Loudcloud

Loudcloud was founded in September 1999 by Silicon Valley legends Marc Andreessen, Ben Horowitz, Tim Howes and Sik Rhee. It sold what would now be called managed cloud infrastructure roughly a decade before the phrase existed: servers, storage, networking and the software on top, run for the customer as a monthly service rather than assembled by them. AWS before AWS.

At it’s core there was a software called Opsware, to provision and configure servers. The business ran into the end of the dot-com boom. It raised $68 million before launch and another $120 million in mid-2000, went public in March 2001 at $6 a share after a far larger valuation was abandoned, and in June 2002 sold the managed services and hosting business to EDS for $63.5 million in cash plus licensing fees. What was left renamed itself Opsware and sold the automation software to everybody else, including to EDS, who now ran the hosting. HP bought Opsware in 2007.

It is recorded here because it is the ancestor of the category. Loudcloud’s argument — that the interesting product is not the machine but the automation that operates it, and that a customer should rent the outcome — is the argument every managed platform in this register still makes. The company that made it first could not survive long enough to be it, and the software outlived the service.

The same story, eleven years later

dotCloud built container tooling to run its customers’ applications, released it as Docker in 2013, renamed itself after the tool and sold the hosting business off; the hosting died in somebody else’s bankruptcy three years later and the tool became an industry standard.

Loudcloud built Opsware to run its customers’ servers, sold the hosting business to EDS in 2002, renamed itself after the tool, and was bought by HP for $1.6 billion five years after that.

Two companies, eleven years apart, and the same discovery both times: the software a hosting company writes to operate other people’s applications turns out to be worth more than the hosting. It is worth knowing which of the two a platform thinks it is in. A host whose automation is its real product has a reason to keep the hosting alive and a reason not to, and this register has more than one record where that question is still open.