Features

Secretive “Intel-killer” has been plagued by delays, slow acceptance 

SAN JOSE– For nearly five years during the high-tech boom, engineers at a small start-up secretly cobbled together a new microprocessor that promised to turn the PC world upside down. 

Rumors mounted as the company, Transmeta Corp., maintained silence but hired talent such as Linux luminary Linus Torvalds while attracting investors including Microsoft co-founder Paul Allen. 

Finally, on Jan. 19, 2000, Transmeta unveiled a new Intel-compatible processor dubbed Crusoe that offered power savings, performance and low cost. 

Media and analysts hailed Transmeta as a potential Intel-killer, a product that could beat the semiconductor giant with small, cheap and energy-efficient chips. 

But in the more than 2 1/2 years since that flashy debut, the Silicon Valley upstart has faced more rough seas than its namesake, Daniel Defoe’s resourceful castaway. 

Despite praise for innovation and some acceptance by computer makers, especially in Japan, Crusoe has not been built into any U.S. manufacturers’ notebooks. And a second Crusoe offering with more power, the TM5800, was delayed for six months by production problems. 

The sour economy also softened demand. Revenues plunged, and Transmeta’s top job changed hands three times. Though its stock jumped as high as $45.61 after the initial public offering, it now trades at about $2 per share, down 95 percent. 

“At the moment, the problem isn’t the product,” said Martin Reynolds, an analyst at Gartner. “It’s figuring out exactly what to do with it, finding the market and making it a success.” 

Transmeta hitched itself to long-sought goals in the high-tech world — the need for performance computers that are both small and don’t drain electricity like water leaking from an old tin bucket. 

The result is a hybrid chip that shifts some of the workload of the microprocessor from the hardware to software, with obvious advantages: with fewer transistors, the Crusoe’s processors are smaller, allowing greater integration of components that previously required two or more chips. They don’t require a fan and can fit into small, nontraditional designs. 

More importantly, they require less power and are far more flexible than hard-wired CPUs like Intel’s Pentium or the Athlon made by Advanced Micro Devices Inc. They could also be reprogrammed to run software designed for non-Intel computers. 

Transmeta executives say their production problems are past, and the company is well-positioned to take advantage of trends toward low-power, high-performance and low-cost mobile computing. 

“We’re a young company,” said Matthew Perry, Transmeta’s chief executive. “It’s taken a little bit for the innovative spirits of (manufacturers) combined with the very innovative technology of Crusoe to get some really interesting devices. You’re just now seeing it.” 

Crusoe will be used in Hewlett-Packard Co.’s upcoming Evo PC tablet, a next-generation lightweight notebook that can be written on using a stylus, and in OQO Inc.’s “ultra personal computer,” which can run Windows XP Professional and still fit into a shirt pocket. 

The chip also is used in lightweight offerings by Sony, Fujitsu, Toshiba and NEC. 

But it’s not yet clear how much demand there is for small gadgets powerful enough that a full-fledged PC operating system can run on them. 

Transmeta shipped just 70,000 chips in the first quarter of this year, compared to Intel’s 35 million overall and 5 million in the mobile market, according to Mercury Research. 

“They do need to increase their unit volume to be a long-term concern,” said Dean McCarron, president of Mercury Research. 

Transmeta, which reports its latest financial results later this month, declined to estimate when it might become profitable. 

It had $216 million in cash at the end of the first quarter of 2002 and reported, for the six months ending March 29, a net loss of $80.6 million on sales of just $5.6 million. 

As it prepares its latest Crusoe, the TM6000, Transmeta faces even stiffer competition from Intel, which is now also pursuing low-power chips, a niche it had largely ignored until Crusoe. 

Some analysts believe Transmeta’s mistake was to take on Intel. 

Within a month of Crusoe’s launch, Intel made low-cost mobile Celeron processors available. More competitive offerings followed, with Intel launching a dozen different chips on a single day last October. 

“It was quite easy for Intel” to reduce the clock speed on its chips and thus decrease their power requirements to come up with products that could compete with the Crusoe, McCarron said. 

And because of their unique design, Crusoe-based PCs only met, or trailed, in performance compared to the competition in devices built to run with Intel or AMD’s chips, said Eric Ross, a semiconductor analyst at Investec Inc. 

“Essentially, they have designed a novel processor that doesn’t run the Intel architecture natively,” he said. “They emulate it in software. So there’s an overhead cost built into it.” 

Additionally, some Crusoe-based computer makers shrunk batteries or added power-hungry extras that negated the power-savings from Transmeta’s processors, which are not among the computer components that consume the most energy. 

Meanwhile, Intel and AMD continue to improve the power, performance and price of their mobile offerings. 

“Intel and AMD have been cascading their most advanced technology down to the lowest-end products,” Ross said. “Transmeta’s a very small player that might get crushed as they get more price competitive with each other, particularly at the low end.”