Power Systems is shrinking at double -digit rate for several years. Perhaps it would be very strange if IBM keeps pouring money into the next Power chip.
What do you think would happen to Power Systems and its staff ? Would some player like SAP buy it ? Would IBM slash more staff from it to make it leaner for a merger ?
13 replies (most recent on top)
The biggest reason Power sales were in decline back at the beginning of 2021 was that the Power9s were old and everyone was waiting for the Power10 (originally announced for 2020) - that is not to say higher commission seeking IBM salesmen are not their own worst enemy. Even as late as 2019 IDC showed Power market share at around 1/8 th of the world server market share (though IBM had a smaller share of that than Inspurpower). I hope the Power stuff keeps going - but they will really need to support OPEL in a two socket server so that the openpower guys can support stuff back to the hardware level and get around the expense and any PowerVM hypervisor nuances for the Linux side of the fence (like raptorcs did with power9).
I really hope the world never cedes computing platform choice to a proprietary x86 monoculture just through laziness or the lack of hardware proficiency of IT managers. I remember x86 being a mediocre compromise when it was introduced, and I'm not convinced it has changed much from that position - people have just slowly found ways to work around the shortcomings to make it look more like a large enterprise class platform (kudos to them). There was a lot more choice of architecture when I first started using computers (Motorola, DEC Alpha, MIPS, Sparc etc) and each had strengths and weaknesses. The Alpha and MIPS machines were my favourites. Unix was originally envisaged to be platform agnostic - the fact that most developers only have access to x86 platforms should never mean they should stop developing code that is portable across hardware platforms - that is simply lazy, poor development practice and short sighted. I think the higher transistor count x86 has come a lot closer to a Moore's law impasse than Power - Intel even ironically turning to IBM (and TSMC) to help shrink the CPU die size further.
The only thing I would like to see totally supplant Power is a plethora of Quantum computing offerings from multiple vendors - of course this would also mean the same would supplant x86 to the despair of many die hard x86 admirers who have never been exposed usefully to anything else.
Power this is your future Partnerships, Redhat, and Hybrid cloud thru the ISV. NOTE traditional legacy business operations are being phased out
https://www.crn.com/news/managed-services/ibm-ceo-arvind-krishna-appoints-new-ecosystems-chief-to-accelerate-hybrid-cloud-ai-sales-offensive
also remember that what makes IBM money it is all the Legacy stuff... I know some people don't want to believe it, but that is where the money is CURRENTLY for IBM as anything IBM Cloud is not working at all!
But the legacy stuff - Z, P, Storage has quarterly double-digit revenue declines, so the margin/profit declines incrementally too! This has been the trend for years. So even though Z is cyclical, the numbers continue to get smaller. Eventually Legacy will die. If IBM keeps the legacy hobby for the declining revenue, then ok - it needs to manage (downsize) those products accordingly, but also build substantial growth elsewhere in order to keep the company alive. But then isn't that exactly what the CEO is striving to do.
The problem with Power is not only that it's a small and shrinking business - IBM knows how to manage a business like that to a "soft landing", although it does make the investment required to do a meaningful P11 unlikely.
The bigger problem with Power is that it is a drag on every software business in IBM and Red Hat. Not only are we pressured/forced by mgmt to port offerings to Power even though the dev/test/release cost is high and the incremental revenue is small to none, but we are also often forced to delay releases for the sake of simultaneous availability on all three platforms (x, p, z).
Don't even mention the fact that there is an entire team whose "job" is to get Power into the big public cloud providers. Seriously.
The hidden "Power tax" on the rest of IBM does not get talked about enough IMO
Notice in the earnings report and analysis meeting that Power was never mentioned. Notice that in Arvin's email to the employees, Power was not on the priority list. Notice how poorly Power performed. Huge drop on year to year compare and the reason that is extra bad is that the previous year was considered very weak which led to a major layoff in the summer/fall of 2020.. Something is up based on management reaction to questions after earnings. This is the year everyone in Power was worried about as far as layoffs because of projected poor sales and Power10 has nothing special vs Power9. Power100 sales will only go to companies that are looking for upgrades for AIX and I which are shrinking businesses. Good news is that if you are on Power and live in Austin there area a ton of tech jobs especially at AMD, Apple, AWS and Nvidia. I suggest jump early while those job posting are still there.
- also remember that what makes IBM money it is all the Legacy stuff... I know some people don't want to believe it, but that is where the money is CURRENTLY for IBM as anything IBM Cloud is not working at all!
Remember: Customers with IBM I (on Power, formerly known as AS/400) are the largest portion of enterprise customers that IBM surely doesn't want to loose...
Unfortunately i see no value in having Power in the current strategy. This is a hardware market that is shrinking. ISV ecosystem is not growing, open source didn’t make its way through. There is little value for IBM to invest more in it.
Let’s wait until after 1st quarter before we start guessing on powers future. P11 HLD was paused but that could be due to other reasons, if it starts up again then power should be around for a while longer but the team will shrink to match the revenue.
Power is alive only because systems top leadership can show they have 3 major HW under them.
Arvind should wake up and sell it (parts which could be sold) or will meet the same fate as oracle's hardware
They will continue to optimize it for AI workloads.
Oracle on power now we are smoking something and not sharing.
I believe Power has faded far quicker than IBM expected, BUT IBM still has quite a competitive edge with the power chip. Remember Power has become a component for Storage, Mainframe, and openpower in addition to the legacy Power systems implementations. That’s quite a lot of IP that IBM can still farm in addition to the legacy OS’s and middleware. So is Power 10 the end? Most likely not. Why? Because Samsung has invested a pile of money into manufacturing the chip sets going forward, and you don’t spend 100’s of millions to billions of dollars on something that is fading away. So what is IBM’s game plan. It most likely centers around “enterprise” implementations (5-10% of the installs) and their go forward strategies. IBM will give up the scaleout implementations (90-95% of the installs), the component / openpower market (farming IP instead) and the Intel compatible/Power LINUX marketplace (Power just wasn’t cost justified and ISV’s didn’t embrace it). Does this mean IBM exits Power? NOPE It just means IBM will farm the IP (mostly OS’s), exit the scaleout marketplace for a partnership with a manufacturer, and invest in Power and Power LINUX where performance is desired. NOTE SAP buying Power is most likely not going to happen Why Because Intel stroked a very very large check to SAP to be the infrastructure of choice for the SAP cloud. Would another ISV stoke a check for Power? Perhaps Oracle certainly has a vested interest in keeping legacy SW viable, and cutting a deal with IBM helps on both the Power and Mainframe front