The pioneer of supercomputers in the 1970s stood on the brink of obscurity 20 years ago but is now surging back to prominence. Its shares have almost doubled over the past 12 months.
The explosion of data - measuring weather, traffic, health and countless other areas - coupled with a desire to tease meaning out of it, demands greater computing power than is accessible via standard machines.
"The assumption was that supercomputers were cliche five years ago. People thought, 'I can run my simulation on my laptop'," said Barry Bolding, a Cray vice president, at the company's Seattle headquarters last week. "That may have been true, so long as the data associated wasn't growing as well. But raw data is being created in exabytes as we sit here. More data means bigger computer, bigger computer means more data."
Experts estimate that 2.5 exabytes - or 2.5 billion gigabytes - of data are now generated every day, and the world's capacity to store that data is doubling every 40 months, which all plays to Cray's strengths.
A basic Cray cabinet costs $500,000 and up and is roughly the size of a refrigerator. Big customers can group 200 or more into massive supercomputers worth hundreds of millions of dollars, such as "Titan" at the U.S. Department of Energy's Oak Ridge National Laboratory.
Titan, completed by Cray last year, is the world's third-fastest supercomputer, takes up the size of a basketball court and can perform more than 20,000 trillion calculations a second.
To be sure, most companies will never need that scale, or can process what they need through multiple machines running in tandem on a high-speed network or in the cloud, which for many projects works out cheaper and more power-efficient.
What makes supercomputers different is that they can make a huge number of interconnected calculations at the same time, rather than a consecutive list of unconnected calculations, which makes them good for running complex simulations and mining unrelated data.
For example, weather apps on smartphones are based on vast models run by research agencies on supercomputers. Financial firms can detect online fraud or cybersecurity breaches in seconds rather than days by using supercomputer models, which would take days on standard set-ups.
"Big data is a new term, but arguably the supercomputer market was the original home of big data, and Cray has been dealing with it forever," said Steve Conway, an analyst at tech research firm IDC.
Market on fire
The Seattle-based company, with just over 900 employees and a market value of around $940 million, has changed ownership several times but was started in 1972 by Seymour Cray, the "father of supercomputing."
With a recent resurgence in supercomputers, Cray is garnering Wall Street's attention. This June, it sold one of its new XC30 supercomputers to the European Centre for Medium-Range Weather Forecasts for $65 million, nabbing a contract from a long-time IBM customer.
That sort of deal is piquing investor interest. Wall Street analysts are expecting revenue of $519 million this year, up 23 percent from 2012, with a gross profit margin around 34 per cent. Its shares are up 91 per cent over the past 12 months while rival Silicon Graphics International Corp's are up 90 per cent. Cray is now richly valued, with a share price 36 times estimated earnings for the next 12 months, compared with 19 times for SGI.
The global market for computers costing more than $500,000 is on a tear, according to IDC, having more than doubled to $5.6 billion in 2012 from $2.7 billion in 2008.
The whole market for high-performance computing (HPC) - essentially any machine bigger than a desktop used for intense computation - is forecast to grow 7 per cent a year through 2017, well ahead of the stagnant business server market.
The U.S. government directly or indirectly accounted for two-thirds of Cray's revenue last year. But the company is reaching out to new customers interested in big data. Last year it set up a new unit called YarcData - Yarc is Cray backwards - to focus on analyzing huge amounts of information and teasing out unseen patterns in a process known as graph analytics.
"Unstructured databases are becoming more prevalent, gathering raw data from everywhere," said Cray's Bolding. "Now you start asking very complex questions, and it starts to create links between sets of data."
The YarcData unit is helping the U.S. government detect fraud patterns in Medicare and Medicaid payments. Private sector customers include medical research group Mayo Clinic and several financial services, life sciences and telecommunications firms, which Cray cannot name for contractual reasons.
New efforts are working and should boost revenue over time, said Sid Parakh, an analyst at fund firm McAdams Wright Ragen.
"This is not a commodity market. It takes years of experience," said Conway at IDC. "It's easy to build a big computer, but it's not easy to build a big computer that works."