The other shoe finally dropped. IBM will transfer its semiconductor assets to Globalfoundries. I say “transfer” rather than “sell” because
Such an arrangement might look unusual at first glance, but IBM’s semi group has been losing a bundle of money over the past few years — $700 million last year and similar amounts previously — and shows no immediate prospects of going cash-flow positive.
As the cost of making chips has risen, IBM’s volume base has declined, making the business less and less viable each year. At this point, IBM is practically the only company using its own Power chips. In the past, it was able to sell them to game console manufacturers like
Last year, in an effort to keep Power architecture alive, IBM contributed it to a consortium called the OpenPOWER Foundation, members of which are developing their own Power chips, mostly for server applications. Founding members include Google, Tyan, nVidia, and Mellanox. But the foundation’s establishment did nothing to change the basic economics of IBM’s chip manufacturing business, which continued to hemorrhage money.
Thus, the odd terms of the deal, which has IBM paying Globalfoundries, like some garbage hauler, to take the material away. The transaction includes a non-recurring pre-tax charge of $4.7 billion, or $3.3 billion net of tax for IBM’s third quarter earnings, a charge that pretty much wiped out the company’s earnings for the quarter. For the company has a whole, EPS, reported this morning, was 2 cents, compared with $3.68 in 3Q13.
However, IBM’s lemons may well be Globalfoundries’ lemonade, as the one key element that IBM has been unable to produce Globalfoundries has in spades: volume. If Globalfoundries can get the semi operations up to a decent scale, it can make money with them.
And there are some real gems among the assets. IBM’s Application-specific integrated circuit (ASIC) specialty business and the radio-frequency (RF) parts that it sells to the phone industry are in high demand. On a call with analysts Monday morning, Globalfoundries' SVP of Technology and R&D Gregg Bartlett noted that IBM’s existing facilities in Essex Junction, Vermont, are insufficient to meet demand and that Globalfoundries will be “expanding” this technology to its other facilities in Germany and Singapore.
Senior Vice President of IBM Systems & Technology Group Tom Rosamilia emphasized on the same call that IBM remains committed to semiconductor R&D, having just pledged $3 billion to it in July. The company will maintain its R&D facility and partnerships in Albany, New York, and expects to lead process-node development from 22nm down through 14nm and 10nm over the next 10 years.
Also, IBM will become a major customer for Globalfoundries, buying Power chips from the manufacturing company for its own mainframe and “scale-out” systems as well as for next-generation storage systems.
This deal has been in the works for a long time. I've been writing about the increasingly desperate dynamics of the semiconductor industry for several years already and even fingered Globalfoundries as IBM’s likely partner this past July, but I was far from the only observer to put two and two together. This was one obvious deal to anyone watching the industry.
It is IBM’s business model to focus on high-value areas where it has a strong advantage. Semiconductor manufacturing just isn’t one of them. On the other hand, the company has major contributions yet to make in semiconductor research and in systems hardware. With the money-losing manufacturing business off its hands, it now has the freedom and scope to do so.
