IBM bids for Sun Microsystems, report says
- Thread starter Idontcare
- Start date
theeedude
Lifer
- Feb 5, 2006
- 35,787
- 6,198
- 126
Aside from patent portfolio. talent, and installed base, it eliminates a major competitor to a lot of their products, which will help pricing power.
Plus Sun has $3B in cash, so really, they are only paying $3.5B for it. That's less than Sun paid for Storagetek.
I believe Sun lost a lot of its market value during this downturn, didn't they lay off some thousands of people already? Perhaps it's not a bad time to buy companies for those who have the cash.
IBM's purchase of Sun (assuming it is real) has got to be about acquiring enterprise customers, followed by getting valuable software IP.
Sun's dying SPARC business has nothing to offer IBM, whose mainframe and POWER (AIX / System I (AS/400)) businesses are doing just fine on their own. Since POWER 4 was introduced, IBM has been steadily acquiring UNIX market share at Sun's expense, and there is nothing to indicate that this was going to change in the next few years. Sure, Niagara is interesting, but I'm not sure the latest Niagara systems are any better at high throughput web/app serving than the arriving Nehalem Xeons. And that is all Niagara is good for, unlike Nehalem. As for Rock, that seems to be a looooong shot. Sun could not bring the in-house developed successors to UltraSparc III/IV to market (they had two botched attempts), and have ended up selling re-badged Fujitsu servers at the high end. Hell, Sun did not even create the initial Niagara design, they bought it from Afara Web Systems. Sun has no credibility left when it comes to their in-house RISC design efforts. Until Sun shows real Rock systems running real workloads their customers care about, it is just a bunch of dream.
IBM's storage business could make use of StorageTek, which is (was?) the leader in mainframe tape backup. But, if that were so important to IBM, why did they not by StorageTek years ago, before Sun let it atrophy?
I'm sure IBM thinks that Sun's software portfolio is valuable, though there is a real problem with how they would make money off of it when Sun is struggling to do so. IBM has done a much better job profiting from Java than Sun ever did, and IBM has a real multi-faceted, multi-billion dollar enterprise software business. IBM already has a successful enterprise DB - do they need MySQL? IBM already has a successful Java middle-ware stack, do they need Glassfish? And what the hell do they do with Solaris? Solaris is still popular, widespread, and highly regarded (you can't say all of those things about AIX), but I can't imagine IBM going with Solaris over AIX. I doubt IBM's current customers would take kindly to any kind of forced migration to Solaris.
I guess all that's left is acquiring Sun's customers?
IBM could use the QuickTransit software they have from the purchase of Transitive in 2008 to allow Solaris customers to run their apps on AIX/POWER without modification. It's not a perfect solution, but it could act as a bridge until most important apps are available natively on POWER, just as Apple licensed QuickTransit when they switched to x86.
Then IBM could cherry pick what parts of Sun's software portfolio they want to use, like ZFS and the QFS and Luster HPC filesystems, and much more, without taking on the task of keeping Solaris viable.
Sun's dying SPARC business has nothing to offer IBM, whose mainframe and POWER (AIX / System I (AS/400)) businesses are doing just fine on their own. Since POWER 4 was introduced, IBM has been steadily acquiring UNIX market share at Sun's expense, and there is nothing to indicate that this was going to change in the next few years. Sure, Niagara is interesting, but I'm not sure the latest Niagara systems are any better at high throughput web/app serving than the arriving Nehalem Xeons. And that is all Niagara is good for, unlike Nehalem. As for Rock, that seems to be a looooong shot. Sun could not bring the in-house developed successors to UltraSparc III/IV to market (they had two botched attempts), and have ended up selling re-badged Fujitsu servers at the high end. Hell, Sun did not even create the initial Niagara design, they bought it from Afara Web Systems. Sun has no credibility left when it comes to their in-house RISC design efforts. Until Sun shows real Rock systems running real workloads their customers care about, it is just a bunch of dream.
IBM's storage business could make use of StorageTek, which is (was?) the leader in mainframe tape backup. But, if that were so important to IBM, why did they not by StorageTek years ago, before Sun let it atrophy?
I'm sure IBM thinks that Sun's software portfolio is valuable, though there is a real problem with how they would make money off of it when Sun is struggling to do so. IBM has done a much better job profiting from Java than Sun ever did, and IBM has a real multi-faceted, multi-billion dollar enterprise software business. IBM already has a successful enterprise DB - do they need MySQL? IBM already has a successful Java middle-ware stack, do they need Glassfish? And what the hell do they do with Solaris? Solaris is still popular, widespread, and highly regarded (you can't say all of those things about AIX), but I can't imagine IBM going with Solaris over AIX. I doubt IBM's current customers would take kindly to any kind of forced migration to Solaris.
I guess all that's left is acquiring Sun's customers?
IBM could use the QuickTransit software they have from the purchase of Transitive in 2008 to allow Solaris customers to run their apps on AIX/POWER without modification. It's not a perfect solution, but it could act as a bridge until most important apps are available natively on POWER, just as Apple licensed QuickTransit when they switched to x86.
Then IBM could cherry pick what parts of Sun's software portfolio they want to use, like ZFS and the QFS and Luster HPC filesystems, and much more, without taking on the task of keeping Solaris viable.
Soulkeeper
Diamond Member
- Nov 23, 2001
- 6,740
- 156
- 106
sun microsystems apparently has 10% of the server market ....
not to mention java and all their services/software engineers
they are kind like a mini-ibm
not to mention java and all their services/software engineers
they are kind like a mini-ibm
Originally posted by: Soulkeeper
sun microsystems apparently has 10% of the server market ....
not to mention java and all their services/software engineers
they are kind like a mini-ibm
More specifically - a fabless mini-IBM.
I long thought SUN and AMD would combine, it always made sense to me from the Intel market segmentation approach with x86 vs. EPIC. Good parallel could have been made by an AMD/SUN merger with x86 vs. Sparc (niagara/rock).
But an IBM/SUN merger makes sense too considering the threat IBM is facing with Intel skipping 45nm and going straight to 32nm with Itanium. That kind of aggression in the big-iron marketspace sets the stage for a hastening market share demise for the next decade to come.
Not too mention SUN is screwed at 45nm, TI left them in a total lurch and they literally have no viable solutions for 45nm high-performance MPU production options. A buyout by IBM would give the SUN team a built-in excuse for not having anything home-grown (non-Fujitsu) in the market for the 45nm cycle.
I think that Sun and AMD might have made some real magic a few years ago when both were stronger, richer, and more relevant. But back then, Sun would have had to pay a premium for AMD, and it still (probably) would not have prevented the problem AMD has had for the last couple of years with a relatively weak lineup compared to Intel. Maybe a Sun/AMD merger would have ended up like AMD/ATI today - lots of talent, some good products and great ideas, but no money.
It does seem that the high end server CPU race is coming down to IBM and Intel. Will Intel really launch a 32nm Itanium design in a timely manner to "jump" past IBM and Power 7? You would certainly think that they could, but Intel has for what ever reasons been chronically late with Itanium releases (Tukwilla at 65nm slipping another 6-9 months is getting ridiculous). Many folks say that Itanium is a profitable business for Intel these days, but is Intel just making the minimum effort and milking the last drop of money out of Itanium, or are they really serious about kicking IBM out of first place in big iron?
While Sun can't be happy about having to look for another fab now that TI left the business, at least they have options. Didn't TI go with TSMC for their 45nm fab business? I know TSMC does not target their foundry processes for high-end CPUs, but neither did TI. I guess Sun could also go to Fujitsu (I have no idea if this is technically feasible regarding Fujitu's process technology). Heck, I thought Fujitsu would buy Sun outright. Sun's guys could work on Niagara, Fujitsu's guys could work on Sparc64 for the big-iron/mainframs, they would save a ton on eliminating duplicate efforts, their x86 server business would have greater economies of scale, they would have a stronger presence in the US and Asia, etc. etc. etc.
It does seem that the high end server CPU race is coming down to IBM and Intel. Will Intel really launch a 32nm Itanium design in a timely manner to "jump" past IBM and Power 7? You would certainly think that they could, but Intel has for what ever reasons been chronically late with Itanium releases (Tukwilla at 65nm slipping another 6-9 months is getting ridiculous). Many folks say that Itanium is a profitable business for Intel these days, but is Intel just making the minimum effort and milking the last drop of money out of Itanium, or are they really serious about kicking IBM out of first place in big iron?
While Sun can't be happy about having to look for another fab now that TI left the business, at least they have options. Didn't TI go with TSMC for their 45nm fab business? I know TSMC does not target their foundry processes for high-end CPUs, but neither did TI. I guess Sun could also go to Fujitsu (I have no idea if this is technically feasible regarding Fujitu's process technology). Heck, I thought Fujitsu would buy Sun outright. Sun's guys could work on Niagara, Fujitsu's guys could work on Sparc64 for the big-iron/mainframs, they would save a ton on eliminating duplicate efforts, their x86 server business would have greater economies of scale, they would have a stronger presence in the US and Asia, etc. etc. etc.
They really don't, not for 45nm. At the time TI made their decision (Jan 2007) SUN was nearly 18 months into designing their 45nm chips for our high-performance C014.B process technology.
They had no idea, the day TI made their announcement internally, Hans Stork (CTO) flew to CA after the meeting to disclose it to SUN CEO same day.
The "transition" plan for SUN was for TSMC to step up to the plate and rush-develop a high-performance MPU sub-node for 45nm special for SUN. To assist in TSMC's development timeline and minimize the slips and delays to SUN we setup a core team tasked with this effort.
Then TSMC said take a walk. So TI partnered with UMC for their own internal 45nm development needs. SUN has been drifting ever since (early 2008).
TSMC does not. TI did. We developed four sub-nodes for every major node, each sub-node had different power/performance characteristics and varying maximum metal level schemes.
The dot-B's and dot-C's were always for SUN, exclusively for SUN. Additionally we developed mobile flows (dot-0, later dot-k, and ultimately named dot-m in the final nodes we developed) as well as high-performance DSP and embedded CMOS (dot-A).
A lot of process technology got recycled naturally, dot-C high performance stuff from 90nm became the template for low-performance mobile xtors for 65nm, etc.
It is feasible, SUN could go with any process, the challenge is that they need about 4 yrs advance notice as to the process technology specifics that they need to design their chips for. No two 45nm flows are the same, they are all unique and different enough that porting one's design from one 45nm node to another is simply impractical.
It is a good question what this will do for Fujitsu. They are using open-sparc, so they suffer no ill effects of an IBM/SUN merger, but if SUN needs to merge to survive then presumably so too must Fujitsu or the same fate awaits them as only a matter of time.
Wow - thanks for this informative and insightful post, especially regarding TI's former fab business!
agreed, that along is worth paying. although not sure how much of it is patented. ibm might as well just dl the source code from sun's site.
I find myself in agreement with Paul DeMone's assessment regarding the "bury some inconvenient technology" rather than let someone snag it and turn it into a potential problem down the road.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter