GlobalFoundries Is The Front Runner To Buy IBM Chip Biz
April 7, 2014 Timothy Prickett Morgan
If the rumor mill is correct–and stories with concrete details that appear in the Wall Street Journal usually are–then it looks like IBM is one step closer to selling off its chip manufacturing business. And it looks like GlobalFoundries, the former chip-making arm of Advanced Micro Devices that is owned by the government of Abu Dhabi, is the front-runner to acquire the business.
According to a report in the WSJ last week, Intel and GlobalFoundries are in discussions with Big Blue to acquire its chip making unit, IBM Microelectronics, and Taiwan Semiconductor Manufacturing Corp, which is another obvious suitor for the unit, has dropped out of the discussions. None of these vendors will confirm that they were in talks and none of them will certainly admit that they have left the bargaining table, either.
IBM’s chip plant in Burlington, Vermont, is pretty old in terms of the process technology it uses, and therefore it is not worth all that much except for the skilled workforce and government permits it has to operate in the area. Not all chips need to be at the latest-greatest process technology, so there is some value even if the equipment is a bit dated. But, just like Moore’s Law makes chips that are three or more years old obsolete, Moore’s Law makes chip fabrication similarly obsolete. IBM’s chip plant in East Fishkill, New York, has the latest-greatest 22 nanometer technologies, which are as good as anything Intel can deliver today. But, rather than having 3D FinFET transistors, IBM is relying on the 22 nanometer process coupled with its copper and silicon-on-insulator (SOI) technologies to push the transistors. TSMC and GlobalFoundries are going to 3D transistors based on the FinFET approach, and so is Samsung Electronics, another potential buyer for the IBM Microelectronics biz. For all we know, IBM was planning to move to FinFET with its own 14 nanometer process node, but what we do know is that it has not done so yet. And that means whoever buys this business is probably going to have to spend billions of dollars upgrading the East Fishkill plant.
GlobalFoundries just spent billions of dollars to the north outside of Albany, New York, in the town of Malta in its Fab 8 chip plant is being upgraded to the 14 nanometer node, so it is hard to imagine why this company would want a second plant. It is far more likely that any of the potential buyers would take the plant if IBM put a substantial portion of its semiconductor patent portfolio into the deal–much like IBM had to goose the System x selloff deal with Lenovo with switches, storage, and systems software to get it done. IBM also accepted about half of the original asking price, too, to get the Lenovo deal done, and based on the rumors going around now as reported in the WSJ, it looks like the same thing is happening again. The scuttlebutt is that IBM wants more than $2 billion for the business, but the interested parties are will to part with something more on the order of $1 billion. IBM would probably have to take a multi-billion write-off on its books just to shut down the chip business, and it would also have no place from whence to source its Power and System z chips, either, since no one else uses the copper/SOI processes that IBM has perfected and used to great advantage.
It is a tricky business, selling off IBM Microelectronics, as you can see. But once IBM lost the game console businesses of Microsoft and Sony to AMD, the fate of IBM Microelectronics was pretty much sealed, no matter how much extra business IBM might try to attract as an ARM chip or Power embedded chip maker. If IBM didn’t buy something on the order of $14 billion in its own shares from Wall Street every year, it could easily indulge in chip manufacturing. Even after losing the game console business, IBM Microelectronics is still bringing in around $400 million in revenues per quarter, about 20 percent lower than it was at five years ago. The problem is that each node gets increasingly more costly to bring to market, and the move from 300 millimeter to 450 millimeter wafers is going to add further to development costs even if it will eventually drop the costs per chip.
To make it plain and simple: IBM made the choice to put financial engineering ahead of real chip engineering, and now it is suffering the consequences. Now, Big Blue needs to find a buyer who will continue making its Power and System z chips and invest in the future to advance process technology. There is no question in my mind that IBM will prevail in this effort, but I think that if you look at the math of the deal very closely, with a wafer supply agreement that is inevitable between the buyer and Big Blue, you will see that it will all net out to a wash and IBM may end up paying slightly higher costs for each chip. But, it won’t have what I presume is an unprofitable business dragging on its Systems and Technology Group business.
All we care about is that someone makes Power chips, and in fact, gets them to market a bit faster than IBM can currently do on its own. IBM will no doubt continue to design Power and System z chips and will no doubt continue to sell systems based on these processors, even if, as I suspect it will eventually do, the company outsources manufacturing to a third party. Again, so long as IBM warrants the devices, this doesn’t matter. Unless, of course, you still think of it as International Business Machines. This is just business. Don’t take it personally.
Personally, I have a hard time with that.