Plotting IBM Micro’s Future

It’s been a wild ride for IBM’s Microelectronics Group.

Neither IBM, nor the other parties involved, have made any public comments about the recent events concerning IBM Micro. Much of the drama has played out in the media.

Based on those reports, here’s a rough outline of the events. Not long ago, IBM put its loss-ridden chip unit on the block to shore up the company’s bottom line. Then, IBM entertained several offers, but GlobalFoundries apparently emerged as the leading candidate to buy the chip unit.

GlobalFoundries wanted IBM’s engineers and IP, but not the fabs. IBM supposedly offered $1 billion to persuade Globalfoundries to buy the whole chip unit, according to Bloomberg.Globalfoundries wanted about $2 billion.

IBM, according to reports, balked and the deal fell through. Perhaps the U.S. Department of Defense (DoD) also stepped in to block the deal. IBM Micro apparently does some sensitive work with the government.

Then, after rejecting GlobalFoundries’ offer–and realizing that it may have lost some credibility in the semiconductor market–IBM recently put out a rather curious press release. In an effort to re-assert itself in the chip business, IBM in July announced it is investing $3 billion over the next five years in two chip-related research programs. The first research program is aimed at the 7nm CMOS node. The second is to explore the various technologies in the post-CMOS era.

Wasn’t IBM already doing that? Now, it’s business as usual for IBM Micro, right? Not so fast. Following the ill-fated sale of IBM Micro, Big Blue must still make some tough decisions about the chip unit. IBM can’t afford to sustain ongoing losses in that unit. And besides, IBM itself is focusing on software and services—and not hardware and chips.

So where does IBM Micro go from here? Here are some of my ideas for Big Blue:

Revamp its fab and chip strategy. For years, IBM justified its existence in the semiconductor business on two fronts. First, it developed and made chips for its own computer lines. Second, it had a decent merchant chip business, namely in ASICs, game machine processors and RF.

But in recent times, IBM lost the game-machine processor business to AMD and others. Making chips for its own systems is critical, but the volumes are miniscule. So today, IBM neither has the volumes, nor the customer base, to fill its 300mm fab in East Fishkill, N.Y. In addition, the fab is in dire need of some serious upgrades.

The solution: Over time, IBM may need to bite the bullet and jettison the East Fishkill fab. It’s simply a drain on the bottom line. Another possible solution is to downsize the plant and create an R&D fab. It’s a tough thing to swallow, but the company should also let the foundries make the chips for its own computers. However, IBM should keep its 200mm fab in Burlington, Vt., which makes RF chips. In RF, IBM can differentiate itself in the market.

Jumpstart the FD-SOI roadmap. IBM and STMicroelectronics are the main proponents of fully-depleted silicon-on-insulator (FD-SOI) technology. Of course, Samsung recently jumped into the SOI market, saying it will provide foundry services for 28nm FD-SOI technology. But still, when IBM put its chip unit on the block, the move raised questions about the long term roadmap for FD-SOI. After all, IBM itself took the lead role in developing SOI-based finFETs for 7nm and beyond. All told, FD-SOI technology seemed to get lost in the shuffle.

The solution: FD-SOI is a promising technology. Long term, SOI-based finFETs appear to be one of the viable options in the next-generation transistor race. So IBM must re-assert itself in FD-SOI and lead the industry in SOI-based finFET R&D. IBM doesn’t require a 300mm production fab to do basic research. Samsung, for one, is more than capable of taking the FD-SOI production ball and running with it.

Re-create Bell Labs. Once upon a time, IBM was a force in the merchant chip market. Those glory days are long gone. However, IBM is still a major force in R&D. In technology, IBM Research is doing impressive work in 3D packaging, carbon nanotubes, DSA, FD-SOI, graphene and spintronics. That’s just the tip of the iceberg.

The solution: Fold IBM Micro into IBM Research. Re-create the equivalent of a Bell Labs in the United States. And by the way, license the IP and make some money.

To some degree, IBM is already doing that. But frankly, the United States is in dire need of the equivalent of an Imec. Belgium-based Imec has assembled the who’s who in the semiconductor industry and has created an R&D powerhouse. In many ways, Imec is shepherding the next-generation chip roadmap. Imec is also doing impressive work in batteries, life sciences, solar and other technologies.

Of course, Albany Nanotech, Intel, Sematech, SRC and U.S. universities are all doing great work. But in many ways, the R&D efforts in the United States lack coordination and the funding is somewhat uneven. The European Union is not perfect, but the EU seems to be much better at coordinating their pre-competitive R&D efforts.

Answer the phone.IBM Micro is still in play. If a company comes up with a good offer for IBM Micro, Big Blue should pick up the phone and listen to offers.

The solution: IBM Micro could remain a viable part of IBM. In the R&D field alone, the industry is need of new breakthroughs in several areas, such as 3D packaging, fab tools, lithography, materials and others. But all told, IBM Micro must re-focus, if not downsize, its efforts. Changing with the times is always a difficult task.