Whenever I think of Transmeta, I picture a game-show contestant who must perform a complex task before the sand sifts into the lower half of an hourglass. In Transmeta's case, the task is the just-announced TM8000 processor, or Efficeon. The sand represents money cash on hand to keep the lights on.
The company has just announced it is ready to ship Efficeon. But make no mistake about it, the sand is still sifting, to the tune of $15 million to $20 million a quarter. With less than $75 million in the bank at the end of September, that gives Transmeta about a year to turn the hourglass back on its head.
So, will they make it? It will be close.
Here's the good news: Efficeon is much improved from the previous generation. Although Transmeta still has trouble competing with Intel's mobile processors on a clock-for-clock basis, it seems now that the company can actually deliver acceptable performance for its primary market: suppliers of innovative, low-power notebooks.
Efficeon is also paired with a name-brand chip set, Nvidia's nForce3 Go120. (Credit that coup to the processor's HyperTransport interface, which undoubtedly made it easy for Nvidia to spin a chip set for AMD's mobile Athlon 64 into the Go120.)
Transmeta also has a loyal set of customers, including Sharp, Sony and Toshiba. But while they haven't abandoned ship, neither have they bet a product line on Transmeta silicon. That won't change with this version of Efficeon.
Transmeta says to watch for business to pick up next spring, with a spate of new notebook designs. I expect them to be experimental products on the periphery, where Transmeta has been positioned in the past. If those products do well and if Transmeta transitions Efficeon to Fujitsu's 90-nanometer process from TSMC's 130-nm fabs on schedule then and only then will OEMs consider a heftier bet.
Transmeta found itself in the same situation two years ago, the last time it scheduled simultaneous changes in foundry partner (IBM to TSMC) and manufacturing process (180 nm to 130 nm). In that case, Transmeta missed its schedule and lost most of its initial momentum.
Presumably, Transmeta has learned from past mistakes. It better have. The company has done a good job cutting spending. But even at the current burn rate, there is precious little margin for error in its race to spin silicon into money before the sand runs out.
Mike Feibus is principal analyst at TechKnowledge Strategies Inc., a market research firm in Scottsdale, Ariz., that focuses on components for mobile systems. Reach him at mike@techknowledge-group.com.