Big-data hotshot Cloudera has a new CEO: Tom Reilly, an industry veteran who has guided other tech firms to profitable acquisitions. But don't automatically assume that the privately held Cloudera is shopping itself around – an IPO is also a real possibility.
Judging from some changes among Cloudera's top brass, it looks like the so-called Red Hat of Hadoop may be looking to make a little trip down to Wall Street.
Mike Olson, one of the cofounders of the Hadoop distie, has been CEO of the company since it was set up in 2008 to commercialize the Hadoop big data muncher that was created by Yahoo! and open sourced. He is now Cloudera's chairman of the board and chief strategy officer, relinquishing his role in the day-to-day operations of the company. Olson is a database expert who was CEO of Sleepycat Software, the makers of the open source BerkeleyDB in-memory database that Oracle bought in 2006.
To create Cloudera, Olson teamed up with Amr Awadallah, who was vice president of product intelligence engineering at Yahoo! – which, if it had any sense, would buy Cloudera before it goes public and get its Hadoop mojo and smarts back.
Doug Cutting, who created Hadoop at Yahoo! based on his interpretation on some papers that Google published about its Google File System for unstructured data and its MapReduce algorithm for chewing on it, joined Cloudera in 2009 from Yahoo! and is chief architect.
Jeff Hammerbacher, who lead the data team at Facebook, a big Hadoop user, is another cofounder of the company and is chief scientist. There are a lot of chiefs at Cloudera.
Speaking of which, the new chief executive officer Tom Reilly started his job this week. Like Olson, Reilly comes from the database hotbed of the University of California at Berkeley, but actually got his bachelors degree in mechanical engineering, and has spent 30 years in the software business.
Reilly is a serial CEO and if history is any guide, Cloudera might get et before it makes its trip down to Wall Street to go public. Reilly was CEO at Trigo Technologies, a provider of master data management software that IBM bought in June 2004 for an undisclosed amount of dough. He then moved on to be CEO of ArcSight, a security company that got some of its early funding from the CIA's capital arm, and was acquired by HP in September 2010 for $1.5bn.
That acquisition was about a 7.5 multiple over ArcSight's annualized revenues at the time – a pretty hefty price, indeed. (But HP wants a software business real bad, which is how it ended up spending more than $10bn for Autonomy two summers ago.)
Cloudera raised $141m in five rounds of venture funding from a slew of investors, including Accel Partners and Greylock Partners, as well as receiving angel funding from a few software players who gotten rich over the years, including Diane Greene, cofounder and the first CEO at VMware, and Marten Mickos, who used to be CEO at MySQL and who is now in that same job at cloud controller software Eucalyptus Software. Jeff Weiner, who is CEO at biz social networker LinkedIn, is also an early investor in Cloudera.
Cloudera, being privately held, does not talk about its revenues, profits, and such, but earlier this year Charles Zedlewski, vice president of products at the company, told El Reg that Cloudera was "steadily moving out of the startup phase." At that time – four months ago – the company had more than 150 paying customers and over 320 employees. By the beginning of 2013, it had more than doubled its bookings and revenues over the same period in 2012.
It is difficult to say how much money Cloudera is making, but let's have a little fun.
In November 2011, Olson gave a presentation at Hadoop World in which he said that the average Hadoop cluster size for the companies that were attending the show was 120 nodes, up from 66 nodes in the fall of 2010. It is hard to estimate the current average cluster size, but in those numbers were some very large clusters, like those at eBay and Yahoo!, so let's say it has slowed down some and that the average cluster size is 256 nodes. Cloudera is cagey about its pricing, with license fees in the range of $2,000 to $4,000 per server node per year for commercial-grade support for its Hadoop stack. Call it $3,000.
That works out to around $115m in support revenues per year. Add in services and other things, and you could push it to $150m. I think it could be as large as $300m annualized run rate based on larger cluster sizes and a top-heavy feature set in Cloudera Manager.
That's roughly twice the bait that the investors have put in, and it also works out to around $940,000 per employee. Let's see if this is reasonable.
Red Hat had $1.13bn in sales in 2012, and has 5,500 employees, or about $205,500 per employee. Hadoop skills are in short supply and in high demand, which is not the same case with Linux – but even still, nearly $1m per revenue per employee is kinda high. So it seems likely that the $150m revenue figure is right.
Provided that Cloudera is not losing too much money with all of the database development (mostly Project Impala) and marketing that it has been doing, then it could fetch a very high multiple if someone were to acquire it or if it were to go public.
If you play around with some Red Hat revenue and market cap numbers, and assume the same ratio of market cap to revenues and also assume that the inflationary effects of Hadoop excitement are the same as actual profits at Red Hat (maybe not a fair assumption), then Cloudera is worth maybe between $1.2bn and $1.5bn. ®