Remember Cray Computer? The company made massive, screaming-fast mainframes back in the 1970s. Founder Seymour Cray was a tech rock star when Steve Jobs was still puttering in his garage and Jeff Bezos rode a bike with training wheels.

Managing today’s crushing mountains of un-crunched data means supercomputers have taken on a new luster, and Cray (NSDQ: CRAY) is having a resurgence. Despite improvements in computing, standard computers can’t process the volume of data generated for government, academic and commercial purposes fast enough.

Market research firm IDC estimates the demand for supercomputers will grow 8.6% a year in the next five years, topping $15 billion in 2019.

And now more commercial customers (as opposed to government agencies) use Cray’s supercomputers to analyze the tsunami of data landing on their servers. Earnings per share more than doubled in 2015 to $1.30 from 60 cents and should grow 15% this year and 20% next. The stock currently trades around $30 per share, and our Target is $40.

The stock is cheap now—down from its 52-week high of $43.79—given a hiccup in production a a few months ago that trashed Cray’s stock price. In early May near term earnings were lowered due to a component shortage. Cray has since secured those components, putting supercomputer shipments back on track. In late June, Cray announced new contracts for various supercomputers.

Beating the Bottleneck

We’ve been eyeing Cray’s stock since late last year when the company’s strong commercial customer base pushed the share price higher. Government customers, outfits like NASA or the National Nuclear Security Administration, have been Cray’s bread and butter. While government agencies make good customers, their orders tend to be big and infrequent, wreaking havoc on the predictable growth Wall Street likes.

With the company doubling the proportion of commercial revenue to more than 15% last year, Cray seemed on the road to more stable growth. Then production delays for Intel’s Knight’s Landing processor and Nvidia’s Pascal chip interrupted that trend. Without these processors, some orders that Cray had hoped to deliver in the second and third quarters hit bottlenecks.

All three of Cray’s major new supercomputers scheduled to debut this year rely on next-generation processors, which hadn’t been released when the company provided 2016 guidance. Clearly, any delay with these processors would affect the distribution of Cray’s revenue across the year.

In February management suggested that the timing of the new chips meant half of 2016 revenue would occur in the fourth quarter. At an investor conference in March, CEO Paul Ungaro emphasized that even with new commercial customers, the company’s revenue flow wouldn’t be steady: “Our business comes in large chunks, and we can’t always control the timing of those orders.”

The explanation didn’t satisfy investors. Cray’s share price fell 20% in May, when an additional 10% of annual revenue was pushed into the fourth quarter. We think the market overreacted. Investors shouldn’t have been surprised about Cray’s well-documented lumpy revenue, which eventually boosts the bottom line, sometimes in a big way. Last year’s revenue benefited from $40 million in orders that shifted from 2014 into 2015.

Intel’s Knight’s Landing processor was officially released in late June. This suggests computers using the chip will ship by year-end. Nvidia’s Pascal processor, which is less critical to Cray making its annual targets, was released in late May. These releases are probably too late to contribute to Cray’s second quarter but should boost the annual number.

One reason for revenue coming in fits and starts is Cray doesn’t sell many of its supercomputers, but the ones it does sell are super expensive. The CrayXC40, used by the British National Weather Service, for example, costs $145 million. Though smaller than the government’s orders, orders for commercial customers are still large, and a few weeks delay by a handful of those customers can make or break a quarter.

Solid Numbers

Although Cray’s numbers are lumpy, the profits still add up on its balance sheet. As of March, the company had $320 million in cash and zero debt. It generated $140 million in free cash flow last year and should grow earnings at least 20% in 2017. Our $40 target is based on a multiple of 22 on 2017 earnings per share of $1.80.

Father of Supercomputing

Founder Seymour Cray (who died in 1996 at age 71) is still called “the father of supercomputing.” Cray’s first supercomputer monitored national security and researched nuclear energy for the Los Alamos National Laboratory. Early versions of Cray’s computers ran at gigaflop speeds: a billion operations per second, which morphed into a trillion operations per second (teraflops) in the mid 1990s. Current models handle 23,000 trillion calculations per second and are needed to crunch the exponentially growing amount of data fed into complex mathematical models.

For example, weather forecasters need teraflop power to warn us of increasingly volatile storm patterns. Hedge funds use supercomputers to predict commodity prices, and pharmaceutical companies need them to unravel how tumors grow.

The company’s Urika-GD computer has been used to detect fraud, monitor cybersecurity threats and improve Major League Baseball lineups. Recent customers include Kyoto University, the Danish Meteorological Institute and Polish steel company Stalprodukt, which uses Cray supercomputers to improve steel-making techniques.