Fujitsu says biz to pick up in 2010
Wins 10 petaflop Japanese super deal
Kuniaki Nozoe, president of Japanese IT giant Fujitsu, gave investors and customers an update on the company's financial situation in Tokyo this morning, and said that thanks to restructuring and acquisitions, he was hopeful that Fujitsu would be able to get profitable this fiscal year and boost profits in the following year.
Nozoe did not hold out much hope for a recovery in fiscal 2009, which ends next March, but said he expected "a mild recovery" to start during that term and added that a "large sales expansion through organic growth is unrealistic."
For the fiscal 2009 year ending March 2010, Fujitsu projects that it will have aggregate sales of ¥4,800bn ($50.4bn) across its disparate hardware, software, and services product lines. That's a 2.3 per cent increase over the ¥4,693bn ($49.3bn) in sales posted in its fiscal 2008 ended in March 2009.
Fujitsu's operating income is projected to be around ¥80bn ($841m, up 16.4 per cent) in the current fiscal year, and it hopes to swing from a ¥112.3bn ($1.2bn) net loss in fiscal 2008 to a ¥20bn ($200m) net income this fiscal year.
Looking ahead, Nozoe said that Fujitsu now hops to boost sales in fiscal 2011 to ¥5,000bn ($52.5bn), an increase of 4.2 per cent over this fiscal year's projected sales, with operating income projected at ¥250bn ($2.6bn) and net income of ¥130bn ($1.4bn). That operating profit is nearly twice as high as analysts had forecast, for what that is worth.
Those are some pretty hefty projections, considering that Fujitsu's previous record for operating income was ¥244bn ($2.6bn) in fiscal 2000 and for net income was ¥102.4bn ($1.1bn) in fiscal 2006.
But selling off its disk business to Toshiba, outsourcing its chip manufacturing, and taking control of the PC and server units in EMEA from former partner Siemens have given Fujitsu the opportunity to cut costs and boost earnings, the company says.
Nozoe said that Fujitsu's goal, in fact, is to get more than 40 per cent of its sales outside of Japan in fiscal 2011, and that it will do this through a "think globally, act locally" strategy, which means getting rid of siloed business units that operate outside of Japan.
Nozoe also said that getting its chip business to profitability was the biggest challenge facing Fujitsu. According to a report from Reuters, Nozoe told analysts that its money-losing chip operation would be profitable sometime between October and December, and would bring in some profits on an annual basis in fiscal 2010.
In the systems arena, Fujitsu has set a goal of peddling more than 500,000 x64 servers in fiscal 2010, a feat made possible now that it has taken control of the EMEA partnership it formerly shared with Siemens.
The company also wants to unify its supply chains into one giant supply chain to cut costs - which IBM did to cut billions of dollars out of its operational costs - and expand sales of its Eternus disk arrays in Europe. Fujitsu also wants to sell 10 million PCs a year and ramp up sales of IT services in Europe and North America though its Fujitsu Services arm, which has 25,000 employees and sales of ¥500bn ($5.25bn) today.
In a related story - as El Reg had anticipated - Fujitsu has taken over the 10-petaflop massively parallel supercomputer contract sponsored by the Japanese government and variously dubbed Project Keisoku or the Next Generation Supercomputer Project. This $1.2bn project called for the three key server makers in Japan - NEC, Hitachi, and Fujitsu - to cooperate in building a 10-petaflop machine made up of scalar processors designed by Fujitsu, vector processors designed by NEC and Hitachi, and a system design created by all three.
But in May, just after Fujitsu announced it had the eight-core Venus Sparc64-VIII processor to be used in the scalar half of the box running in the labs, NEC announced that although it had completed the design of the vectors with Hitachi, it was pulling out of Project Keisoku because it was bleeding red ink in general and making the vectors for the machine would only make matters worse.
In any event, the Rikagaku Kenkyusho (Riken) research lab in Kobe, Japan, that is managing Project Keisoku has indeed announced that it will now build a completely scalar supercomputer based on the future Fujitsu Venus chips.
This was no surprise, since it had little choice to save political face. The surprise is that this Sparc machine will not run Solaris - well, at least a surprise for Oracle as it acquires Sun Microsystems - but rather a Fujitsu variant of Linux. If Fujitsu sticks with the 128 gigaflops per chip rating it announced for the Venus prototype in May, then it will take 625,000 cores (that's 78,125 chips) to hit a peak theoretical performance of 10 petaflops.
Fujitsu hasn't said if it will make any money on the chips or not, but presumably taking over the Project Keisoku machine all by its lonesome means it's making money on the deal, particularly given that the Japanese government doesn't really have a Plan B if it wants to use indigenous technology - though to some ways of thinking, Fujitsu already was Plan B.
This means that the Japanese government has to either come up with the cash to keep Fujitsu whole or shift to x64, Sparc, or Power chips made by non-Japanese companies. This is politically impossible, and given the stakes it's a wonder that the Japanese government didn't just come up with the extra cash needed to make NEC and Hitachi whole and get the hybrid box that Riken clearly wanted.
As we reported earlier this week, the Japan Atomic Energy Agency is having Fujitsu build a 200-teraflop x64 blade server cluster based on Intel's Nehalem EP Xeon 5500 processors, as well as giving the nuclear fission and fusion lab some Sparc64 iron as it works on software development for the Project Keisoku cluster.
Fujitsu and Riken say they will get bits of Project Keisoku operational in fiscal 2010 and will have it completely installed by 2012. The data center and research labs for the supercomputer are under construction now. You can find out more here. ®
Sponsored: Hyper-scale data management