Big Blue’s Chip Business Is Probably On The Block

February 17, 2014 Timothy Prickett Morgan

This rumor, should it turn out to be true, will not come as much of a surprise to IBM i shops that read The Four Hundred. Time and time again, as I have watched IBM‘s quarterly financial results, I have wondered at what point the chip manufacturing business would be too expensive for Big Blue to stay in. We may be approaching that point.

Tongues have been wagging for a long time on this issue, particularly in the wake of IBM’s decision to sell of its System x X86 server business to Lenovo for $2.3 billion a few weeks ago. IBM has been leaving various parts of the hardware business for several decades, including the manufacturing of memory chips and disk drives in IBM’s Rochester, Minnesota, labs in years gone by. Hitachi bought the disk business, and it is now part of Western Digital and, interestingly, there are only two suppliers of enterprise-grade disk drives left. (Seagate is the other one.) Ricoh has the IBM high-end printer business, Lexmark has the low-end printer business, Lenovo has the PC business, and Toshiba has the retail store systems business.

A report in the Financial Times started off the latest round of hardware selloff rumors, citing unnamed sources who claim that Big Blue has hired investment bank Goldman Sachs to try to scare up possible buyers for its chip plants in East Fishkill, New York, and Burlington, Vermont. The Vermont plant has pretty old technology, while the East Fishkill plant has recently been upgraded to etch chips using the 22 nanometer processes that the impending Power8 processors, due later this year, will employ. The FT report said that IBM was not determined to ditch the chip business, but was exploring its options. A separate report in the Wall Street Journal corroborated the rumors but cleared nothing up.

IBM has been on the forefront of chip-making technology since it entered the computer age 50 years ago, and it has a wealth of intellectual property related to chip design and manufacturing as well as a certain amount of stature amongst its peers. But this is a game of volumes just like the PC, disk, and memory business. There is simply no way for a low-volume chip manufacturer to pay the billions of dollars that are necessary to invest in chip making equipment and design tools. While IBM was manufacturing PowerPC chips for embedded devices such as cars and electronics as well as the custom Power chips for Microsoft Xbox and Sony PlayStation, and Nintendo Wii game consoles, the numbers worked. Power chip volumes were higher a decade ago, and mainframes sold in higher volumes and at higher prices, too. But IBM has lost the Microsoft and Sony console business to Advanced Micro Devices and Nintendo is not doing great with the Wii consoles.

Toni Sacconaghi, who is the tech analyst at Sanford Bernstein and without a doubt one of the smartest of the Wall Street IT watchers, figures that the IBM Microelectronics portion of the Systems and Technology Group had $1.75 billion in sales in 2013 and booked a $130 million pre-tax loss. Sacconaghi figures that IBM might bring in $1.45 billion in sales for chips in 2014 and book another $130 million pre-tax loss. This is not helping IBM toward its 2015 Roadmap plan of bringing $20 per share to the bottom line as 2015 comes to an end.

When you consider that each process node costs incrementally more to develop, and plants keep getting more expensive as well as chip etching becomes more complex, it is no wonder that IBM wants out. IBM will probably try to sell the fabs to another chip maker that is looking for a guaranteed customer to recoup those costs. IBM could probably afford higher Power and System z chip costs if it didn’t have to do the major investment in chip research and plant. The only obvious buyers are GlobalFoundries, the chip-making spinout from AMD and already an IBM partner and a company that operates its own high-tech chip fab up near Albany, New York. Taiwan Semiconductor Manufacturing Corp, which makes Sparc, Sparc64, and Opteron processors already, might be interested in the IBM chip business, and there is an outside possibility that Intel might be or possibly the Chinese government. The odds of a Chinese company being able to buy IBM’s chip business is about the same as you or me winning Powerball. GlobalFoundries is the best partner for a lot of reasons, and the question is what does IBM want for it. The business has a lot less revenue than the System x business, and is not making money.

The important thing is that IBM can’t mess this up or it could jeopardize its Power Systems and System z business, which still generate plenty of revenues for the company and which drive an even larger amount of software and services revenue. This is a very delicate situation, indeed. Maybe Lenovo will want to get into chips and save IBM’s cookies? Somewhere back in the IP pile is a license from Intel to make X86 processors. Now wouldn’t that be funny?

The other important thing is that IBM remains committed to the Power processor even if it might be looking at farming out the way they get made. This is good news for IBM i and AIX customers alike. Linux customers have options, with Linux available on X86 servers, System z mainframes, and soon ARM servers.

“We are clearly committed to the Power platform,” explained Doug Balog, general manager of the Power Systems division at IBM, when I talked to him last week about Samsung joining the OpenPower Consortium and Google, building its own custom Power8 servers, which it is testing in its labs right now as I reported over at EnterpriseTech last week. “You can see that based on the OpenPower initiative and new membership but also based on our soon-to-be-coming Power8 chip and system designs. There is no confusion within IBM on the importance of the Power chip or the clients that use them.”