. Intel went so far as to invest $850 million in memory-chip makers to encourage them to implement the Rambus design. With this critical backing, virtually every memory company, including Infineon, anted up for a license on RDRAM. The news helped fuel Rambus' smash IPO in May 1997.

But what nobody counted on was a series of technical snafus at Intel. Rambus' technology may have been fast, but it was cumbersome to implement. And, says analyst Dean McCarron of Mercury Research, "Intel didn't perceive the market and political realities."

The market reality was that Rambus technology was expensive. The political reality was that the company had alienated many chipmakers. "From the beginning, Rambus has been arrogant," says Desi Rhoden, chairman of JEDEC and CEO of Advanced Memory International. "[Rambus] used to say, 'We're going to bury the memory industry.'" The company was also known for its heavy-handed tactics. For example, according to Rhoden and three other sources, Rambus' contracts forbade customers from saying anything negative about Rambus in public. According to those sources, Rambus threatened unspecified retribution on several occasions when people made statements that Rambus took exception to. "Lots of people got slapped around," says Rhoden. (Rambus vice president Avo Kanadjian replied in an e-mail that such "comments appear to be misinterpretations.")

Increasingly frustrated by these Rambus-related headaches, Intel began backing away from its erstwhile protege. Just as the company was getting set to introduce its Rambus-only Pentium 4 last year, it announced that it would soon make versions of the chip that would be SDRAM-compatible. "We made a big bet on Rambus, and it did not work out," Intel CEO Craig Barrett told the Financial Times last fall. "It was a mistake to be dependent on a third party for a technology that gates your performance." To be sure, RDRAM production is expected to rise on sales of the Pentium 4, but it has a long way to go to catch SDRAM: In the first quarter of this year, RDRAM production accounted for only 5.6% of the market, according to Gartner Dataquest, compared with 83.6% for SDRAM.

By late 1999 the bulk of Rambus' SDRAM patents, which it had been industriously filing and amending for years, finally began to be issued by the Patent Office. Soon thereafter, the company started approaching its RDRAM customers to let them know that it expected to be paid for SDRAM too.

"People were incensed," says analyst Bert McComas of InQuest Market Research. It was as if a microbrew that had built its entire brand identity on the notion that it was better than Budweiser had then turned around and demanded a royalty on every case of Bud. But Rambus' sweeping licensing program went even further. Late last year a Rambus executive told industry weekly Electronic News that it wanted to license not only the entire SDRAM market, but also every product that connected to SDRAM. If successful, says McComas, Rambus "would end up taxing everybody" in the industry.

On Wall Street, the company's new patent strategy drew cheers. "Rambus has the long-term potential to become the most powerful intellectual-property company on the planet," raved Morgan Stanley analyst Mark Edelstone to the Wall Street Journal last March. (Edelstone did not return FORTUNE's calls seeking comment.)

Despite resentment toward Rambus, chipmakers representing nearly half the market capitulated and agreed to pay license fees on SDRAM. Hitachi caved after being sued by Rambus, for expedience: It didn't want litigation to impede the planned spinoff of its chip unit. Samsung executives, meanwhile, vowed to resist, according to McComas--and then also signed a license. (Hitachi and Samsung declined to comment.)

In June 2000, soon after the Hitachi settlement, Rambus' shares topped out at $127, giving the company a stunning $13 billion market value. Buoyed by the euphoric stock price and the momentum of its licensing deals, Rambus began pounding on the doors of the remaining industry holdouts. Only six weeks after first contacting Infineon in late June, Rambus sued for patent infringement. Two other companies, Micron and Hyundai, realizing that they were next on the list, sued Rambus instead, seeking rulings that they weren't infringing the SDRAM patents.

As quickly as the euphoria had materialized, it evaporated. Battered by fears that Rambus was losing the support of Intel and that Rambus litigation could be protracted and costly, Rambus' stock began a steep decline. Things got only worse as the Infineon trial approached. In March, for example, the judge issued a pretrial ruling that defined the Rambus patents in a way that made it virtually impossible for Rambus to prove infringement. On the day of that ruling, the stock dropped by a third. (Lately the stock has been trading around $10.)

At the Infineon trial, the heart of Rambus' defense was that it had never stolen ideas. Au contraire: Rambus argued that it had turned to patents in self-defense because other companies were trying to steal its ideas--which Rambus had shared under nondisclosure agreements--and convert them into a "public domain" version of Rambus' design. Never mind that Rambus, one of the most aggressive companies around, never mentioned those alleged thefts at JEDEC, let alone accused any company there of violating its nondisclosure pacts.

During the trial Rambus' lawyers were outmaneuvered at every turn by a team from the New York office of Kirkland & Ellis. Both Tate and Crisp of Rambus admitted trying to fold JEDEC technology into Rambus' patents.

"What you did," Infineon lawyer John Desmarais asked Crisp on the stand, "was work on new claims for the Rambus pending patent applications, and your intent was to make them broad enough that they would cover an SDRAM using the features that you had seen at prior meetings. Isn't that a fact?"

Crisp answered, "In some cases that was true."

In light of admissions like that, it's not surprising that the jury found that Rambus had committed fraud and slapped the company with $3.5 million in damages (which were automatically reduced to $350,000 under a punitive-damages cap).

The judge hearing the Infineon case, Robert Payne, concluded that Rambus' patents hadn't even been infringed. If other courts adopt his interpretation, Rambus will have virtually no way to prove that other companies infringed either. That could cost Rambus a significant chunk of its revenues, seriously imperiling its growth prospects.

Court cases are never over till they're over, of course. But things are looking mighty gloomy for Rambus. To reverse its defeat in the Infineon trial, Rambus will probably have to get both the jury's fraud verdict and the judge's patent and infringement rulings overturned. It's not impossible, but it's a very tall order, and one that's likely to consume lots of time and money--with no guarantee of success. Rambus currently has plenty of cash on hand, but its legal fees devoured a scary 23% of its revenues last quarter. Shareholders will tolerate that kind of bleeding for only so long. (Asked about Rambus' legal fees, Kanadjian answers only, "Rambus is committed to protecting its [intellectual property].")

As FORTUNE went to press, the judge in the Infineon case was planning to hear post-trial motions on July 6 that could bar Rambus from pursuing a second case against Infineon in Germany, due to go to trial in late July. And testimony suggested that at least one of Rambus' SDRAM licensees may be released from its contract should a judge rule essentially that Rambus' sleazy behavior deprives Rambus of its rights to wield its patents. What's more, Rambus is being investigated by the Federal Trade Commission, which is looking into Rambus' actions at JEDEC, according to two lawyers involved in the Rambus litigation. That is no idle threat. In 1996 the FTC charged Dell with hiding patents from a standards body and ordered the company to give up the patents in question. The FTC is also investigating Sun, which settled a suit earlier this year involving similar allegations.

Was Rambus' goal to monopolize the entire DRAM industry? Or was the company just seeking to raise the costs of using SDRAM as a way to force people to use Rambus technology? Industry observers differ. But they do agree on one thing: The Rambus campaign was an act of unparalleled hubris. The company was profitable, but that wasn't enough. Says one chip executive: "They got greedy."

FEEDBACK: nvarchaver@fortunemail.com