Recently AK appeared at Intel's CEO new strategy event. A collaboration was declared to include semi development, and Power and Z processors mentioned. Details are blured but given AK exiting HW, looks like a similar exercise to HP with Itanium at the time. Seems no longer IBM can afford the high costs of development as Power has failed to replace x86 and Z alone cannot keep aupporting apower processors development. But Intel could do it in return for IBM helping tyem with their failing fabs, and provide I0 and know how.
I'm speculating but if so, soon Systems will be sold to Intel in return for a 10 years processors contract.
Win win?
14 replies (most recent on top)
3zqy Why did you pick TCS? (I’m just curious why them vs someone else) A second question is why did you include Z into the equation? IBM has lot of reputation built into Z and that lift vs Power or storage is pretty tall
TCS will buy power and Z business?
https://www.indiatimes.com/technology/news/ibm-power10-akhtar-ali-rahul-rao-india-manufacturing-531660.html
Intel's BoD fired two CEOs in 4 years, recognized the end of an old business model and the geopolitical tailwind, made a $20B bet to right the ship, and brought in a true technical leader as the new CEO. You really have to give them the credit. This is in contrast to the dysfunctional BoD of the company, which sat on their hands for 7 years to let Rometty and her minions destroy the company piece by piece.
No, Intel will not buy Power and Z. Why bother to burden themselves with two outdated and obsolete legacy HW businesses. Their future is to get business from google, microsoft, and AWS, all are busy working on their own microprocessor designs. Each of them will need millions of units every year to fill up their cavernous data centers. The world has moved on, ibm's legacy HW business is only attractive to bottom feeders. Look at who brought Lotus notes.
You are absolutely correct. The problem in the US and with US companies is that it is all about the greed of executives... Intel and other big tech companies are no different than IBM. Most of these IBM companies are slowly dying...
You realize Intel is 3-5 years behind the competition. They are still making 10NM technology, and have fallen WAY behind the Chinese and Koreans. IBM has partnered with Samsung who is sampling 7NM technology for Power 10. It doesn’t come to market till end of 2021 - early 2022. Building fabs and bringing them up to speed is a 3-5 year effort. GF gave up on their efforts as it was too expensive, and it came to market way too slow. Intel is in the same boat as GF in that they are trying to hop a generation. It can be done, BUT it’s not cheap or easy. My guess is that’s why they are partnering with IBM. IBM has the basic knowledge of how to do it (If they have not laid those folks off), but it’s a long term effort. IBM labs are some of the best in the world, BUT remember, first die sets to come off a new Intel/IBM process are 3-5 years out on the time horizon
If they have a plan they need to announce it soon, power is bleeding talent at a fast rate and many of the older workers left behind plan are planning on leaving after p10, they will need to rebuild but the unknown is k–ling them, if there is a plan get it out now before it’s not repairable
Power will never grow again but if they need to keep it around for existing customers they need to stabilize it
@1dih @1hkc so the fab shouldn't have been sold right? It was critical for Z... Apple has no fab but is doing OK. Z is not going anywhere, tead the design services provided by Intel, they cannot copy the Z architecture, nor the ISA of the chips. It could be a deal to assure Z supply. ...
Thebonly reason Z is around mainly is the software and relability of operation. A lot of this could move to x8 even ARM with added security and reliability that IBM can sell its know how to Intel. Hardware itself is dead. HP moved its Itanium software to x86 for enterprises.
All speculation here but to think IBM will join Intel's fab was impossible nor to think they will make Power processors at Samsung orveven at Global Foundries. ..things change.
AK has already recognized that Z is important part of strategy in his bullet points so he's definitely not sharing the business or anything with intel.
Answer this if you think otherwise:
Who are the customers that are going to use IBM's hybrid cloud? Hint: those are the same customers who rely heavily on mainframes (Z).
I do think some of what being said is right.
There was some overlap with design teams in power and Z and i think they're probably cutting of power development and so those people would develop ip for z and x86.
Ibm would be dead without Z
IBM is turning over control of critical aspects of Power and Z to be managed by a competitor who is ruthless about crushing competition to x86? Brilliant!
@1ksi Intel can make the processor chips only. Dell can make the servers. IBM keeps the high margins Z and Power software. Not sure it's worth for Dell to buy thevSystems but can package the Intel's made processors into servers. IBM makes no PCBs or metal racks for Z maunframe servers even anyway. But costs of keeping the engineering team to only develop Power and Z circuits it's high. At Intel they could consolidate circuits into main x86 too. Why develop multiple floating point units for example, or memory controller or PCIe Gen 5 if such IP could be used and mux and match accross different ISA..just like Zvand Power share circuits, now Intel could share such IP accross i86 too. Win win. Also for IBMers working on Power and Z processors moving to Intel I think it's good for all parties if it happens. Intel needs help too as AMD threatens them with TSMC. HP gave Intel their Enterprise, like Power, Itanium design, keeping the software and marketing, vservices. Looks possible to make an agreement. That's why AK bothered to show up at Intel's CEO new strategy event. Should be fundamental and AK gets rid of the hardware. Storage could go to Dell indeed. But Intel has storage Nandflash too.
There is a pile of IP in the Power and Z processors. IBM has contracted with Samsung to manufacture the Power chip, but I’m not sure they contracted to share IP. I can’t see Intel signing up to jump into the server manufacturing business, but I could see IBM joining forces with Intel to optimize the IP stack. Dell/EMC has a much better shot at partnering with IBM, because they have an established channel and would like to gain storage share along with having a high performance AI offering (eg Power). Same with HPE and they need an Itanium replacement which goes off of support in July. We’ll see, but IBM definitely wants to dump the scale out HW and one of these folks is going to get a “deal”
@ @1qxf OP here - only the design of Z and Power chips not the software stack certainty not the service and enterprise selling - that is to stay within IBM as part of cloud. No more "Systems" division. Just the hardware diversified no way the software, applications etc. That's what HP has done with its Itanium. IBM doesn’t care for ARM to replace z or Power. ..why?
Intel's new design service optimized on their fabs.
Remember that the US worries about fabs out of the country, that includes Samsung, especially for Government use, with Power and Z. Both companies can tap the Government$billions available to build US fabs. Samsung may still have access to past tech but how does it look for big Intel to need IBM? Bad in my view unless they will buy them including a big portfolio of patents. .. yes, all hardware Quantum, AI chips whatever. IBM left with applications on top of the hardware. Remember even DB2 can run on x86... maybe crippled intentionally but in principle can work. Intel could use IBM's chops security too...
This is speculation but IBMers moving to Intel will be better for them than now, expanding their opportunities to design x86 circuits too.
Here's Intel's design services - need all be made in the USA especially as the Gov is an important customer I think.
https://www.intel.com/content/www/us/en/foundry/intel-foundry-services.html
All these services are new. Not just a fab services but also actual design and chips on Intel's fabs. Will take a few years I think but in my view Samsung will lose future Power and Z. Not a big revenues loss . How many wafers are needed? Not many given the handful chips sold in Z and Power systems anyway.
The IBM press release (cleverly?) mentions Power and Z as if to imply they are involved in this partnership, but they are not. (Samsung is IBM's partner for Power manufacturing).
If you read details, this is about x86, GPUs, potentially ARM, and even RISC-V – the latter two being much more attractive alternatives to Power for anybody serious about a non-x86 direction.
Even if Intel did buy Systems (and it won't), it would be a catastrophe. Every previous time Intel has dipped its toes into some market beyond chips, it has lasted only as long as Intel believed that the business was driving sales of more chips (see, for example, it's brief and unhappy ownership of Lustre. I firmly expect DAOS to go the same way when Intel realizes it isn't driving Optane sales.) What is there in the Systems portfolio that would survive that kind of scrutiny from Intel? Certainly not of software - Intel has proven over and over it can't do software businesses; certainly not Power - a niche market, and currently under contract to Samsung; not Z - that's an enterprise sale not a hardware speeds'n'feeds sale. Maybe one or two things in Storage, but not remotely enough to justify buying whole thing.
https://newsroom.ibm.com/Intel-Teams-with-IBM-for-Advanced-Semiconductor-Research-and-Development