__We're about to find out if the ultra-chip startup can beat Intel at a new game: mobile, the market of the future. __
It's Thai night in Geek Heaven. "The standard collection of engineers," David Ditzel says, sweeping his hand over the 50-odd employees who file into his company's largest meeting room for dinner. As Transmeta's patriarch and CEO, Ditzel is playing host. "A few ponytails, a little bit of orange hair. A dog. I like this, people getting together."
It's an understatement about as extreme as the Pope saying he likes to see people in church. For the better part of the last five-plus years, the satay-munching Transmetans before me, and 190 more of their colleagues, have been nothing if not together - locked away in an office building in Santa Clara, California, across town from Intel, unable to speak to anyone about their mission. Their isolation has been so complete that a Stanford professor who saw several of his best students disappear into the company calls Transmeta "the Cult" - a reference supported by a certain glow in their eyes, and by their collective focus, their intolerance of criticism, and their disdain for Intel, the establishment player to beat; for journalists; for the hustling mass of dot-coms. For everything, in short, that is not Transmeta.
The mood is clearly set by Ditzel himself. His minions look to him for a nod before answering even my simplest questions. The one topic that produces unanimous ebullience: what an incredible place Transmeta is to work. One after another, they rave about their employer. "We're all from places like Stanford and MIT, where we're used to being the top 1 percent," says one. "Now we get to be the top 1 percent of the top 1 percent!"
"This is not the next Web-based rip-off scheme," blurts another. "Nowhere else will you find people of this quality."
It's easy to see why they like it. Two-thirds of Transmetans are engineers. There's endless lab space, workdays that run into the night - sometimes the next morning - and free Altoids. It feels kind of like college, except nearly everyone has an office. And together they've already spent hundreds of millions of dollars in venture capital - supplied by Microsoft cofounder Paul Allen and über-financier George Soros, among others - without giving anyone on the outside a clue as to what they were doing. All that and Dave still buys them dinner four nights a week.
This group's confidence is inspired by a sense of having identified a chink in the armor of one of the world's most powerful (and profitable) tech companies. After five years inside a black hole - during which world-class programmers like Linux inventor Linus Torvalds and Quake coauthor Dave Taylor went in, and not a single press release came out - these engineers have emerged with a weapon they claim will penetrate the fortress that is Intel.
As the world learned not long ago, this weapon is called Crusoe - a two-headed line of tiny, software-based microprocessors being produced by Transmeta's fabrication partner, IBM. The first, called TM5400, is a 700-MHz chip for the ultrathin, ultralight category of Windows notebooks, optimized to run software written for Intel chips on a fraction of the juice it takes to power a Pentium. The second, TM3120, is a 400-MHz chip designed to run Internet appliances on MobileLinux, a version of Linux crafted by Torvalds.
No need for a history lesson here. These engineers know that many have tried to out-Intel Intel with marginally faster or cheaper chips, only to get bought out (Cyrix) or become a distant also-ran (AMD), thanks to Intel's stranglehold on the supply chain and its mastery of the Pentium architecture. But Transmeta has no plans to follow such a perpetually vulnerable and temporary strategy. Instead, Crusoe is aimed at a market that Intel has all but ignored, the market of the future: mobile.
Wait, you say. Doesn't Intel have the lion's share of today's mobile marketplace? It does. But Transmeta feels tomorrow's mobile market will be much broader, and besides, Intel has practically backed into its 90 percent share in notebooks. A Moore's law-based development scheme allows Intel to continually increase the speed of its Pentium chips, but this scheme is optimized for desktop machines. The latest Pentium iterations inevitably require more power and produce more heat, and so they need fans - relegating the latest technology to machines with an unlimited power supply and no space concerns. Only later does the chip technology trickle down to notebooks.
By contrast, Crusoe, with one-quarter the transistors of a Pentium, is designed to save as much space - and power - as possible. The beneficiary, Transmeta hopes, will be the consumer, who will finally see lightweight notebooks that run at high speeds and consume far less power and run much cooler than Pentium-based systems. With the help of Transmeta's so-called LongRun power optimization technology, these notebooks will also run all day on a single battery.
The possibilities of a software-based chip - one that makes upgrading your laptop more like getting a tune-up than overhauling the engine - have OEMs abuzz, to say the least. "If Transmeta delivers, this chip can change the whole cost of ownership," says Michael Gale, chief Web officer for Micron PC - which, as of early May, was still working exclusively with Intel. "What they offer philosophically is very, very interesting. They extend the relevance of the machine. People are going to begin to ask if they really want to go the disposable route for the sake of a little extra power now - or do they want to be upgradable?"
In an effort to extend today's definition of mobile, Transmeta is also challenging both halves of the Wintel duopoly - with the chip and the MobileLinux OS in tow - in what could be an emerging class of Internet appliance: the webpad. Transmeta promises that under average use, the product, resembling an Etch A Sketch with a Web browser, will go weeks without a recharge.
"This is a new space and no one knows where it's going - but we do know that having a large market share elsewhere isn't all that advantageous here. This is a new beginning," Torvalds says. "Being a hungry company allows you to do the things we're doing. It allows you to respond to market pressure. It's not that Intel and Microsoft can't make the change. It's that they're caught up in what they're doing now."
For all its promise, when Transmeta officially launched at a villa above Silicon Valley in January, Ditzel announced no customers. In the months that followed, only S3, an Internet appliance company, and Quanta, a notebook manufacturer, came forward with commitments, offering only loose time frames and few details. And as of mid-May, no major OEM would admit on the record to having done anything more than talk with Transmeta.
There's also the question of how well Crusoe performs. Because of the extra translation step a Crusoe takes to be x86-compliant, a 700-MHz chip won't appear to be running as fast as a 700-MHz Pentium. Think of a race between two cars on a straight road. Each car goes 100 mph, but one weaves ever so slightly. The car that moves in a straight line will win the race, of course, because it is covering less ground. But it's still not clear exactly how much extra ground Crusoe has to cover. At the launch, Transmeta officials said that a 700-MHz Crusoe performs about as fast as a 500-MHz Pentium III, but they've yet to make the chips widely available for testing, and have even lobbied publicly for a new benchmark standard - which critics contend means Transmeta is afraid of a side-by-side analysis.
Combine the benchmark issue with the absence of customers, and industry analysts are wondering whether Ditzel has really drawn up a blueprint for the next millennium, or just written another chapter in the history of Intel. "Everyone is asking Dave where his customers are. So far he hasn't given an answer," microprocessor industry analyst Linley Gwennap said recently. "And they haven't demonstrated performance. If it were any good, they'd be sending it all over the place. Most of the time, when companies don't answer these kinds of questions, it's because they have something to hide."
"Transmeta has put new twists on old ideas," added Keith Diefendorff, editor in chief of Microprocessor Report, the sacred text of the industry. "I don't think what they've done is radically new."
Ditzel - just back from Phoenix, where he spent three days at Paul Allen's annual CEO confab - sees little need for concern in such commentary. In fact, he actually agrees with Diefendorff's comment - Transmeta didn't invent Crusoe's underlying technology - but feels that as criticism goes, it's like saying "We've heard of software before."
As for Gwennap, Ditzel says the analysts don't yet understand the scope of Transmeta's vision. As a result, they've focused on the types of issues that, while seemingly important now, will be moot once Crusoe comes to market.
If it's true that analysts don't yet grasp Transmeta's plans, it's Ditzel's own doing. He stonewalled the analyst community until the January launch, and has allowed only limited access since then. It was an unusual move in the microprocessor industry, where analysts are usually briefed in advance so they can assess the technology and communicate its potential to both the media and the markeplace. But Ditzel insists that the strategy was necessary to avoid tipping Transmeta's hand to Intel. "Our first goal was to get a leg up on the competition," he says. "It takes so long to build a chip. If we could get a five-year lead, we could really get a chance to establish ourselves."
In the months since launch, Ditzel has worked to raise his profile, reciting the Crusoe mantra to prospective customers and in industry keynotes from London to Tokyo to Paris, where he's headed tomorrow, after one more helping of Thai food and a good night's rest. "I want to tell you what my life is like," he says, digging into a plate of sticky rice. "Before, I wasn't doing social events. I didn't want to show up and have nothing to say when someone asked me what I did. Now I can."
But he won't budge on the details. He says customers will come forward when they're ready. And then the skeptics will go away.
Four weeks pass. I'm still grappling with Transmeta's lack of customers. Ditzel's a brilliant engineer, but I wonder if he's incapable of conveying the practical advantages of his technology to prospective customers. And then I see a news report of AOL CEO Steve Case at Internet World, showing the same webpad that Ditzel demonstrated at the Transmeta launch. Case is outlining AOL's co-marketing agreement with Gateway to push these Linux-based products, and he's mouthing Ditzel's vision.
After Case's presentation, Transmeta executives fess up, detailing the AOL/Gateway/Transmeta relationship, which they had planned to announce by early June. The trio will offer both wireless and tethered webpads, as well as a "residential gateway" - a home server "the size of a lava lamp" that will cache Web pages through a landline and serve them wirelessly at 11 Mbps. The products will be available by year's end.
While they're at it, the execs tell me that Transmeta has secured a tier-one notebook vendor that's set to unveil a Crusoe-powered machine at PC Expo. The 700-MHz notebook will weigh less than 3 pounds and run all day on one battery; it will be rolling off production lines late this year.
Ditzel has been trying to design the ultimate chip throughout his career. Though he spent years at Sun and AT&T's Bell Labs, before Transmeta he was perhaps best known for a paper he coauthored in 1980 while still a graduate student at Berkeley, titled "The Case for Reduced Instruction Set Computing." Arguing that microprocessors could be made more efficient by moving much of their complexity to software (the RISC approach), it set off a holy war inside the microprocessor industry; Intel's complex instruction-set computing (CISC) approach has prevailed.
But Ditzel didn't give up: He changed his focus to a design that would be x86-compatible because it was primarily software-based. Transmeta was founded in April 1995, with a mission to develop such a design on the back of a number of established programming techniques. Notable among them: "dynamic binary translation."
While the technique's origins are obscure, the architecture is clear. Essentially, it's a layer of software that envelops the chip's hardware component and interfaces with the OS, allowing designers to make radical upgrades. The software translates x86 instructions into native Crusoe instructions, allowing the chip to run Wintel programs, and stores the instructions in memory so they have to be translated only once. That way, the chip performs better as it "learns." Because it's software-based, the chip itself is far smaller and lighter, and it runs cooler.
In 1994, while at Sun, Ditzel and Transmeta cofounder Colin Hunter experimented with binary translation techniques in an attempt to make Windows run better on Sun workstations. They were convinced of the idea's merit, but Sun wasn't interested in changing its processor design. So Ditzel and Hunter decided to leave. "A lot of what we ended up with can be traced back 30 years or so - various ideas and things I've been personally involved in," says Ditzel. "What really made the difference here was learning how to put different pieces together in new ways, to make the whole thing real."
It was with the promise of producing cool new technology that Ditzel and Hunter recruited fellow Sun employees Bob Cmelik and Doug Laird. When they weren't developing the technology, the team would look for funding. But in the beginning, it wasn't forthcoming - because Ditzel had yet to identify a market. "Our only idea was to make a cool chip," says an early employee. "We didn't start out to make it low power. No one was saying anything about Linux back then. We were a hammer searching for a nail. Thank God we found one."
To keep the company going, Laird and Ditzel did some outside consulting, earning enough to move Transmeta from Ditzel's living room to a building in Redwood City. By mid-1995, Cmelik hit upon the phrase "code morphing" to explain Transmeta's technology. It helped do the trick, triggering an investment from Walden Venture Capital. Other investments, including funding from Paul Allen's Vulcan Ventures, quickly followed - and the real hiring process began.
In search of engineers, Ditzel worked the students and alumni from Berkeley, Stanford, the University of Washington, and MIT. For veteran talent, he raided SGI, HP, and Sun, but avoided Intel - to keep his mission a secret and to stave off lawsuits. Soon, the company outgrew its offices and moved to Santa Clara. "We were trying to hire anyone who knew anything about dynamic translation," says Ditzel. "If we couldn't get them, we had them sign nondisclosures and made them friends of the family."
Recruiting top engineers in a talent-hungry Valley is no small task. Getting the software and hardware coders - traditionally like oil and water - to work together (and keep quiet) is even more of a challenge. Keeping them loyal is another issue entirely. Insiders credit Laird with holding on to the team members while the Internet was happening all around them. "Too many people come to a job saying 'This is the way it's always been done.' Here they have to say 'This is what we need to do to solve the problem,' even if it's not the most logical approach," Laird says of his management style. "What I like is a get-it-done attitude. You attack the problem. If you need help, scream. If people aren't in attack mode, they don't belong here."
Even as his engineering team has grown, Laird has remained intimately connected with the labs. "Even though he's managing all these people, Doug is still designing," says Marc Fleischmann, who was recruited from Hewlett-Packard. "At HP, once you have three people reporting to you, you never get your hands dirty again."
By 1997, Transmeta had begun to attract attention by filing for patents and spurred even more speculation when it recruited Linus Torvalds. On geek sites and Usenet groups, some argued that the Torvalds hire was Paul Allen's idea, an attempt to contain the Linux movement by locking up its champion; others thought the hire meant that Transmeta was a Linux company. Both sides were wrong: Ditzel says he hired Torvalds because he's a world-class programmer.
In May 1998, the first shipment of silicon arrived from IBM's foundry; a couple of days later, engineers had the processor running. "It was a miracle that it worked at all," says engineer Scott Draves. "But it was unbelievably slow. It took half an hour just to boot up. It was running software a hundred times slower than a normal Pentium."
That first hardware design led to another - and in the meantime, the engineers began to see that if they could solve the performance riddle, Crusoe could power high-performing, browser-capable systems on a battery life rivaling that of handheld devices. This realization helped the company focus on a core advantage: power. This, in turn, made the mobile market an obvious target. "We knew it was going to be low power, but that turned out to be the chip's strongest asset," explains Frank Priscaro, director of brand development. "So we came up with the phrase 'We rethought the microprocessor in order to solve the problems of mobile computing.'"
On the strength of that directive, Fleischmann fielded a team of 10 engineers to work on LongRun, which can monitor a system, determine changes in power requirements, and alter the voltage and clock frequency so the processor is always running at the minimum power level required. It's a feedback loop, written by the engineers in nine months, that makes SpeedStep, a feature from Intel that senses whether a notebook is plugged in or running on battery, seem antiquated. At the January launch, Transmeta demonstrated the technology by running a DVD movie on a Crusoe machine and a Pentium machine, side by side. The Pentium required 7.1 watts of power to play the movie, compared with Crusoe's 1.75.
By 1999, Crusoe had gone through four hardware revisions and, according to Laird, gobbled up several hundred million dollars. Meanwhile outside Transmeta, the Internet, Linux, and wireless technology had begun working together to drive the mobile market, which increased Torvalds' importance on the inside. "Originally I came on as just a code-morphing person," Torvalds says. "About a year ago, when we were ready to come out of the closet, so to speak, we wanted a demonstration that showed us to be more than an Intel clone. We were trying to demonstrate a product that wasn't a PC or a laptop and that would use the Transmeta CPU; and it would have to run Linux, naturally, because we would need to demonstrate something without a hard disk that could access the Internet. That's where the webpad came in."
It was another well-timed development. No one knows how webpads will fare, but it's clear that demand for mobile Internet access is heating up - analysts predict that the number of mobile Internet appliances coming off production lines will outnumber desktops three to one by 2003. The company began to receive interest from potential partners, particularly in Taiwan and Japan. It was time to start thinking about revenues.
For some, it was not a moment too soon. "If you told me up front it was going to take so long, I probably wouldn't have joined," says Godfrey D'Souza, Transmeta's first employee. "But now I can see the end. I can see people paying money for something I helped make. It's not over until that happens."
Across town, at Intel headquarters, Manny Vara says Transmeta is still a long way from home. "Transmeta says it's better and we're arcane," he adds. "A lot of companies have come and gone. But we're still here. You have to ask why.
"A lot of what Transmeta claims to be inventing, we're already doing," Vara continues. "We have notebooks under 3 pounds. We're working with computer manufacturers to design new systems that reduce power consumption. You won't know about Transmeta's products until you can walk into a store and try them with your own hands. Until then, everything is just speculation."
Vara is responsible for articulating Intel's take on Transmeta to the media, but although hundreds of stories have been written about Crusoe - in the days following Transmeta's launch, reporters covered everything from the technology to the food and the valet parking - Vara has been lonely. He says he's received only "a handful of calls" and no one has quoted him.
He can take some comfort from the fact that Jim Chapman is listening.
Chapman, the senior VP of sales and marketing, is one of very few Transmeta employees who wears a suit and tie to work and actually schedules morning meetings before 10. His team is made up of a four-man staff and a handful of sales guys in Tokyo and Taipei. Among the ranks of upper-level execs, Ditzel, Laird, and Chapman are Transmeta's holy trinity: Ditzel is the visionary, Laird is the get-it-built engineer, and Chapman is the marketing maven, the one who has to identify a viable market - which means that the real pressure these days is on him.
While Ditzel pooh-poohs Intel for its inelegant design choices, Chapman is fearful. An 11-year veteran of Intel's marketing department, he also spent four years at Cyrix trying to compete with Intel, and left vowing never to take on the juggernaut again. But Ditzel changed his mind.
Chapman knows full well why Intel hasn't been able to deconstruct Crusoe's technology in the media: It's a function of Transmeta's silence strategy. By keeping the technology under wraps for five years, and maintaining a tight control on test chips even after launch, Transmeta has befuddled Intel. To wit, he walks me through a copy of Intel's competitive response - the argument the company is making to its customers - obtained from a Taiwanese manufacturer. "Stuff like this falls into our hands all the time," he says.
Judging from this document, in the context of my conversation with Vara, I have to agree that Intel doesn't know any more about Crusoe than the analysts I've spoken with. Its most powerful argument is the same one I've heard over and over: Transmeta has no customers. It's a slippery slope. Transmeta needs only one significant customer and Intel's bluster goes away. "A little weak, isn't it?" Chapman says, slapping the document down on a stack of papers.
Still, until Transmeta makes a major customer announcement, I tell Chapman, Intel doesn't need to say any more than it has. Chapman concedes, but says those days are over. Beyond owning up to the webpad relationship with Gateway and AOL, he details a recently closed round of funding - $88 million. And then he drops the bomb about a major customer unveiling Transmeta's product at PC Expo. "This is going to happen," he says. "All of what Intel is saying, that we haven't substantiated our claims? Just wait."
What Chapman means is that either Toshiba, IBM, Dell, Gateway, Compaq, Hewlett-Packard, or Sony will announce a high-end Windows notebook powered by Crusoe. Chapman won't say which. Dell, I figure, is out, given its just-in-time manufacturing process and its relationship with Intel. Any of the others could be the one.
Someone knocks, then pokes his head in. "I need to talk with you about Toshiba - about your meetings next week," he says, and quickly disappears.
Chapman starts talking benchmarks. He shows me a slide outlining Crusoe's 8.9 hours of battery life versus a maximum of 3.5 for a similar Intel chip, and says that, using standard benchmarks, Crusoe actually outperforms the Pentium when running on batteries. "These are actual numbers," he says. "It's been tested and it's ready to go. It's because of these results that this tier-one company is going to launch." Lastly, he shows me a slide that says Transmeta will follow the PC Expo announcement with a July advertising campaign, and then - if all goes well - an IPO as early as August.
When presented with this information, Vara concedes that Intel isn't powering the AOL/Gateway devices, but he brushes aside the funding announcement. "You can slice that banana in different ways," he says. "$88 million, 9 companies - less than $10 million a pop. That's a fairly small investment. I'm not saying it's chump change, but it's not a huge risk, either." (I learned later from a source that the round's lead investor, Sony, contributed $33 million.)
Another call to an Intel executive ups the ante: According to the source, at PC Expo Intel will announce a new low-voltage, 700-MHz Pentium processor of its own. At that time, the source says, every major notebook vendor will introduce notebooks that incorporate the new chip, which is designed to prolong battery life. The source won't say how long these systems will last on one charge, though, because battery life is dependent on the entire system design, and this will vary by vendor. So how can Transmeta make an "all day" claim? "They're bluffing," the source replies. Until Transmeta systems are available for testing, it's a safe answer.
Since my meeting with Chapman, it has become apparent that Transmeta's prospects may be even rosier than he let on. Although none of the notebook OEMs were willing to commit to Transmeta on the record as this story went to press, not-for-attribution accounts suggest that as many as six of the top ten notebook OEMs are planning to display working products powered by Crusoe in Transmeta's booth at PC Expo. What's more, one major vendor appears to have committed to taking shipment of as many as 100,000 Crusoe chips by November.
Given Intel's announcement, customers will be able to compare the products side-by-side (Transmeta has secured a booth adjacent to Intel's) and decide which chip really delivers. Even without the benefit of that comparison, it would seem that Transmeta at least has physics on its side. Intel's strategy of dropping voltage to prolong battery life has to affect performance. "If a high-end chip is designed to run at 800 MHz with a 3-volt supply, and you cut it back to 1.5 volts, you'd better not try to run the thing any faster than 400 MHz," says John Wharton, a Stanford lecturer and microprocessor consultant whose clients include HP and IBM.
Transmeta's LongRun technology works the same way - except that the voltage is dynamic. It's set according to the needs of the application, while Intel's fix is hardwired.
Wharton also claims that because Crusoe is software-based, the re-creation process will be more fluid. "Intel may put 500 or 1,000 man-years into designing [its next-generation chip] Itanium. The next Transmeta chip may require 10 or 20, or maybe 50," he says. "Intel is a smart, flexible, financially successful company, and I expect that to continue for many years. But Transmeta is doing something quite innovative. I give them better odds than I ever gave Cyrix, National, or IBM."
The support of notebook OEMs would give Transmeta some buoyancy, but the Internet appliance space may be the wild card that determines whether Transmeta becomes more than a boutique shop. Although Intel is a tough rival in notebooks - which is where the company will leverage OEM relationships - it has no clear advantage in this emerging category.
Are webpads just one more technology that no one asked for, like videophones and HDTV? It would be easy to dismiss them as such if it weren't for AOL's Steve Case. Case more than anyone else has managed to predict what consumers want out of the Net. And although no one at AOL would talk about Transmeta as Wired went to press, Case appears to be backing Crusoe. "We've been endorsed by the largest Internet appliance vendor, the one with the most substantial strategy in the industry, to be the building-block architecture for this category of products," Chapman gloats.
AOL envisions Web access becoming important throughout the home, whether the user is making popcorn in the kitchen or watching TV in the living room. It sounds silly, until you think about the possibility for Time Warner tie-ins, and the number of people who even now are spurred by their TVs to get up from their couches and vote on the question du jour from their desktop PCs. Or until you see how the webpad will facilitate Middle America's QVC obsession and enable commercial-induced impulse buying. AOL can no doubt price these devices attractively - as loss leaders, perhaps - or even lease them. For Transmeta, signing on AOL is huge: If only 1 percent of AOL's 20 million customers opt for a webpad, that's 200,000 Crusoe chips.
The analysts are sold: IDC sees the Internet appliance device market growing to 89 million units by 2004. Granted, that number includes devices such as WAP phones, for which Transmeta has no working design. But the company feels the two aren't mutually exclusive. As consumers continue to warm to mobile surfing, they'll clamor for a richer, full-screen browser experience.
Which is not to say Ditzel won't change his focus if webpads don't pan out. Although Crusoe, at 400 MHz, is a power hog compared with the dominant cell phone chip, manufactured by ARM, Ditzel suggested at the launch that any battery-operated product with a browser is ripe for a Crusoe chip, whatever form those devices may take.
But then, maybe the man had been silent for so long that he got a bit caught up in the moment. "Dave should be writing screenplays instead of running Transmeta," teases former Sun colleague Brian Case. "His formula for generating suspense is unparalleled."
Meanwhile, Ditzel's employees are trying to stay focused and not think about the IPO. On my last visit, someone had posted a mock ad for a Transmeta candy bar: "Creamy Rich IPO Inside, Surrounded by Nuts!"
If and when that IPO happens, it will most likely make up for the fortunes these engineers passed up by not jumping to dot-coms three years ago. But it will also help develop Crusoe's next iteration in the lab - a chip that, of course, no one's talking about. One thing's for sure: This chip won't take another five years, and it most likely won't look much like the current version. The engineers are already working off a 128-bit-wide architecture, while Intel's Itanium, which has yet to hit the market, is still a 64-bit design. Theoretically, 10 years from now, Transmeta will be free to implement a 3,000-bit-wide architecture, or whatever other technology is state of the art.
All this has some people already claiming that they knew Transmeta when. "Intel, with its strong brand and technology, has a strong hold on the industry - but a company like Transmeta can step up to fill holes in their mobile strategy," says Micron's Michael Gale. "We're one of the last true-blue companies, but we want technology that'll allow us to make the best products out there. The game isn't just about Intel anymore."
"Starting now, I think the most challenging tasks will be carried out using software analogous to Transmeta's code morphing, in lieu of silicon," says Stanford's Wharton. "If the company is talking only about portable network devices, it's just because it's trying to focus. Everything Transmeta has said about power and portability is true, but it wouldn't surprise me if the company's aspirations are much broader than it has let on."
As both critics and supporters have surmised, Ditzel has a flair for the dramatic; he's written an impressive screenplay - as far it has gone. The challenge, of course, is for him to finish that script in the public eye. If he can get that far, we already know the ending - and, for the first time in 20 years of Silicon Valley drama, it won't have Intel coming out on top.
