Cray posts modest loss, awaits Q4 revenue wave

Technical issue with 'Titan' super – now fixed – eats into profits

The essential guide to IT transformation

Supercomputer maker Cray turned in a slightly better than expected second quarter, with a modest bump in revenues and a slight loss.

This is a win in the ever-choppy supercomputer racket – and particularly for Cray in a second quarter. For the past three years, Cray has been cramming a large portion of its sales in the final quarter of the year thanks to processor and GPU coprocessor transitions and the normal end-of-year budget crunch that all IT vendors generally see.

Very few organizations want to buy last year's model if they can avoid it, unless their applications will work well on it and they can get a good deal.

In the quarter ended in June, Cray's product sales were down 9 per cent to $62.4m, but its services revenues were up 41 per cent to $22.1m. Total sales were actually up a fraction of a point to $84.2m. However, as Cray is building machines now for acceptance later this year, its costs went up (as expected), and it also nearly tripled its research and development expenses while boosting its sales, marketing, and other costs a little bit. That pushed Cray to an operating loss of $9.8m, and after posting an income tax benefit, Cray reported a net loss of $150,000 compared to a net income of $147.4m in the year-ago period.

Cray booked the sale of its "Gemini" and "Aries" interconnect business to Intel in the year-ago period, which gave it $139.1m net of expenses from the deal. If you take out the Intel deal, the swing from profit to loss is a lot smaller, obviously.

Cray's non-GAAP numbers, which take out the effects of the interconnect sale as well as the acquisition of cluster maker Appro International and some other costs, show Cray reporting a $7m net loss in the second quarter of this year compared to a $12.8m gain a year ago. That difference is a little bit smaller than the net change in product costs plus the increase in R&D spending.

In its conference call with Wall Street analysts, Cray did not elaborate on what projects ate up $19.9m in R&D budget during the quarter, but presumably there are costs to bring Intel's latest "Ivy Bridge-EP" Xeon E5 processors to the "Cascade" XC30 supercomputers that are based on the Aries interconnect. There should not be much cost there, given that the Ivy Bridge processors are socket-compatible with the current "Sandy Bridge-EP" Xeon E5s used in the XC30s.

Cray is also working to integrate Nvidia's Tesla GPU coprocessors and Intel's Xeon Phi x86 coprocessors into the XC30 machines. Both of these projects are expected to be completed in a matter of months. Hopefully Cray is up to something else interesting with all that R&D spending, and will talk about it soon.

But CEO Peter Ungaro did say in the conference call that gross margins were adversely affected by a "technical issue with a component" that resulted in a higher-than-expect cost for that component as it was deployed in the "Titan" XK7 ceepie-geepie system at Oak Ridge National Laboratory.

Ungaro said that without this technical issue, gross margins would have been "very strong," and that it would not have an impact on future quarters. El Reg has contacted Cray to get some insight into exactly what the issue was, but the company has not returned calls at press time. Whatever the issue was, Cray fixed it – Titan has been accepted by Oak Ridge, and the remaining revenue for the system has been booked.

On the storage front, Ungaro said storage sales were "a little light" in the quarter, but that Cray had bagged two big Sonexion storage system sales, one at the European Centre for Medium-Range Weather Forecasts and the other at UK's national supercomputing service at the University of Edinburgh.

These Sonexion sales were tied to XC30 systems, which Cray is obviously very pleased about. But Cray has aspirations to sell Sonexion arrays and its Cray Cluster Connect, or C3, implementation of the Lustre clustered file system outside of its own system sales, and is expanding out from traditional HPC workloads to big data and other kinds of jobs.

Ungaro said on the call that the YarcData division, which is peddling the Urika graph analytics appliance based on Cray's homegrown "ThreadStorm" processor and XMT system, was "showing signs of real traction."

Unfortunately, the customers who are using the Urika appliance are not the kind who like to talk about their systems. Ungaro said that one big government customer made a "significant addition" to its Urika system and two new government agencies bought the appliances, too, during the quarter. He didn't say which government, but presumably it is the US federal government.

Cray has sold a Urika appliance to a telecommunications firm that is using it to see how service can impact customer churn, a financial services firm is using a Urika to do analytics for cross-selling and upselling to customers, and a bank is using the graph analytics appliance to do risk management and compliance.

Cray ended the quarter with $251m in cash and investments. Cray's inventory is sitting at $126m at the moment, and CFO Brian Henry reminded Wall Street that the company does not buy parts until it anticipates deals, and the reason inventory is so high right now is because Cray has a good feeling about the second half of the year.

In fact, Cray is raising its revenue guidance for 2013, saying it will be $20m higher than previously expected and will come in around $520m. Revenues in the third quarter are now expected to be around $90m, and about half the year's revenues to be recognized in the fourth quarter.

Cray had been expecting for around $135m in revenues for the third quarter when it provided guidance three months ago, so $45m of sales slipped into Q4 and another $20m was added. (The odds are that the revenue push has to do with timing of the ramp of Intel's Ivy Bridge Xeon E5 chips, but neither Cray nor Intel will talk about that.)

As long as Cray can get it done by December, Wall Street will be happy. ®

Next gen security for virtualised datacentres

More from The Register

next story
6 Obvious Reasons Why Facebook Will Ban This Article (Thank God)
Clampdown on clickbait ... and El Reg is OK with this
No, thank you. I will not code for the Caliphate
Some assignments, even the Bongster decline must
Caught red-handed: UK cops, PCSOs, specials behaving badly… on social media
No Mr Fuzz, don't ask a crime victim to be your pal on Facebook
Barnes & Noble: Swallow a Samsung Nook tablet, please ... pretty please
Novelslab finally on sale with ($199 - $20) price tag
Ballmer leaves Microsoft board to spend more time with his b-balls
From Clippy to Clippers: Hi, I see you're running an NBA team now ...
Banking apps: Handy, can grab all your money... and RIDDLED with coding flaws
Yep, that one place you'd hoped you wouldn't find 'em
Video of US journalist 'beheading' pulled from social media
Yanked footage featured British-accented attacker and US journo James Foley
Call of Duty daddy considers launching own movie studio
Activision Blizzard might like quality control of a CoD film
Primetime precrime? Minority Report TV series 'being developed'
I have to know. I have to find out what happened to my life
prev story


A new approach to endpoint data protection
What is the best way to ensure comprehensive visibility, management, and control of information on both company-owned and employee-owned devices?
Implementing global e-invoicing with guaranteed legal certainty
Explaining the role local tax compliance plays in successful supply chain management and e-business and how leading global brands are addressing this.
Maximize storage efficiency across the enterprise
The HP StoreOnce backup solution offers highly flexible, centrally managed, and highly efficient data protection for any enterprise.
How modern custom applications can spur business growth
Learn how to create, deploy and manage custom applications without consuming or expanding the need for scarce, expensive IT resources.
Next gen security for virtualised datacentres
Legacy security solutions are inefficient due to the architectural differences between physical and virtual environments.