IBM is exploring the sale of its chip division -- a move that could have profound implications for the chip and server industries.
SAN JOSE, Calif. -- When IBM sold its x86 server business to Lenovo last month, I predicted Big Blue might get out of hardware altogether by 2020. Looks like I may have been too conservative -- a deal could come as early as this year.
IBM has retained Goldman Sachs to explore possible sale of its chip division, the Financial Times reported Thursday. The company is not committed to a sale and could seek a partnership for the group, the report said. IBM declined to comment on the report, but the long-rumored move is not surprising.
Nevertheless it would have implications that ripple throughout the semiconductor and electronics industry. Without IBM's leadership, the Common Platform alliance among it, Globalfoundries and Samsung could disintegrate. The partnership, which licensed its 20-nm FinFET technology to Taiwan's UMC, provides a credible alternative to foundry giant TSMC as well as Intel which is increasing its foundry business.
GloFo and Samsung are already the next largest foundries following TSMC. However, they might lose momentum if they could not share R&D expenses, which are rapidly rising for chip makers. Last year, the trio announced it hammered out techniques to handle the need for double-patterning lithography, one of the chief bugaboos for chips starting at the 20nm node.
IBM has been one of the leading advocates of silicon-on-insulator technology, an alternative to the CMOS approaches of Intel and TSMC. Without Big Blue's backing SOI's promise of a more energy efficient process may fall by the wayside. New York State's initiative to become a global center of chip manufacturing excellence could lose a significant partner without an IBM chip division. GloFo is already ramping a big presence in the state, but the area's momentum could suffer if Big Blue's fabs go dark.
IBM contributed through-silicon via technology to Micron's hybrid memory cube, a 3-D stack of DRAM chips seen as one of the leading efforts in next generation memories. Micron's HMC will go on and IBM likely will still be a major user of the technology, even if IBM gets out of chips. However, an IBM exit from semiconductors begs the question of what future chip technology might go undiscovered once its researchers refocus on other areas such as data analytics.
The move could have plenty of ripple effects on systems, too. IBM's high-end computers depend on its proprietary mainframe and Power processors. They drive supercomputers in national labs and major banks around the globe. IBM would no doubt have whoever buys the group continue to make the chips. But a separate company might not have the incentive and insight to drive the silicon as fast or far as IBM's own engineers, opening opportunities in high-end servers for chips from Intel and Oracle.
Next page: a viable rationale