Nebula, Inc.

~ the goodbye ~

OpenStack came out of NASA. Chris Kemp was chief technology officer at Ames Research Center when the agency's Nebula compute project was merged with work from Rackspace to create it in 2010, and in March 2011 he left to start a company around the result. It was registered as Fourth Paradigm Development, took the name Nebula from the NASA project it grew out of, and was based in Menlo Park before moving to Mountain View. Devin Carlen and Steve O'Hara founded it with him, and a good part of the NASA team that had written the original components followed.

The product was one box. Nebula One was a hardware appliance you put in a rack next to servers you already owned, and it turned them into a working private OpenStack cloud without anyone on your staff having to become an OpenStack engineer. That was a good read of the problem. OpenStack in 2013 was famous for being powerful and miserable to operate, and a company that wanted the thing without the operations burden had nowhere to go.

Money followed the idea rather than the sales. A series A in May 2011, then a $25 million series B in the autumn of 2012 led by Comcast Ventures and Highland Capital, with Eric Schmidt's Innovation Endeavors joining. CIO.com named Nebula one of ten hot cloud companies to watch in March 2013. Nebula One shipped generally on 2 April 2013. Almost exactly two years later the company was gone.

~ the chronicle ~

~ what the box was competing with ~

Everything that could go wrong for a private-cloud appliance in 2013 went wrong at once. Amazon Web Services was eating the demand for anything that looked like infrastructure you had to own, so the customers who might have bought a private cloud were instead deciding they did not want a data center. The ones who did want OpenStack on their own hardware increasingly had vendors like Red Hat and Mirantis offering it as software with support attached, which does not require capital approval for a rack unit.

A hardware appliance is also the slowest possible way to learn you are wrong. Software companies find out in weeks. Nebula spent years building and supporting physical units before it had enough signal to know how big the market really was, and by then $38.5 million was gone.

~ 1 april 2015 ~

The shutdown letter went up on the company site and was confirmed on Twitter the same day, which meant a fair number of people in the industry spent a while checking the date on the calendar before believing it. The wording put the blame squarely on the market. The company said it was deeply disappointed that the market would likely take another several years to mature, and that as a venture-backed startup it did not have the resources to wait.

That turned out to be roughly accurate. Private OpenStack did become a real business, mostly for Red Hat and for telecom operators building their own infrastructure, several years after Nebula ran out of runway. Being early is indistinguishable from being wrong when the money is on a four-year clock.

~ from a distance ~

This was an enterprise procurement product, sold to people with a data center and a budget line for one, and none of that was ever within reach of an individual engineer in Ankara or anywhere else. What reached me in those years was the name OpenStack itself, which was in every infrastructure conversation between about 2012 and 2016 and then quietly stopped being in any of them. Nebula is one of the reasons that arc looks the way it does.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »