Although the project seemed threatened in early 2006 from all sides these were minor compared to the problems to come. The biggest concern at the time was lack of an LCD panel manufacturer, but Negroponte and CTO Mary Lou Jepsen managed to charm another eccentric Taiwanese billionaire. Wen-Long Hsu—founder of southern Taiwan’s Chi-Mei conglomerate—is the owner of the world's largest collection of Stradivarius violins, and he played one for them when they visited to sign contracts.

By the fall, everything was working great in prototype form. Quanta agreed to run its first batch, and even agreed to run a suspend-resume hibernation test cycle 1000 times on each test machine. Normally, test units were give this cycle four times, so it was a particularly unusual request. Then, at 3am on the first day of mass production, Jepsen got a call. Everything was shut down; the laptops were going to sleep and not waking up.

"All hell was breaking loose." She hauled ass to the manufacturing lab with a few other guys and started pumping the caffeine.

Eventually a Quanta guy named Gary Chang and an OLPC guy named Richard Smith ("He's from Arkansas, looks like surfer dude") solved the problem. "We were calling it the second shot from the grassy knoll," says Jepsen. Apparently, as the system was shutting down, electromagnetic noise was corrupting data, screwing up the instructions that told the thing how to wake up again.

At around the same time, the maker of the wireless chips, Marvell, decided to update the firmware for the radio, and they started to crash. "We had four people in four time zones working on that problem," said networking engineer Michail Bletsas. "Mark Foster in Taipei, me in Boston, someone in India, and someone in Santa Clara. We had to program a workaround on the fly: It's in the radio, something you're not supposed to touch under normal consequences."

"A lot of those stories weren't told," says Jepsen. "We weren't hiding it, everybody knew, but we weren't broadcasting it. We figured it all out, and shipped a million of them."

Threat Level Rising
By late 2006, Intel had finalized its specs for the Classmate PC. Though it would cost $30 to $40 more than the XO—the "$100 laptop" in the end cost $188—the Classmate had a faster processor, Intel brand equity and the option of Windows XP as the OS. (Bulk buyers could also opt for Linux.) It was seductive in that it wasn't the revolutionary product that the XO was, but something more familiar, and in line with what ministers of education might have been considering already. What's more, it was a reference design that regional companies could license and customize to fit their needs. And, perhaps, countries rife with pirated software infrastructure had plenty of free programs to run from the black market.

As it began pilot program, Intel's strategy was seen as more traditional too: Laptops could go to teachers, or loaned to students. It did not enforce Negroponte's logical but strict mandate, that the laptops be given to the children, and that they should only be deployed when there are enough to go around.

In the middle of 2007, Intel and OLPC entered into a partnership that was probably more of a hindrance to each other's initiatives than any sort of help. From the start, the deal was vague, more of a mutual appreciation society than a true strategic alliance. Six months later, it had dissolved in acrimony. OLPC accused Intel of pitching Classmate to would-be XO customers; Intel griped that OLPC wouldn't stop asking that the Classmate be discontinued in favor of the XO.

Meanwhile, Intel's more profit-minded operatives were hanging out in Taiwan, spinning the baby laptop idea to one of Quanta's arch competitors, a little known company called Asus.
On June 8, 2007, while both the XO and the Classmate were still deep in pilot testing, Asus introduced the Eee PC, a $400 mini-notebook running a warm-n-fuzzy flavor of Linux. Not only did it resemble the Classmate more than a little, it was unveiled at a press conference hosted by none other than Intel. It would be ready for sale worldwide by that winter, and when it did become available, boy did it sell like hotcakes.

Sales Figures, Sales Facts
"Selling like hotcakes" is an expression that doesn't mean anything in particular. In many cases, "selling a million" doesn't really mean anything specific either. I've heard OLPC people say they've hit the million mark, but in terms of actual shipments, it's not true.

Due to issues that have nothing to do with hardware—and largely to do with Negroponte's greater mission of educating the world's poor—the XO spent most of 2007 in beta testing. In early November, OLPC launched the "Give 1 Get 1" $400 charitable promotion for US buyers, but the first real bonafide XO deployment happened in Uruguay in on December 1. Confirmed orders might have topped a million at this point, but the number of existing XOs, both sold in the US and deployed en masse to schoolchildren in Peru and Uruguay, hovers around 500,000.

Ask Intel how many Classmate PCs are out in the wild, and you get a vague stat, somewhere in the "hundreds of thousands." Intel, too, promises large numbers to come. Portugal will be buying 500,000 of them for the coming school year, for instance.

The Eee PC, though, is already nearing 2 million sold, having hit 1.7 million in the first half of 2008. It is on target to reach a promised goal of 5 million by the end of the year. (By contrast, OLPC will most assuredly not reach 1 million by the end of 2008.)

The success of the mini notebooks has largely been due to price (even expensive ones rarely touch $600) and their intentionally internet-friendly design (you're not going to load up Photoshop CS3, but browsing and email checking work fine). They are also boosted by the negativity surrounding Windows Vista: By running Linux or Windows XP, they present a desirable alternative to the bulkier, more expensive, resource-heavy machines required to run Microsoft's latest OS.

In the wake of the Eee's success, over 40 mini notebooks have hit the market over night. The top four best-selling notebooks on Amazon fall into this catetgory.

At this point, even if the millions of third-world students eventually get laptops, it's unlikely that the XO will be the one they receive. Still, the past two years are definitive proof that Negroponte can take credit for the birth of an entirely new kind of PC.

And Negroponte does claim credit for the Eee PC's success. In fact, he says it's why he introduced the next version of the XO laptop—a radical two-touchscreen device aimed at a $75 pricetag—so early.

Encore?
I asked him why, with the first XO so clearly in its early stages of shipment, would he show off the XO-2. Sure, he doesn't have customers at Best Buy who may hold off because they know what's coming, but it seemed to take away from the momentum of the original device, not to mention confirming some of its criticisms (underpowered, cramped keyboard, etc.).

"When we announce something now that will be in play two years from now, it's partly to give the manufacturers something to start copying now," he says, elaborating, "If you go back two years and you look at the press, [the XO] was dismissed, it was not possible. Then came the Classmate, then Asus. If I underestimated anything, it was how fast people would [copy] it, even if they didn't get down to the same price or didn't have the same features. It was a movement—a hardware trend—that happened because of OLPC."

He also hopes that the announcement of the XO-2 concept, one that only exists in pictures, will stimulate small developers who work on components. Jepsen's new company Pixel Qi will focus on the next-generation of LCD touchscreen, one that can be made as cheaply as current screens today, but have capacitive touch built right into the active matrix, making it thinner than an iPhone screen. Others who saw the XO-2 renderings have already begun pitching solutions to the group.

Not a Manager
If there's one criticism made against Negroponte that's indisputable, is that he changes his tune.

In the beginning, Negroponte repeatedly affirmed that the XO was to run "Linux or some other open source operating system." After a long struggle that could easily be the subject of another series, the XO has recently been made capable of booting both its own Linux OS with Sugar interface, as well as Windows XP. (Critics say that Negroponte never allowed OLPC's Linux OS to mature so that it could stand up to pressure from the Windows advocates.)

Likewise, he was adamant at the beginning that his laptop be the only one shipped to these third-world educational programs where there isn't so much a "market" as there is a case for charity. He says now that if there is a true market—schools and families with the means and desire to buy their own laptops—others can serve it.

Inside OLPC, the leader's mercurial nature and changing priorities proved too much for the talent he had assembled. On the software side, Walter Bender and Ivan Krstic left after open disagreements with Negroponte—mostly pertaining to the adoption of Windows, but also to the overall goals of the program. Jepsen left in January 2008 in what she says was an amicable split, though other hardware experts including laptop maestro Mark Foster had abandoned ship earlier, possibly because they couldn't get along with Jepsen. Most people seem rankled by the credit that Yves Behar took as the "OLPC designer," most notably in a Wired article that would seem laughable to anyone who read Part 1 and Part 2 of this series.

When talking to staff members, there is a sense that no one really got along, and that the religion that Negroponte had instilled in his lieutenants, enough to get them to hang together for two years, has dissipated. The rocky Intel alliance and the move toward Windows were just the final disillusionments. Negroponte spoke the painfully obvious to BusinessWeek last March: "I am not a CEO. Management, administration and details are my weaknesses."

Pulling an Obi-Wan
Still, Negroponte and whoever has stuck by him charge onward. He said, to us and to others, "OLPC is not a laptop company." He himself said that to be taken seriously, you have to build hundreds of thousands of laptops every month; Quanta currently outputs a reliable stream of around 50,000 per month. Now that the mini-notebook movement is in full swing commercially, perhaps the focus should veer from hardware development. Why then stay in the hardware game? Perhaps it's telling that, on the OLPC website's own "Progress" page, nothing is mentioned after December 2007.

Bletsas—who remains hard at work on OLPC today—says that if OLPC does not stay in business, the laptop makers who followed the XO design cues will start doing what they do best: bumping the specs, upping the prices and keeping product too expensive for the foundation to use it in its educational mission. "Unless we keep designing, showing the world it's doable, I don't think they will follow in that path," he says. "If we stop at this stage, they are not going to come down enough for us to use their machines. We have to push them at least one step further."

Want more on OLPC's secret origins? Jump back to the earlier sections:
Part 1 - Genius, Hubris and the Birth of the Netbook
Part 2 - US and Taiwan's Hardware Lovechild

The pot boiled over on the 31st, when Mandriva's CEO François Bancilhon all but accused Steve Ballmer of dirty dealing in an open letter.

Wow! I'm impressed, Steve! What have you done to these guys to make them change their mind like this? It's quite clear to me, and it will be to everyone. How do you call what you just did Steve? There is various names for it, I'm sure you know them.

The One Laptop Per Child program and Intel said Friday that the chip maker would join the board of the nonprofit and contribute funding...One Laptop Per Child's computers will continue to use processors from Intel rival Advanced Micro Devices Inc., but Intel is expected to work with the project on future technical developments.

According to Ars Technica, Asus will first build its own edition in July; Asus will begin the official Intel Classmate PC production in September. There's no mention of the quantity of installed flash-memory, but Ars does say that the Classmate PC may give the OLPC XO a run for its money:

Although the lower cost and specific design features make the XO laptop a good fit for poor nations, the Classmate PC reference design will almost certainly outperform the XO in many scenarios.

Asustek to launch $249 Classmate PC design in July [Ars Technica]