I have not heard anything in a long time on the Power Processor/Server teams.
Are they growing or losing people or what ?
7 replies (most recent on top)
The Power11 is coming in 2025 with new AI capabilities:
https://newsroom.ibm.com/blog-ibm-power-modernizes-infrastructure-and-accelerates-innovation-with-ai-in-the-year-ahead
https://community.ibm.com/community/user/power/blogs/val-besong/2024/11/12/ibm-power-modernizes-infrastructure-and-accelerate
IBM has been taking out cost for so long that their value add in the world is now in question. Does IBM really think that they'll just be a lead chip designer while everything else in the POWER world is farmed out to others? OpenPOWER now manages the ISA, Samsung fabs the chips, AIX is being replaced by Linux, and IBM is getting out of what remains in HW and SW.
This answers the OP's question...IBM took an $8B business and turned it into maybe a $500M business, and that's being generous. IT executives all over the place are probably asking the same question the OP has when considering where to spend their money: "I have not heard anything in a long time on POWER..." They haven't heard anything because IBM no longer DOES anything.
Yeah, a real confidence booster. IBM certainly got their cost take out.
It’s all about IBM’s ability to design chips. They are very very good at it, and have some of the best toolsets in the world. As such it makes far more sense to farm the IP, via selling the designs and exit all other parts of the business via partnerships. NOTE the partnerships are a give to get. IBM exits most phases of the Power HW business and most likely the HW storage business, as their design captures the legacy SW (think OS’s) and peripheral Storage business. The partner manufacturers benefit as they don’t have to invest much in design as IBM has done the heavy lifting. Essentially everything IBM moved to Mexico in 2024 will be partnered out (most likely Foxconn) as IBM exits the manufacturing game (except for mainframe). It’s a win for IBM as their primary goal is cost take out.
For many (probably most) workloads, it is much easier to say "just throw some more Intel/AMD cores at the problem and it'll work out fine". Whether it's a cloud offering like AWS, or something done in-house, you can rest assured that your favorite app will run fine on Intel/AMD, and it will be supported for a long while because that's what the rest of the world uses.
IBM is no longer the big fish in the sea. It is a specialty provider, and so it has to be extra-special to get anyone's attention these days. Selling POWER may work, but it didn't work very long when Sun did it with SPARC. I see no reason why IBM will get anything out of it except maybe a (very) few years of license revenue.
Power is suffering the consequences of a failed Intel abandonment strategy. When IBM exited Intel, Power was supposed to back fill via its Linux on Power offerings and new cloud power offerings offerings. Problem was Linux on Power wasn’t and still isn’t ready for prime time, and Power cloud offerings are an also ran. As such almost every small and medium ISV said “no thanks” and adopted Intel Linux solutions instead or just redeployed on one of the major Intel cloud provider offerings. Heck of a job Power exec’s You took an 8 billion a year business and shrank it to 500 million. Major Power ISV’s (think enterprise players like SAP) did take the LOP plunge as they couldn’t afford to abandon their large customers. The results of this Power strategy has been a shrinking of the scaleout customer sets, (yep they have and continue to move to Intel), and a stagnation of the enterprise power customers (eg no growth but rather a replacement environment) There is a reason P11 is going to be more of a RAS offering and less of a bang for your buck offering. The enterprise power customers are demanding “mainframe reliability” in a Power package. It’s just a matter of time before IBM exits the HW and SW side of the power house (think partnership) and just farms the IP design side of the Power house
unfortunately not good...
PowerPC Future patches https://gcc.gnu.org/pipermail/gcc-patches/2023-October/633510.html