01/19/2011 02:33 pm ET | Updated May 25, 2011

China, U.S. and the Politics of Failure

Who killed Evergreen Solar?

This month, the Massachusetts-based company announced it would shutter its local factory and lay off 800 of 925 employees. The firm lost $54 million in the first nine months of 2010. A European joint venture limped offstage last year.

The company isn't quite dead -- it plans to reposition itself as a wafer maker -- but the future looks cloudy.

Conventional wisdom -- predictably -- blames Chinese solar panels. But it's not a fully adequate explanation. First Solar, the largest solar module maker in the world, has led the industry in cutting costs. It comes from Arizona and runs factories in Europe and Malaysia.

SunPower makes some the world's most expensive silicon solar modules. Yet it has managed to defy predictions about its impending death for years. Suntech, a Chinese company, has a new factory in Arizona.

How come they -- and others -- are surviving?

Blame it on the rogue ripple, or a series of small, seemingly sound management decisions that have a massive and unforeseeable impact on the destiny of a company.

Back in 1994, serial entrepreneur Jerry Kaplan formed an online auction site that quickly gained buzz with the business press and web consumers. Legendary VC firm Kleiner, Perkins backed it.

It wasn't eBay. The company was called Onsale -- eBay emerged the following year. Onsale specialized in selling PCs and other remainders from manufacturer and distributor warehouses. Ebay trafficked in used goods from consumers. Ebay exploded. Onsale ended up on the scrap heap of history.

"If you don't get the model exactly right, capitalism can be unforgiving," Kaplan, laughing, told me once.

The same happened to Direct Hit. Who? What? Back in 1998, Direct Hit, backed by Draper Fisher Jurvetson, was the next big thing in search.

What happened? Direct Hit based its back-end on servers from Sun. The company soon maxed out its ability to index the web. By contrast, Google built itself around cheap, replaceable Intel servers.

The core of Evergreen's technology is called String Ribbon. Traditionally, silicon manufacturers create a round log of pure silicon and saw off thin wafers, sort of like prosciutto. Square solar cells are then cut from the round wafer: silicon sawdust and the moon-shaped wafer edges need to be recycled.

In the String Ribbon process, two strings are placed in molten silicon. Gradually, as they are pulled apart, a thin sheet of silicon forms. No sawing, no lost corners, no waste. Brilliant idea.

Unfortunately, in the early version of String Ribbon, the silicon could only be stretched so far, leading to rectangular wafers. Unusually sized wafers led to unusually sized solar cells, which in turn required tweaks to solar modules.

Third-party module makers -- faced with the prospect of having to retool factories to adapt to Evergreen -- passed. Instead of selling wafers, the company's module group became Evergreen's captive customer.

When solar demand outpaced supply, the company could garner a profit. Then the recession hit. Silicon dropped from a few hundred dollars a kilogram to $45. Module prices, goosed by an influx of suppliers, plummeted.

Suddenly, weird and isolated weren't virtues. Evergreen managed to drop its costs to a very competitive $1.88 a watt toward the end of 2010, but it couldn't reverse the momentum. (Interestingly, Luminus Devices, an MIT alum with an odd-sized LED, is also undergoing challenges.)

Evergreen has now figured out how to make normal sized wafers. It could work, but the company carries debt and a whiff of failure. What kind of volumes can they achieve? What is the efficiency roadmap? What are the warranties and reliability test results?

China's impact can't be ignored. At times, the company has bought glass from China instead of nearby New Hampshire to cut costs.

Nonetheless, human decisions played a part in the drama. Don't forget: others have dodged the bullet. First Solar bet on technology -- cadmium telluride -- others found too complex. SunPower was an early mover in developing solar power plants.

AQT Solar guessed correctly that it could make complex CIGS solar panels with standard factory equipment. Others spent hundreds of millions on custom equipment.

And look at Innovalight. The same year Evergreen opened its module factory, Innovalight decided not to build one -- instead, it retrofitted its business to just sell its solar ink to others. The shift lowered revenue but also the need for capital. It now has five customers, all Chinese.

Luck or better foresight?

Maybe you need a bit of both.