Hortonworks is zigging while the rest of the market is zagging, and it doesn’t seem to make sense.

I’ve been a harsh critic of Hortonworks’ business model in the past, and lambasted its financials when it first filed to go public. But, over the past several quarters Hortonworks has demonstrated an ability to sell lots of free software, suggesting I was premature in my eulogy.

Given this fact, and the overall hotness of big data, it can be confusing why Hortonworks’ stock price sits at bargain basement levels while the overall market for Hadoop is booming at a 53% CAGR.

At least, until you look at the bottom line.

Selling hotness

Open source is hot. Big data is hot. But somehow Hortonworks is not. At least, not at the level its financials and its markets would seem to deserve.

Valued at just $858 million (as of this writing), well below the $1 billion valuation it earned after its IPO, Hortonworks’ stock price has been sagging over the last few weeks. This despite announcing yet another quarter of triple-digit year-over-year growth, with billings surging 104% and subscription revenue booming by 129%.

Oh, and the company generated eight deals over a $1 million, while also adding 152 new unique support subscription customer logos (up 311% year-over-year).

What’s not to love?

In fact, according to Drexel Hamilton analyst Brian White, Hortonworks remains “one of the few pure plays on the Big Data movement, the only publicly-traded, independent Hadoop player and one of the few plays on the open source software movement.”

Call me when you’re profitable

All of which sounded great until Hortonworks acknowledged that it will keep losing money until early 2017. To be fair, Hortonworks has not made a secret of this: The company keeps making progress toward profitability, as CFO Scott Davidson noted in Hortonworks most recent earnings call: “We are on the march to…narrow the adjusted EBITDA loss.”

Even so, as one financial analyst told me, “Many Investors don’t believe in Hortonworks’ revenue model, and profitability is the litmus test.”

Red Hat, which pioneered the support subscription model that Hortonworks follows, saw its stock spike 20% in 2015. But Red Hat has also demonstrated, quarter after quarter, that it can profitably sell open-source software (services).

Hortonworks, to date, has not.

This wasn’t an issue back when the market was running on unicorn fumes. Lately, the tenor of the market has changed, however, and investors have lost patience with money-losing companies. Even poster children for big data.

It has become clear that Hortonworks, not to mention still-private peers like Cloudera, DataStax, MapR, and MongoDB, can make lots of top-line revenue. What is less clear, however, is if any of these can turn a profit. The first company that manages to do that will win, and win big.

Until then, former Wall Street analyst Peter Goldmacher’s suggestion that the biggest beneficiaries of big data wouldn’t be the ones selling big data infrastructure seems more and more prescient.