Friday being the natural day to publish a feature about Transmeta and the power-saving technology in its then-flagship Crusoe processor, it's sad to report that Crusoe appears somewhat shipwrecked. It's now nine years on from the company's incorporation, and four both from the chip's debut and from when I first met vice-chairman and chief technology officer David Ditzel. Once CTO at Sun and ex-director of SPARC labs, he can be presumed to know a thing or two about processor design. Linux progenitor Linus Torwalds was on board too.

Yet Transmeta's balance sheets continues to haemorrhage red ink. Having lost $591 million over that four year period, $25.5 million of it in its second quarter of 2004 - more than the $22 million lost in the equivalent quarter of 2003 - the company's time is running out.

What happened? Back in 2000 when Crusoe was launched, when the economic landscape looked very different from today's. Ditzel was predicting that his company would make serious inroads into Intel's market share as result of its then-revolutionary power-saving processor technology. We would all have these CPUs in our laptops and would be living with battery lives of four to eight hours.

We do now enjoy the battery life that Ditzel predicted, or close to it - but the chips providing it aren't Transmeta's, they're Intel's. As we've previously noted in these pages, Intel's mantra that 'only the paranoid survive, coined by Intel's then-boss Andy Grove, remains securely in force at the Santa Clara semi-conductor giant.

In a move straight out of capitalist for dummies version 1.0, Intel raised the barrier to entry not just by making ever-more complex chips in ludicrously expensive, multi-billion dollar fabrication plants, but by making best use of the ideas presented to it by its competitors.

Transmeta's technology, called LongRun, is nonetheless very impressive. Using code-morphing software to translate x86 native code into Crusoe-compatible instructions, the CPU adjusts its clock speed to the requirements of the application thousands of times per second, the idea being to provide the right amount of CPU power appropriate to the circumstances. Early impressions were that the system worked well if slowly. The first chips ran even slower than had at first been claimed.

However, Crusoe consumed a measly 1W - compare that to the three-figure consumptions of today's top chips - which made them ideal for applications such as blade servers as well as laptops. Unfortunately, many companies who might have bought those new blade servers simply evaporated as the economic technology boom turned into a mudslide. And, despite the early sign-up of IBM to the notion of a Crusoe-powered ThinkPad, Big Blue pulled the plug on its development, despite having a reasonably specified example running at PC Expo in July 2000. When we saw it though, no-one was keen to allow journalists any hands-on experience with it, which raised doubts even then. IBM is not listed on Transmeta's list of OEM customers with whom it has worked.

In many ways compatibility was and remains one of its main problems. The risk for an OEM entails building a product that cannot be guaranteed to be compatible under all circumstances. As AMD found out, entering the CPU market when there's an established player with the dominance of Intel is very tough rock face to climb. Only in the last five years or so has AMD managed to banish the gremlins - whether real or perceived - of non-compatibility.


The other one is of course that Intel has managed to steal Transmeta's clothes with its versions of the Pentium M for mobiles. Neither users nor OEMs care - in the ultimate analysis - about the technology used to provide benefits such as battery life. As long as they get it, that's enough. And if you can have it without having to worry about compatibility - which you presumably don't if the chip has Intel printed in the top - then that's what you're more likely to buy. It's tough competing against Intel, which dominates the semi-conductor industry in way almost unparalleled in any industry.

That's especially true when the behemoth has contracts with companies such as Dell, which buys chips in vast quantities, big enough to allow it to sell at a reasonable price while making comfortable margins.

So, locked out by contracts, incompatible technology and both mind and market share, what next for Transmeta? It would appear to have few avenues of escape while attacking the PC-compatible laptop business.

It is selling a special version of Crusoe into embedded markets and is making noises about quiet PCs - low power means lower cooling requirements and so lower noise, and is a market niche slated to grow. It also claims a number of OEMs selling tablet and notebook computers. One other way forward to license parts of its technology to those building applications demanding low-power consumption and for whom x86 compatibility and/or high performance are not issues.

Those aside, and given the state of its balance sheet, it's hard to see how Crusoe is ever going to escape from its island - if at all.