I believe CMP=Chemical-Mechanical Planarization (though I've also heard ...Polishing). Some people think of CMP as a publishing firm, but that might be associated with UBM, not IBM.
There's no doubt that IBM has played a significant role in the development of semiconductors. I always looked forward to briefings from their research team. But they never were a big merchant chip producer; didn't have the infrastructure for mass market sales and were primarily there to serve their own systems needs (the M stands for Machines, not Microprocessors). However, there are a lot of other names that created the semiconductor industry - names that are more in the dust than in purified sand (with some allowances for major buy-outs): AT&T (Bell Labs), Motorola (my upbringing), Fairchild, National, RCA, DEC (...Equipment...) - and those are just the American brands. IBM also suffers from being geographically away from the mainstream of our industry (in spite of what the state of New York likes to advertise), so there wasn't a lot of circulation of the talent outside of the blue walls.
I did a lot of designs using IBM PowerPC embedded processors for data communications products. They were wonderful chips, and the documentation was excellent. One of my favorite designs used an IBM 401GF, which was a PowerPC in an 80-pin TQFP with multiplexed address/data. It was a great way to talk to a Galileo (now Marvell) chip that had a multiplexed address/data, and the 401GF worked very nicely with synchronous SRAM. Most recently I designed with the PPC405EP, a very nice SoC with built-in SDRAM interface, Ethernet, and PCI.
IBM sold the embedded PPC technology to AMCC (now Applied Micro), who IMO let the technology languish (sigh). IBM was very open with embedded PPC and even gave away data books on real paper. The last time I looked at the AMCC site I couldn't even see what embedded PPC chips were available, much less get technical reference manuals. Great way to compete, guys.
I'm not sure why Applied Micro is even considered a "primary source" for PPC anymore, try and buy some actual parts through distribution and practically all you can actually get your hands on is from Freescale, although all I've really tried to get is MPC5xx or 55xx, maybe neither of them makes the full range nowadays, but Freescale seems to at least have pdf's online.
@Rick, thanks for the post. I still dont understand why IBM will get out of semiconductors because as mentioned in article IBM is still a chip technology leader, publishing top papers at events such as ISSCC, IEDM. FinFET may be the future of VLSI and hence IBM should stick to its R&D instead of coming out of semiconductor.
It's alright being an innovator but at the end of the qarter, shareholders need the highest possible return on investment. Fabs cost a lot of money and if you can't keep them working with commercial orders, they are a real drain on resources. I agree with the article, GF is the partner of choice in this reagrd, IBM should concentrate on design and application services...