Slashdot Mirror
IBM Pays GlobalFoundries $1.5 Billion To Shed Its Chip Division
helix2301 writes with word that Big Blue has become slightly smaller: IBM will pay $1.5 billion to GlobalFoundries in order to shed its costly chip division. IBM will make payments to the chipmaker over three years, but it took a $4.7 billion charge for the third quarter when it reported earnings Monday. The company fell short of Wall Street profit expectations and revenue slid 4 percent, sending shares down 8 percent before the opening bell.
I can't speak for any of IBM's other decisions, but in this case I have to strongly disagree with you. The IBM semiconductor business has been losing money hand over fist recently. They can't compete with Samsung or TSMC on price and volume, and there's not enough interest in specialty chips or POWER to make up the slack. It costs at minimum $5 billion to build a new fab, and IBM would have to build at least one, maybe two new fabs, not to mention updating their existing fabs, in order to be competitive with the big guys.
So, IBM could spend $5 billion - $10 billion just to catch up to their competitors, and still be at a very serious risk of the division being unprofitable, or they could spend $1.3 billion knowing for certain that the bleeding will stop. I only wonder what took them so long.
Also, for what it's worth, IBM is allegedly doing this deal in part so that it can focus more money into design research. They've announced a $3 billion investment into their semiconductor research division, which they aren't getting rid of. The implication is that the manufacturing division was crowding out any other R&D spending, and that IBM can now focus on high margin ARM-style licensing instead of getting dragged further into a war with TSMC et al. that they would inevitably lose.
A couple of reasons. First as others have mentioned IBM still needs the lines. IBM's processor design is fairly integrated. It needs custom circuits which really on their fabrication technology. Their chip design process is the antithesis of fabless development. So they can't just shut off lights to the fab without crippling the Power, Mainframe, and high end storage business for years to come. The other issue is there are customers getting chips manufactured. If you shut that down there are typically very steep contract termination fees. This is really 6 year wind down with higher costs every year. The fab business is very cutthroat, you have to hit fairly high yields and have the line near full capacity or else you lose a boatload because the fixed costs are so high. They've been getting out of this business for years, especially on the low end. This is just the last step.