by Arik Hesseldahl

Cloudera’s IPO is overshadowed by a rival it won’t mention

Apr 26, 2017
Big DataCloud ComputingHadoop

Big data's poster child is selling itself to investors as a machine learning company, but is really still in the Hadoop business just like Hortonworks, the competitor it would like to ignore.

One of the original poster children of the big data software craze, Cloudera, is due for its long-awaited IPO this week. Sometime Thursday afternoon, its shares will price somewhere between $12 and $14 and will open for trading on the New York Stock Exchange the following morning, raising about $200 million in the process.

Make no mistake, this IPO qualifies as what’s known in venture capital circles as a down round. Essentially the new investors are buying shares in a company that is worth less than it was during its prior funding rounds.

When Cloudera last raised money in 2014 it took a jaw-dropping $740 million from the chipmaker Intel, which paid a little less than $31 a share for a stake in the company amounting to about 19 percent. That investment implied that Cloudera was then worth about $4.1 billion. Depending on its final price, when trading on Cloudera shares opens on Friday, Cloudera will be worth less than half that: About $1.8 billion.

What happened to that valuation? A lot of things, but as the IPO process progressed, it became clear that there was no way that Cloudera could justify a valuation north of $4 billion, especially when compared to a publicly held competitor it has tried mightily to ignore: Hortonworks.

Cloudera and Hortonworks compete in selling and servicing open-source software for analyzing data called Hadoop, and a set of related technologies. Hadoop was a central factor in the “big data” hype cycle that rocked the enterprise software world earlier this decade. That hype cycle has cooled, and shifted toward newer, shinier buzzwords including artificial intelligence and machine learning.

And yet the more you look at Cloudera’s S1 filings with the U.S. Securities and Exchange Commission, Cloudera minimizes mentions of Hadoop, which is still very much its primary line of business, mentioning it 14 times in a 150-page document. The new emphasis is on “machine learning” a phrase it repeats 70 times. 

But Hadoop is still without question, Cloudera’s main business, just as it is the main business at Hortonworks. At Cloudera, Hadoop gets packaged with proprietary software in an approach it calls Hybrid Open Source Software or HOSS. The HOSS approach, it argues, is aimed at large corporate customers who want an “enterprise-grade” product.

Hortonworks has for its part sought to stay true to its roots in Hadoop while evolving into adjacent businesses in order to address wider markets. It has built enterprise-grade products for security, governance, and operations. One example is a dedicated streaming analytics platform called Hortonworks DataFlow intended for use with Internet of Things applications. Cloudera customers are told to use Spark Streaming which isn’t comparable.  

As you work further through the document you first see Cloudera start to squirm awkwardly on the subject of Hortonworks in the only place the latter company warrants a mention. In the section where Cloudera names its competitors, and break them into four tiers. Hortonworks and another Hadoop company MapR appear only in the fourth, as if to say “Oh, them… Are they still around?”

Yes it is, very much so. Hortonworks has ties to all three of those first-tier competitors: Hewlett Packard Enterprise (HPE) is a strategic investor in Hortonworks, and the other three, IBM and Oracle and Teradata are partners who resell Hortonworks. Where Cloudera sees competitors, Hortonworks sees allies. 

And the further you compare it Hortonworks, the more it makes sense to compare them, because it turns out that Hortonworks stands up pretty well.

Consider annual revenue: Cloudera, founded in 2008, posted revenue of $261 million in its most recently completed fiscal year. By itself that figure would be fine for a software company first coming public. And yet when compared to Hortonworks, founded in 2011, it’s easy to ask yourself “Is that all?”

Despite a three-year operational head start over Hortonworks, Cloudera’s annual revenue was only $77 million ahead of Hortonworks. (Hortonworks has been public for nearly three years and posted revenue of $184.5 million in its fiscal 2016.) Put another way: Cloudera had a three year head start, and yet the smaller Hortonworks has already grown to 70 percent of its size.

By one key financial metric — billings, or the combined value of invoices sent to customers in a period — Hortonworks is growing faster than Cloudera. At Hortonworks billings grew year-on-year by 63 percent to $270 million while at Cloudera they grew 54 percent to $320 million. Yet again, Cloudera seeks to change the subject. Billings, Cloudera argues in the filing “are not an accurate indicator of future revenue.”

None of this is to suggest that Cloudera won’t have a successful IPO this week and make its venture capital investors — mainly Greylock and Accel who combine for nearly 29 percent of Cloudera’s equity — very happy. Tech IPOs have remained popular. And most of the tech companies who completed their IPOs in 2016 got off to good starts but some have seen their share prices dip below the offering price since then. We’ll soon see if this one was worth the wait.