Intel maintains its dominance of the microchip market by suing companies that compete with it. Tim Jackson tells the story of one of these battles
In the mid-1980s, George Hwang decided to start a new business combining his own expertise in process technology and manufacturing with the talents of Bob Woo, a friend with an expert command of floating-point logic design.
Intel’s 386 chip had been launched shortly before and was earning spectacular profits. Hwang saw that the 386 created an opening for someone to start up a new semiconductor business.
In order to keep the size of the 386 chip to a minimum, Intel had followed its usual practice of not including specialist hardware inside the chip to carry out floating-point operations such as long division. This made sense for the majority of customers, who would never need to do heavy-duty number crunching. But for financial analysts, engineers, and people using computer graphics for design, a floating-point unit, or FPU, was essential: it could do arithmetic 10 or 100 times as fast as a standard chip. Ever since the 8086 generation, Intel had sold this unit as a separate “math co-processor”. Installed next to the main processor, the device would perform the arithmetic when needed.
As soon as the 386 began shipping in volume, it was clear that the co-processor that accompanied it – to be known as the 80387, or simply “387” – was a big seller. Hwang did a little market research, and concluded that his new company could make a lot of money if it could develop a competitor. “Intel was the only supplier, and theirs was not a very high-performance part,” he said.
Hwang came to the conclusion that if he and Woo could design a competing co-processor that did the job as well as Intel’s or better, and could price it at $400, then that one product alone over a three-year lifetime could sell one million units, and thus bring in $400m of revenue – more than enough to kick-start a serious semiconductor business.
Hwang’s business, ULSI Systems, started up in the spring of 1987. To make sure that its own design would be fully compatible with the 386, ULSI had to find out precisely how Intel’s co-processor communicated with the main processor chip. By July, however, the company had not got very far. Then things began to look up: in August, Intel published a detailed guide to the workings of its co-processor in IEEE Micro, a respected industry magazine. This answered most of the outstanding questions, and allowed ULSI to make progress designing its own co-processor to do the job five to 10 times faster than Intel’s.
In December, Alfred Chen, one of the ULSI engineers, gave his chief executive a shock. He showed Hwang a copy of a document called a “target specification” for the 387, marked “Intel proprietary”. The “T- spec” was an early outline of how the chip would work. It had been superseded by the version that went into manufacturing, and the key information that it contained had now been published by Intel in the trade magazine anyway. But Hwang, realising that he could not condone the holding of Intel documents on ULSI’s premises, called his lawyer. Alan MacPherson’s response was swift: he took possession of the document, and made sure there was nothing else belonging to Intel on the premises.
Later, in a tense meeting with Intel representatives, Hwang offered to open his laboratories to Intel lawyers in order to prove that everything was above board: “We told them, `Our part is totally different from yours, and we’ll open our database [to prove it]. We’ve done the architecture, we’ve done the algorithm, but we haven’t gone far into the implementation.'”
Intel’s response was cautious. Tom Dunlap, head of Intel’s legal department, thanked Hwang for his openness, but suggested that the best way to resolve the issue would be for ULSI to drop the project completely. “No. The object of this meeting is to assure you that we’re not using your trade secrets,” replied Hwang. And there the matter was left to rest.
Hearing nothing further from Intel, and receiving no response to MacPherson’s follow-up letters, ULSI went on with the design of their 387-killer. As the chip neared completion, Hwang started talking to venture capitalists – and to AMD, a semiconductor manufacturer that had no equivalent of the 387 of its own, and was the company that could most obviously benefit by making and marketing ULSI’s design.
Then disaster struck. On the very day that Hwang received the AMD agreement for signature, seven officers from the Santa Clara Police Department arrived at ULSI’s premises, armed with a search warrant entitling them to scour the premises for evidence that the company had stolen Intel’s trade secrets. Echoing the hostile trade relations between the US and Japan, the police officers angrily ordered all the engineers in the building with Oriental looks to stay in one place, while allowing the Caucasians to walk about.
Five days later, Intel filed a civil suit against ULSI. Dunlap proposed that the company should agree not to transfer its technology to anywhere other than Hewlett- Packard, the chip foundry where Hwang had already made arrangements to fabricate the co-processors. Hwang asked his lawyer how long the injunction was likely to remain in force. “At worst, the case will take six months to go away,” Hwang’s lawyer told him.
Since Hwang thought it would take more than six months getting the chip into production, they decided there would be no harm in agreeing to the injunction.
But Hwang and his lawyers had never fought Intel before. Dunlap fully appreciated that every extra month he could keep ULSI and Hwang in court would be one extra month that Intel had a monopoly over the 387, and one extra month during which venture capitalists would be unwilling to put any money into Hwang’s operation.
The stress of the search and the civil lawsuit led to strains between Hwang and his business partner, which culminated in an acrimonious parting. It was not until January 1991 when two other later entrants had already started selling their own 387 clones, and the ULSI co-processor had been sitting on the shelf for a year and a half, that Hwang was able at last to put a product on the market. Despite pricing his chip at a bargain-basement $220, Hwang still faced suspicion from customers who wanted ULSI to give them legal guarantees of protection against Intel before they would buy.
In July 1991, following the civil suit and a criminal prosecution that were still in progress, Intel opened a third front against Hwang. It sued for infringing a patent covering a method of making floating-point calculations accurately. At first Hwang thought the case was laughable: the method for carrying out the calculations proposed in the patent was one that had been adopted as an industry standard by a 20-strong committee of which Intel was a member.
But Intel was able to refile the patent to remove the controversial issues, and use it to sue ULSI. Instead of choosing Santa Clara as the venue for the case, Intel took it before a judge in Oregon, where the company was a prominent employer. Intel won; ULSI appealed and won; Intel appealed again. It was not until 1994 that the patent infringement claim was finally thrown out.
But by then, it did not matter. The civil case had run on for so long that Hwang had run into a new problem: Hewlett-Packard notified Hwang that it was getting out of the “chip foundry” business altogether. Since Hwang had agreed to the injunction banning the transfer of his technology to anywhere but Hewlett-Packard, he now faced the risk of being unable to build any more chips while the trial continued. This was the straw that broke the camel’s back.
Intel, then, had managed to take ULSI to court nearly a year after it first became aware of the appearance of the T-spec on ULSI premises, and after ULSI had offered Intel access to records to prove that there had been no foul play. No judge ever had a chance to assess the merits of Dunlap’s accusation or Hwang’s defence, yet Intel had kept ULSI busy with the proceedings for nearly five years. Developing his co- processor had cost Hwang $2,5-million; fighting Intel had cost $3,5-million.
Inside Intel: How Andy Grove built the world’s most successful chip company, by Tim Jackson, is published by HarperCollins. Also see http://www.inside-intel.com