Hortonworks breaks free from death spiral, but cloud is still a threat

Hortonworks finally had a good quarter financially, but it's not out of the woods yet.

Image: iStockphoto/kentoh

It hasn't been a good year for Hortonworks. At least, not by Wall Street's standards. Until Hortonworks reported earnings this past week, the company's valuation had been in a death spiral, dropping from $1.1 billion to under $400 million over the course of the year, fueled by missed revenue targets and illusive profitability. In Q3, Hortonworks kept missing on earnings targets but delivered better-than-expected revenue, with promises to exceed analyst expectations in Q4.

Is the sun finally shining on Hortonworks? While it's still too soon to call definitively, there are signs that Hortonworks may be on the financial mend.

Running a tight ship

Ovum analyst Tony Baer has asked if we've "reached peak Hadoop?," but according to Hortonworks' results the answer is clearly "no."

"[E]ven after years of commercial packaging, Hadoop is still a vendor-curated collection of projects that is complicated to deploy," Baer wrote. Yet, enterprises have been willing to go through the pain because of the outsized potential if they figure out Hadoop and its big data peers.

For Hortonworks, this translated into a 47% revenue increase over the previous year. Last quarter missed earnings sent the stock down 22% in a day. Last week on rosier projections of revenues, the stock jumped 22% in a day. Clearly it's not a stock for the nervous investor.

SEE: Hadoop numbers suggest the best is yet to come

Even so, subscription revenue has outpaced total revenue, resulting in higher-margin revenue. The company did 10 deals worth over $1 million each, the most seven-figure deals the company had ever managed in a single quarter. Further adding to the relatively positive news, the company's renewals returned 140% of their original value.

Couple this with much tighter sales execution--the field now reports directly to CEO Rob Bearden, a notoriously tough taskmaster--and things are looking up for a company whose stock has invariably trended down.

On-ramp or exit?

The one area that doesn't bode well for Hortonworks, or any of its on-premise Hadoop peers, is cloud. Despite CEO Rob Bearden's optimism that "we view cloud...as an on-ramp," with "some of [AWS'] workloads migrat[ing] to us," this is more wishful thinking than anything else. Hortonworks' head of strategy, Shaun Connolly, has argued that modern applications are "inherently hybrid," spanning data center and cloud, because data lives in and must be processed in both places, the trend is clearly away from data center and toward public cloud.

SEE: Some Hadoop vendors don't understand who their biggest competitor really is

In such a cloud world, the real on-ramp leads to AWS (or Microsoft Azure or another public cloud option) and away from on-premise solutions. This is certainly the way Amazon thinks of it, as revealed on its latest earnings call: "We continue to invest in...technologies that make integrations easier [to] help[] companies move from an on-prem or a hybrid IT environment into AWS."

In a battle between on-ramps, the safer money would be on AWS. This is why Hortonworks would be wise to stop positioning against its on-premise Hadoop rival, Cloudera, and instead fixate on AWS.

Not that Hortonworks is completely poo-pooing the cloud, given that 25% of its customers run Hortonworks in a public cloud environment today. Throughout its earnings call, Hortonworks talked up its partnership with Microsoft Azure, touting Azure HDInsight that integrates with the Hortonworks Data Platform ("We actually got very, very focused on selling the HDI service in the quarter" because "We get paid irrespective of where [a] workload executes"). The company also pitched its integration with AWS.

All of this is positive, but given the inherent need for flexible data infrastructure, Hortonworks needs to get cloudier, quicker, if it hopes to match its customers' shift to the public cloud. Far too often the on-ramp leads to AWS or Azure, and not the other way around. Unless the company can nail this problem, no amount of improved execution or expense management will save it.

Also see