SoftBank grabs ARM
July 18, 2016 6:20 AM Subscribe
ARM Holdings, the British microprocessor design company whose designs are in almost all mobile devices,
is being acquired by Japanese conglomerate SoftBank for £24.3bn.
ARM emerged from Acorn Microsystems, the company behind the BBC Micro computer of the 1980s, and its initials originally stood for the Acorn RISC Machine, the innovative RISC microprocessor powering the Acorn Archimedes (a more powerful though ultimately commercially unsuccessful contemporary of the Amiga and Atari ST). ARM's core product is the ARM architecture, a power-efficient RISC architecture, which it licences to chip makers; ARM chips power almost all mobile phones and tablets, and smaller versions are used as microcontrollers in embedded devices. More recently, ARM has been targeting the server market, traditionally a stronghold of Intel, with some success; meanwhile, Intel announced sweeping job cuts in April.
ARM emerged from Acorn Microsystems, the company behind the BBC Micro computer of the 1980s, and its initials originally stood for the Acorn RISC Machine, the innovative RISC microprocessor powering the Acorn Archimedes (a more powerful though ultimately commercially unsuccessful contemporary of the Amiga and Atari ST). ARM's core product is the ARM architecture, a power-efficient RISC architecture, which it licences to chip makers; ARM chips power almost all mobile phones and tablets, and smaller versions are used as microcontrollers in embedded devices. More recently, ARM has been targeting the server market, traditionally a stronghold of Intel, with some success; meanwhile, Intel announced sweeping job cuts in April.
From the article:
"The fall in the value of the pound against the yen has made the deal about 12% cheaper for SoftBank than before the referendum result on 24 June and 30% cheaper in the past year."
posted by dowcrag at 6:29 AM on July 18, 2016 [6 favorites]
"The fall in the value of the pound against the yen has made the deal about 12% cheaper for SoftBank than before the referendum result on 24 June and 30% cheaper in the past year."
posted by dowcrag at 6:29 AM on July 18, 2016 [6 favorites]
They're going to double the number of jobs in the UK.
posted by Segundus at 6:37 AM on July 18, 2016 [4 favorites]
posted by Segundus at 6:37 AM on July 18, 2016 [4 favorites]
It's hard to see this as anything but pure Brexit collateral damage.
Possibly; though OTOH, ARM's shares (being traded both in London and on the NASDAQ) went up against the pound as it fell.
posted by acb at 6:39 AM on July 18, 2016 [2 favorites]
Possibly; though OTOH, ARM's shares (being traded both in London and on the NASDAQ) went up against the pound as it fell.
posted by acb at 6:39 AM on July 18, 2016 [2 favorites]
i thibk Olivetti were major stakeholders already, but it's hard to see this as anything other than the UK beginning to shut up shop.
posted by Artw at 6:44 AM on July 18, 2016 [1 favorite]
posted by Artw at 6:44 AM on July 18, 2016 [1 favorite]
If SoftBank do as they said (keeping ARM at arm's length, with headquarters in Cambridge, and funding an expansion), this could be promising. ARM have been starting to eat Intel's lunch in the server space, and this could accelerate. (I probably wouldn't want to have Intel shares now; they've tried three times to come up with successors to their x86 cash cow, and failed each time, and their attempts to take on ARM on their home turf by heavily subsidising anybody who wants to put a low-power x86 CPU in their tablet have been comical above all else.)
From what I hear, the main threat to ARM's position could be open-source instruction sets like RISC-V, which would let clients tailor the instruction set in a way they can't do with ARM (which makes sense for platforms where instruction-level binary compatibility is not a consideration but efficiency is).
posted by acb at 6:45 AM on July 18, 2016 [2 favorites]
From what I hear, the main threat to ARM's position could be open-source instruction sets like RISC-V, which would let clients tailor the instruction set in a way they can't do with ARM (which makes sense for platforms where instruction-level binary compatibility is not a consideration but efficiency is).
posted by acb at 6:45 AM on July 18, 2016 [2 favorites]
Another angle here is Japan's central bank has an unlimited amount of yen at hand to give to anyone it wants.
yen per dollar since the 70s
(I suspect Japan Inc is most happiest with the yen around 130)
Japan central bank's printing (as % GDP) vs the Fed:
https://fred.stlouisfed.org/graph/?g=5J1Z
Per-capita basis:
https://fred.stlouisfed.org/graph/?g=5J32
Shows the BOJ is printing ~$10,000/yr per working-age Japanese person now, while the Fed's peak printing was $6,000/yr back in 2010.
posted by Heywood Mogroot III at 6:52 AM on July 18, 2016 [2 favorites]
yen per dollar since the 70s
(I suspect Japan Inc is most happiest with the yen around 130)
Japan central bank's printing (as % GDP) vs the Fed:
https://fred.stlouisfed.org/graph/?g=5J1Z
Per-capita basis:
https://fred.stlouisfed.org/graph/?g=5J32
Shows the BOJ is printing ~$10,000/yr per working-age Japanese person now, while the Fed's peak printing was $6,000/yr back in 2010.
posted by Heywood Mogroot III at 6:52 AM on July 18, 2016 [2 favorites]
This seems like just such a sounder investment than LinkedIn.
posted by Navelgazer at 7:18 AM on July 18, 2016 [7 favorites]
posted by Navelgazer at 7:18 AM on July 18, 2016 [7 favorites]
Winterhill, What convinces you Brexit had nothing to do with it?
I'd guess that SoftBank has been considering a move like this for years. Who wouldn't consider buying ARM if the price was right? Then it drops %12 in price (or %30 depending on your time frame).
posted by bdc34 at 7:32 AM on July 18, 2016
I'd guess that SoftBank has been considering a move like this for years. Who wouldn't consider buying ARM if the price was right? Then it drops %12 in price (or %30 depending on your time frame).
posted by bdc34 at 7:32 AM on July 18, 2016
i thibk Olivetti were major stakeholders already, but it's hard to see this as anything other than the UK beginning to shut up shop.
No, not true.
Although Brexit-related issues could be said to be interpreted this way, foreign direct investment is required in the UK, particularly as there is a trade deficit. Macroecon, innit.
Also, there is no way that this was dreamed up in the last 2-3 weeks post Brexit. Brexit is just a nice cherry on top of the pricing, and assures the deal goes through.
Since the Pfizer and Cadbury takeover issues, the government will seek to implement new rules about commitments in writing on hq/jobs in the UK before final approval.
posted by C.A.S. at 7:33 AM on July 18, 2016 [2 favorites]
No, not true.
Although Brexit-related issues could be said to be interpreted this way, foreign direct investment is required in the UK, particularly as there is a trade deficit. Macroecon, innit.
Also, there is no way that this was dreamed up in the last 2-3 weeks post Brexit. Brexit is just a nice cherry on top of the pricing, and assures the deal goes through.
Since the Pfizer and Cadbury takeover issues, the government will seek to implement new rules about commitments in writing on hq/jobs in the UK before final approval.
posted by C.A.S. at 7:33 AM on July 18, 2016 [2 favorites]
Another angle here is Japan's central bank has an unlimited amount of yen at hand to give to anyone it wants.
But so does any other typical central bank.
Speaking at a very high level of generalization, printing yen weakens the currency and raises the price of imports.
posted by praemunire at 7:39 AM on July 18, 2016
But so does any other typical central bank.
Speaking at a very high level of generalization, printing yen weakens the currency and raises the price of imports.
posted by praemunire at 7:39 AM on July 18, 2016
I can't believe a £24 billion corporate buyout has been planned and executed in the time between the Brexit referendum and now. These things take months, often years, to sort out.
Perhaps it's a better bargain for Softbank than it was three weeks ago, but do people really believe that Masayoshi Son got up one morning last week and thought "you know what'd be a laugh, let's buy ARM while the pound's cheap"?
Perhaps he was using a variation on The Shock Doctrine.
posted by fairmettle at 7:39 AM on July 18, 2016 [1 favorite]
Perhaps it's a better bargain for Softbank than it was three weeks ago, but do people really believe that Masayoshi Son got up one morning last week and thought "you know what'd be a laugh, let's buy ARM while the pound's cheap"?
Perhaps he was using a variation on The Shock Doctrine.
posted by fairmettle at 7:39 AM on July 18, 2016 [1 favorite]
Asset stripping is not investment or a sign of a healthy economy.
posted by Artw at 7:43 AM on July 18, 2016 [3 favorites]
posted by Artw at 7:43 AM on July 18, 2016 [3 favorites]
Winterhill, What convinces you Brexit had nothing to do with it?
I mean that was the first thing winterhill explained.
Is this how it's going to be for the next few months? Everything that happens, good, bad, indifferent or irrelevant, is going to set off a chorus of 'Brexit doom, brexit doom!' like a chorus of euro-loving frogs?
posted by Segundus at 7:43 AM on July 18, 2016 [1 favorite]
I mean that was the first thing winterhill explained.
Is this how it's going to be for the next few months? Everything that happens, good, bad, indifferent or irrelevant, is going to set off a chorus of 'Brexit doom, brexit doom!' like a chorus of euro-loving frogs?
posted by Segundus at 7:43 AM on July 18, 2016 [1 favorite]
Well, if it's related to the pound being in the toilet, yes.
posted by Artw at 7:44 AM on July 18, 2016 [2 favorites]
posted by Artw at 7:44 AM on July 18, 2016 [2 favorites]
This is a stellar purchase made sweeter by Brexit. Intel's dominance in the server market generates massive profits and if ARM based servers like the X-Gene 3 can crack into that market it could be a game changer. Yeah x86 is currently dominant and will likely continue to be so for an extended time period I could see some of the super scale companies latching onto a new technology if it offers an increase in computing performance or a decrease in operating costs.
posted by vuron at 7:53 AM on July 18, 2016 [1 favorite]
posted by vuron at 7:53 AM on July 18, 2016 [1 favorite]
So what would you have said about it if SoftBank had announced they had considered making this move but in current conditions had decided not to go ahead after all? Would it by any chance have been hard to see that as anything other than another sign of Brexit doom?
posted by Segundus at 7:55 AM on July 18, 2016 [3 favorites]
posted by Segundus at 7:55 AM on July 18, 2016 [3 favorites]
Asset stripping is not investment or a sign of a healthy economy.
I'm mostly disappointed that it wasn't one of the big Indian or Chinese firms. Maybe in a decade or two, even Nigeria and Kenya can take a whack at it.
posted by indubitable at 7:55 AM on July 18, 2016 [2 favorites]
I'm mostly disappointed that it wasn't one of the big Indian or Chinese firms. Maybe in a decade or two, even Nigeria and Kenya can take a whack at it.
posted by indubitable at 7:55 AM on July 18, 2016 [2 favorites]
Would it by any chance have been hard to see that as anything other than another sign of Brexit doom?
I would have seen it as dodging a bullet.
posted by Artw at 8:01 AM on July 18, 2016
I would have seen it as dodging a bullet.
posted by Artw at 8:01 AM on July 18, 2016
Yet a "disruptive" company losing hundreds of millions of dollars a year in an increasingly hostile regulatory environment is supposedly worth twice ARM's selling price.
posted by gyc at 8:07 AM on July 18, 2016 [2 favorites]
posted by gyc at 8:07 AM on July 18, 2016 [2 favorites]
As I said in the Brexit thread, a company I was intimately acquainted with was bought by SoftBank in the 1990s. Masayoshi-son himself embarked on a world tour and told the employees that he planned a huge expansion of the company and even alluded to a '300 year plan', which was treated as you might expect.
The 'huge expansion' turned out to be in the company's indebtedness, as SoftBank took the massive debt they'd created in the course of the leveraged deal and put it on the acquired company's books. The company had been truly world-class, hugely profitable and was showing every sign of being willing and able to cope with the forthcoming disruptions in the industry (it had already executed and very niftily abandoned one mistaken foray into online, escaping with a profit and a ton of experience), but now all the profits went into debt maintenance. R&D, expansion, new products were either abandoned or launched under extreme constraints and after a while the company lost its pre-eminent position, shrank and was eventually sold off in chunks to its competitors.
This isn't the whole story, of course, but I have the strongest possible memory of a conversation I had with the local MD after his bit of the company had been stripped and closed down, after a decade which had seen outrageous success. I asked him why what had just happened wasn't actually criminal, and he sighed and said 'It should be, it absolutely should be'. (He left the industry, despite being headhunted by people you know, and worked in charity thereafter; the company had recruited and paid for the best, and they had seen everything they'd worked for thrown away with absolutely nothing they could do about it, and not everyone got back on the horse.)
It was the day I looked into the heart of the beast and found it loathsome in its inhumanity and cold-souled deceptiveness.
Perhaps SoftBank has changed since then, but I doubt it. I don't doubt that Softbank could have put this offer together in a very short time; it has access to some very deep pockets and a history of lining them well.
ARM is going to be unrecognisable in five years' time. If it takes that long. I'm sorry for all the excellent people who work there, and sad for all the brilliant physical and financial engineering that has gone into its own enormous success.
This is just the beginning.
posted by Devonian at 8:10 AM on July 18, 2016 [29 favorites]
The 'huge expansion' turned out to be in the company's indebtedness, as SoftBank took the massive debt they'd created in the course of the leveraged deal and put it on the acquired company's books. The company had been truly world-class, hugely profitable and was showing every sign of being willing and able to cope with the forthcoming disruptions in the industry (it had already executed and very niftily abandoned one mistaken foray into online, escaping with a profit and a ton of experience), but now all the profits went into debt maintenance. R&D, expansion, new products were either abandoned or launched under extreme constraints and after a while the company lost its pre-eminent position, shrank and was eventually sold off in chunks to its competitors.
This isn't the whole story, of course, but I have the strongest possible memory of a conversation I had with the local MD after his bit of the company had been stripped and closed down, after a decade which had seen outrageous success. I asked him why what had just happened wasn't actually criminal, and he sighed and said 'It should be, it absolutely should be'. (He left the industry, despite being headhunted by people you know, and worked in charity thereafter; the company had recruited and paid for the best, and they had seen everything they'd worked for thrown away with absolutely nothing they could do about it, and not everyone got back on the horse.)
It was the day I looked into the heart of the beast and found it loathsome in its inhumanity and cold-souled deceptiveness.
Perhaps SoftBank has changed since then, but I doubt it. I don't doubt that Softbank could have put this offer together in a very short time; it has access to some very deep pockets and a history of lining them well.
ARM is going to be unrecognisable in five years' time. If it takes that long. I'm sorry for all the excellent people who work there, and sad for all the brilliant physical and financial engineering that has gone into its own enormous success.
This is just the beginning.
posted by Devonian at 8:10 AM on July 18, 2016 [29 favorites]
This is one of Britains greatest technical success stories of recent years, which are years where success such stories haven't exactly been common, pissed away like nothing.
posted by Artw at 8:12 AM on July 18, 2016 [2 favorites]
posted by Artw at 8:12 AM on July 18, 2016 [2 favorites]
The other low-cost RISC processor is MIPS, which is in decline as ARM has come to dominate the phone/tablet market. Having everyone's apps being able to run on the chip's instruciton set, several companies designing their own ARM processors (including Apple and Samsung), and all the companies offering various tools and accessories, it is hard for a second instruction set to compete, which is why RISC-V is not really a threat.
I wouldn't count Intel out on the server side, though. ARM has a long, long way to go to be more than a niche in this market and Intel is not standing still.
posted by eye of newt at 8:15 AM on July 18, 2016 [3 favorites]
I wouldn't count Intel out on the server side, though. ARM has a long, long way to go to be more than a niche in this market and Intel is not standing still.
posted by eye of newt at 8:15 AM on July 18, 2016 [3 favorites]
If I were Intel today I'd be dusting off all the mobile projects and pumping money into them.
posted by Artw at 8:20 AM on July 18, 2016 [4 favorites]
posted by Artw at 8:20 AM on July 18, 2016 [4 favorites]
Could Devonian's mystery SoftBank acquisition be Ziff-Davis?
posted by notyou at 8:21 AM on July 18, 2016 [1 favorite]
posted by notyou at 8:21 AM on July 18, 2016 [1 favorite]
MIPS itself was bought a few years ago by the British company Imagination. It's doing a bit better than it was under the hedge fund that previously owned it.
posted by monotreme at 8:24 AM on July 18, 2016
posted by monotreme at 8:24 AM on July 18, 2016
This is huge. I hope to see more variations on the RPi.
posted by aspersioncast at 8:26 AM on July 18, 2016
posted by aspersioncast at 8:26 AM on July 18, 2016
The ARM acquisition hasn't made SoftBank's Corporate history page yet, but dang, that's a lot of corporate entities over the years!
posted by notyou at 8:30 AM on July 18, 2016
posted by notyou at 8:30 AM on July 18, 2016
If I were Intel today I'd be dusting off all the mobile projects and pumping money into them.--Artw
Intel is doing just the opposite. They recently cancelled all their mobile projects. Intel learned the hard way that in the mobile market it is all about the instruction set. They probably lost billions trying to be an alternative to ARM in the tablet/phone market. You can still buy nice tablets with Intel processors (from Acer and others), and you get more power for less money, but no one is buying them.
It is kind-of the same thing in the server market, but with the Intel instruction set. With Network Functional Virtualization (NFV), companies can have their software run on any server in the cloud as needed on a moments notice--but that often only works if it is a known instruction set. If ARM can get into enough servers, then maybe people will port to work on either, but they have a long way to go and, as I mentioned, Intel is not standing still.
And, of course, Intel dominates on desktops. Microsoft tried to change this, but their ARM machines running Windows flopped.
posted by eye of newt at 8:30 AM on July 18, 2016 [5 favorites]
Intel is doing just the opposite. They recently cancelled all their mobile projects. Intel learned the hard way that in the mobile market it is all about the instruction set. They probably lost billions trying to be an alternative to ARM in the tablet/phone market. You can still buy nice tablets with Intel processors (from Acer and others), and you get more power for less money, but no one is buying them.
It is kind-of the same thing in the server market, but with the Intel instruction set. With Network Functional Virtualization (NFV), companies can have their software run on any server in the cloud as needed on a moments notice--but that often only works if it is a known instruction set. If ARM can get into enough servers, then maybe people will port to work on either, but they have a long way to go and, as I mentioned, Intel is not standing still.
And, of course, Intel dominates on desktops. Microsoft tried to change this, but their ARM machines running Windows flopped.
posted by eye of newt at 8:30 AM on July 18, 2016 [5 favorites]
MIPS itself was bought a few years ago by the British company Imagination. It's doing a bit better than it was under the hedge fund that previously owned it.--monotreme
I personally work with MIPS processors. The companies that make them are converting to ARM. I am sorry to say that MIPS is doomed, in my opinion.
posted by eye of newt at 8:34 AM on July 18, 2016 [3 favorites]
I personally work with MIPS processors. The companies that make them are converting to ARM. I am sorry to say that MIPS is doomed, in my opinion.
posted by eye of newt at 8:34 AM on July 18, 2016 [3 favorites]
I'd guess that SoftBank has been considering a move like this for years. Who wouldn't consider buying ARM if the price was right? Then it drops %12 in price (or %30 depending on your time frame)."Sterling down about 12% against yen since June 23 to last Friday. so deal about £3bln less BUT ARM shares up about 12% in same time period"
posted by Auz at 8:41 AM on July 18, 2016 [1 favorite]
MIPS has been doomed for what, 10 years now? It's a slow death.
Most specifically: MIPS doesn't do low-power well, which kills them in mobile. (Our MIPS rep showed off a powered-by-MIPS Android phone a few years ago. "How's the battery life?" "Um.")
posted by We had a deal, Kyle at 8:41 AM on July 18, 2016 [3 favorites]
Most specifically: MIPS doesn't do low-power well, which kills them in mobile. (Our MIPS rep showed off a powered-by-MIPS Android phone a few years ago. "How's the battery life?" "Um.")
posted by We had a deal, Kyle at 8:41 AM on July 18, 2016 [3 favorites]
it is hard for a second instruction set to compete, which is why RISC-V is not really a threat
Assuming that binary compatibility is important. Given the growth in VMs and JIT compilers, and open, extensible toolchains like LLVM, this is not as much the case as it once was. If, say, a wearable device with a custom-tailored instruction set is more power-efficient/performant, requiring developers to use a custom LLVM backend for building binaries for it (or shipping the binaries in a bytecode which is compiled to the native machine code at install time) is not a huge deal.
There is the possibility of vendors shipping special optimising compilers tailored to their architecture (Intel have a C compiler which is highly rated), though these days, it'd make more sense to build these as optimisation steps in the LLVM toolchain than to build a compiler from the ground up.
posted by acb at 8:45 AM on July 18, 2016 [5 favorites]
Assuming that binary compatibility is important. Given the growth in VMs and JIT compilers, and open, extensible toolchains like LLVM, this is not as much the case as it once was. If, say, a wearable device with a custom-tailored instruction set is more power-efficient/performant, requiring developers to use a custom LLVM backend for building binaries for it (or shipping the binaries in a bytecode which is compiled to the native machine code at install time) is not a huge deal.
There is the possibility of vendors shipping special optimising compilers tailored to their architecture (Intel have a C compiler which is highly rated), though these days, it'd make more sense to build these as optimisation steps in the LLVM toolchain than to build a compiler from the ground up.
posted by acb at 8:45 AM on July 18, 2016 [5 favorites]
For what its worth, the view of one of ARM's founders.
"ARM is the proudest achievement of my life. The proposed sale to SoftBank is a sad day for me and for technology in Britain."
Hermannhauser (twitter)
Asset stripping is not investment or a sign of a healthy economy.
Do we have evidence that this takeover will lead to asset stripping? I'm agnostic but it doesn't make sense to strip and cripple a growing company.
This is still a rapidly growing player in the industry with a lot of upside. I have a good friend who left Intel to work on a low-energy cost server farm start-up based on ARM chips.
posted by C.A.S. at 9:03 AM on July 18, 2016 [2 favorites]
"ARM is the proudest achievement of my life. The proposed sale to SoftBank is a sad day for me and for technology in Britain."
Hermannhauser (twitter)
Asset stripping is not investment or a sign of a healthy economy.
Do we have evidence that this takeover will lead to asset stripping? I'm agnostic but it doesn't make sense to strip and cripple a growing company.
This is still a rapidly growing player in the industry with a lot of upside. I have a good friend who left Intel to work on a low-energy cost server farm start-up based on ARM chips.
posted by C.A.S. at 9:03 AM on July 18, 2016 [2 favorites]
Put it like this - ARM is ARM Because they took a longshot on devolving a new kind of processor back in the 80s and leapfrogged the competition. Reading the descriptions of SoftBank above do you see them making a similar kind of investment or do you see them at best riding the current technology to the end of its lifespan?
posted by Artw at 9:14 AM on July 18, 2016 [4 favorites]
posted by Artw at 9:14 AM on July 18, 2016 [4 favorites]
Given the (lack of) penetration by AMD into the server room, never mind the (perceived) lack of success by SPARC in the same space, it's hard to forecast the future of either Intel or ARM in the server room. There's no synergy from my phone and my server running on the same instruction set, so while I could see ARM succeeding in the server room, it would be on its own merits and not due to dominance in the embedded world. (Assuming no game-changing smartphone-cloud integrated platform.)
As far as investing in their purchases, I don't have an insider's view of how Sprint is really doing, but since they were bought by Softbank, their network has gotten better, in part due to money Softbank has been injecting.
posted by fragmede at 9:15 AM on July 18, 2016 [2 favorites]
As far as investing in their purchases, I don't have an insider's view of how Sprint is really doing, but since they were bought by Softbank, their network has gotten better, in part due to money Softbank has been injecting.
posted by fragmede at 9:15 AM on July 18, 2016 [2 favorites]
Considering Softbank's own extensive stake in mobile hardware and services, it would behoove them to not ratfuck the CPU supplier to their other operations, wouldn't it?
I mean, it might happen anyway through poor management but that isn't the same thing as acquiring a company specifically for the purpose of laundering their own debt.
posted by at by at 9:23 AM on July 18, 2016 [1 favorite]
I mean, it might happen anyway through poor management but that isn't the same thing as acquiring a company specifically for the purpose of laundering their own debt.
posted by at by at 9:23 AM on July 18, 2016 [1 favorite]
It's hard to see this as anything but pure Brexit collateral damage.
I'd be more inclined to blame the Snooper's Charter for providing a distinct incentive to migrate ARM development off the United Kingdom.
posted by ocschwar at 9:29 AM on July 18, 2016 [1 favorite]
I'd be more inclined to blame the Snooper's Charter for providing a distinct incentive to migrate ARM development off the United Kingdom.
posted by ocschwar at 9:29 AM on July 18, 2016 [1 favorite]
I think it's a bit Eurocentric to say this has to do with Brexit. Probably it would have happened whichever way the vote went. This is more about the rise of Asian countries as players on the international stage. Softbank is executing on their "2.0 transformation strategy". Back in March, they spun out an international "arm," and they have been gathering up cash to make a purchase like this for a while now.
posted by mantecol at 9:38 AM on July 18, 2016 [1 favorite]
posted by mantecol at 9:38 AM on July 18, 2016 [1 favorite]
Given the (lack of) penetration by AMD into the server room, never mind the (perceived) lack of success by SPARC in the same space, it's hard to forecast the future of either Intel or ARM in the server room. There's no synergy from my phone and my server running on the same instruction set, so while I could see ARM succeeding in the server room, it would be on its own merits and not due to dominance in the embedded world. (Assuming no game-changing smartphone-cloud integrated platform.)
Intel's success in the server room was driven partly by their being at the forefront of Moore's Law. Buy an Intel server board, and you have the longest amount of time before software bloat makes you replace it. It was also driven by the "nobody ever got fired choosing $VENDOR" culture of the corporate MBA layer.
Well, Moore's Law is coming to an end. And considerable amount of progress has gone into facilitating not just a halt to software bloat but a retreat from it, especially on the server side, where it pays to be the one hiring the CS PhDs who can tell an O(log(n)) method from a hole in the ground.
At that point, where does it leave Intel? Selling chips with huge amounts of real estate devoted to floating point processing which the server people don't need. Counting on a receding tide of software bloat to produce a demand for their product, and never ending Moore's Law progress to produce a supply, which is coming to an end.
Meanwhile, the ARM and (yes, them too) SPARC instruction sets are tailor made for server products that aren't, oh, calculating protein folding. Put lots of cores on a die. Put in a good mesh for concurrency. And let the Rust programming language finish the case for your server line. (Code that will not compile if it isn't thread-safe!)
I would not buy Intel stock today. If SoftBank and Oracle don't find a way to screw the pooch, I expect them to take over the server board business in short order.
posted by ocschwar at 9:48 AM on July 18, 2016 [4 favorites]
Is SPARC still a going concern outside of niche areas (i.e., large institutional clients with ongoing SPARC orders too big for Oracle to rationalise out of existence)?
posted by acb at 10:06 AM on July 18, 2016
posted by acb at 10:06 AM on July 18, 2016
If Oracle is stupid enough to kill SPARC off instead of expanding the line, then ARM will move into the server room instead.
If I were a big database client, I'd much prefer SPARC. But at Oracle, the MBAs call the shots. So your implied hunch is probably correct.
So it goes..
posted by ocschwar at 10:15 AM on July 18, 2016 [2 favorites]
If I were a big database client, I'd much prefer SPARC. But at Oracle, the MBAs call the shots. So your implied hunch is probably correct.
So it goes..
posted by ocschwar at 10:15 AM on July 18, 2016 [2 favorites]
ARM's founder's view, to the FT:
"SoftBank is a reasonable acquirer but this is sadly one of the unintended consequences of Brexit. The fall in sterling has made this very cheap and while I suspect they were considering it for a long time, they have acted now because the opportunity is there."
The Buzzfeed summary of what he said to the FT is here, since it's behind a paywall.
posted by dowcrag at 10:17 AM on July 18, 2016 [3 favorites]
"SoftBank is a reasonable acquirer but this is sadly one of the unintended consequences of Brexit. The fall in sterling has made this very cheap and while I suspect they were considering it for a long time, they have acted now because the opportunity is there."
The Buzzfeed summary of what he said to the FT is here, since it's behind a paywall.
posted by dowcrag at 10:17 AM on July 18, 2016 [3 favorites]
We're betting on Oracle?
And that's the REAL Godwin's Law for technical threads.
"As the length of discussion tends towards infinity, the probability of a mention of a piece of technology languishing in Oracle's portfolio tends towards one."
posted by ocschwar at 11:35 AM on July 18, 2016 [2 favorites]
And that's the REAL Godwin's Law for technical threads.
"As the length of discussion tends towards infinity, the probability of a mention of a piece of technology languishing in Oracle's portfolio tends towards one."
posted by ocschwar at 11:35 AM on July 18, 2016 [2 favorites]
Being acquired by Oracle is probably better than being acquired by SCO, but not by much. These guys seem somewhere in the middle.
posted by Artw at 11:41 AM on July 18, 2016
posted by Artw at 11:41 AM on July 18, 2016
Considering Softbank's own extensive stake in mobile hardware and services, it would behoove them to not ratfuck the CPU supplier to their other operations, wouldn't it?
The problem with this logic is that ARM isn't a CPU supplier. They sell sample designs and specifications, and their customers modify them (for example by adding cell phone radios), and then fabricate the chips themselves or contract out for that. Most often the mobile phone manufacturer is buying one of those chips - usually (but not always) a Samsung or Qualcomm chip. The big exception is Apple, who contract with Samsung and TSMC to fab their own ARM-derived Apple A9.
posted by atbash at 12:27 PM on July 18, 2016 [1 favorite]
The problem with this logic is that ARM isn't a CPU supplier. They sell sample designs and specifications, and their customers modify them (for example by adding cell phone radios), and then fabricate the chips themselves or contract out for that. Most often the mobile phone manufacturer is buying one of those chips - usually (but not always) a Samsung or Qualcomm chip. The big exception is Apple, who contract with Samsung and TSMC to fab their own ARM-derived Apple A9.
posted by atbash at 12:27 PM on July 18, 2016 [1 favorite]
Arm is a pure IP company; it sells licences to designs. Like a publishing company, it doesn't have any assets to strip but the IP in what it produces and the brains to produce it. I don't know what SoftBank wants to do with the company, but I don't imagine it's got much interest in chip design - I can see the various IP blocks being sold on to people who'd like to control the market and have large cash reserves to make that happen.
I can think of a couple of candidates.
posted by Devonian at 12:50 PM on July 18, 2016 [4 favorites]
I can think of a couple of candidates.
posted by Devonian at 12:50 PM on July 18, 2016 [4 favorites]
We don't yet know when Softbank began its internal deliberations about acquiring ARM, but according to the NYT, "Mr. Son said he first spoke with ARM’s chairman about two weeks ago regarding a possible takeover, and added that the deal came together quickly."
Brexit was 3.5 weeks ago, so it would certainly have been an influence - perhaps it tipped their decision the over the edge.
posted by adrianhon at 4:01 PM on July 18, 2016
Brexit was 3.5 weeks ago, so it would certainly have been an influence - perhaps it tipped their decision the over the edge.
posted by adrianhon at 4:01 PM on July 18, 2016
I think Devonian nails this play. The talk of doubling the workforce is code for loading on mountains of debt in a phoney expansion that the parent company is handsomely compensated for in consulting fees and whatnot, before the expansion is declared a failure and the company is liquidated by auctioning off the IP to be weaponized by the world's deepest pocketed monopolists.
posted by [expletive deleted] at 4:10 PM on July 18, 2016 [4 favorites]
posted by [expletive deleted] at 4:10 PM on July 18, 2016 [4 favorites]
Speaking at a very high level of generalization, printing yen weakens the currency and raises the price of imports.
Which is what Kuroda wants, to make Japanese exports more competitive, boost import prices to get some good ol' wage-price inflation going, the only way it's going to be able to liquidate its central government debt, short of Kuroda just buying it all and forgetting about where he put the bonds.
The 80 yen regime was killing Japan, Inc.
Like I said, I think 120 is the sweet spot and 150 is tolerable.
Back when the yen was at 80 it hit me that Japan didn't have any macro-economic problems it couldn't print itself out of .
posted by Heywood Mogroot III at 5:48 PM on July 18, 2016
Which is what Kuroda wants, to make Japanese exports more competitive, boost import prices to get some good ol' wage-price inflation going, the only way it's going to be able to liquidate its central government debt, short of Kuroda just buying it all and forgetting about where he put the bonds.
The 80 yen regime was killing Japan, Inc.
Like I said, I think 120 is the sweet spot and 150 is tolerable.
Back when the yen was at 80 it hit me that Japan didn't have any macro-economic problems it couldn't print itself out of .
posted by Heywood Mogroot III at 5:48 PM on July 18, 2016
It's hard to see this as anything but pure Brexit collateral damage.
I'd be more inclined to blame the Snooper's Charter for providing a distinct incentive to migrate ARM development off the United Kingdom.
I may just be undercaffeinated, but how does the Snooper's Charter provide such an incentive again?
posted by iffthen at 5:56 PM on July 18, 2016 [1 favorite]
I may just be undercaffeinated, but how does the Snooper's Charter provide such an incentive again?
It could be interpreted as mandating to UK-based designers of software and hardware to obey and keep secret directives from GCHQ. So, for instance, they could require that the microcode or logic on an ARM core was changed in a way which, combined with other tools, would give them an unpatchable back door to any ARM-based system.
posted by acb at 2:27 AM on July 19, 2016 [3 favorites]
It could be interpreted as mandating to UK-based designers of software and hardware to obey and keep secret directives from GCHQ. So, for instance, they could require that the microcode or logic on an ARM core was changed in a way which, combined with other tools, would give them an unpatchable back door to any ARM-based system.
posted by acb at 2:27 AM on July 19, 2016 [3 favorites]
There's no microcode in ARM (simplistically, that's practically one working definition of RISC, inasmuch as that term still has meaning). It's also difficult to hide functions in there, as manufacturers get what's basically a logic-gate-level design (albeit partitioned into many larger functional units) to implement themselves using standard tools, and that includes a verification scheme that exercises (to some extent) all of the chip's functional parameters. Plus, OEMs get a lot of freedom to reconfigure, omit and add to those functional blocks - so what do you hide, and where do you hide it?
It is possible to envisage mandating some closed tool chain designed to actively hide things from the manufacturers during the implementation and production process, but vanishingly improbable that it could be kept secret for any length of time. It's also very difficult to imagine keeping the design of the toolchain itself sufficiently secret, as the whole business needs a lot of engineers - only one needs to be sufficiently motivated to slip the smallest detail out to the hacker community, and the gaff is blown.
Which is not to say that the IP Bill isn't grotesquely wrong for even going down that path - it is, and it speaks volumes about the technical awareness of the lawmakers - but I would be very surprised if anything like this could be done without it being obvious that it was being done. It's not like slipping a back door into a router BIOS.
posted by Devonian at 8:36 AM on July 19, 2016 [2 favorites]
It is possible to envisage mandating some closed tool chain designed to actively hide things from the manufacturers during the implementation and production process, but vanishingly improbable that it could be kept secret for any length of time. It's also very difficult to imagine keeping the design of the toolchain itself sufficiently secret, as the whole business needs a lot of engineers - only one needs to be sufficiently motivated to slip the smallest detail out to the hacker community, and the gaff is blown.
Which is not to say that the IP Bill isn't grotesquely wrong for even going down that path - it is, and it speaks volumes about the technical awareness of the lawmakers - but I would be very surprised if anything like this could be done without it being obvious that it was being done. It's not like slipping a back door into a router BIOS.
posted by Devonian at 8:36 AM on July 19, 2016 [2 favorites]
ARM's business model makes compliance with the Snooper's Charter difficult.
So the Crown orders ARM to change its business model.
Or, ARM's business model integrates a "get off this island now" component.
posted by ocschwar at 10:29 AM on July 19, 2016
So the Crown orders ARM to change its business model.
Or, ARM's business model integrates a "get off this island now" component.
posted by ocschwar at 10:29 AM on July 19, 2016
Couple of notes...
Intel: If there's one truism I've cultivated in 30 years in the IT business it's that betting against Intel is the ultimate fools errand. ARM is legitimately challenging Intel in workstation & server. End of Intel hegemony? You mean like AMD was going to end it in the early 2000s? Yeah...good times. Intel is remarkably good at finding itself at a disadvantage, backing up to regroup, and then applying cash like napalm to engineer themselves out of the corner. Have you seen the current gen Xeon? Anyone want to take a side bet on how long Intel stays out of the mobile processor market? Hell...there are still billions of decedents of the Intel 8051 shipped every year.
MIPS : The "not-ARM" choice in the embedded/mobile market, especially in networked devices. MIPS ceded the workstation/server/super market a 15 years ago, but there's some life in that area with folks like Longsoon. There have been some recent painful losses (Cavium moving from MIPS to ARM and the loss of the exclusive deal with Broadcom (also to ARM)) but it's awfully premature to declare MIPS 'dead' (it's been done before). STMicro seems to be doing great with STM32, for example, and between Russian and Chinese MIPS processors (as the "not-Intel" or perhaps "not-Western" options) as well as RISC-V there's a good bit of R&D life left there. RISC-V holds out the promise of a really open source platform.
SPARC: Fujitsu is keeping the SPARC64 torch alive as a general purpose processor, but they're bit-players. Oracle is grudgingly keeping the line going, but Niagara and follow on designs seem to be evolving towards custom Oracle/MySQL processors that happen sometimes be good at other highly threaded workloads. SPARC has ceded the embedded market and never played in the mobile market. If any arch is dead, it's SPARC.
PPC/POWER: Still a lot of life in PowerPC in the embedded market, though a no-show in mobile. PPC ceded the other markets a long time ago to POWER. POWER is delivering really, really powerful POWER7 & 8 processors in the workstation/server/super space, but no idea how many they're selling. IBM is now licensing POWER, which is exciting, but the ecosystem doesn't seem to be evolving that much.
ARM: ARM does seem to winning on all fronts. We'll see how long it lasts. FWIW...ARMs business model in no way limits their ability comply with any Snooper mandate. They just have to make the surveillance IP part of the required core ARM definition and the licenses will have to include it to meet the license terms. Hopefully they won't, but you know...
And since we're (sorta) on the topic, one of my sentimental favorite chips, the 32-bit Hitachi SuperH, had all it's patents expire. This is allowing an open source project to reimplement the architecture (details: http://0pf.org/j-core.html). They want an end-to-end open source processor ecosystem and are making good progress. The SH4 is their goal, and it's a great little chip.
posted by kjs3 at 3:40 PM on July 19, 2016 [5 favorites]
Intel: If there's one truism I've cultivated in 30 years in the IT business it's that betting against Intel is the ultimate fools errand. ARM is legitimately challenging Intel in workstation & server. End of Intel hegemony? You mean like AMD was going to end it in the early 2000s? Yeah...good times. Intel is remarkably good at finding itself at a disadvantage, backing up to regroup, and then applying cash like napalm to engineer themselves out of the corner. Have you seen the current gen Xeon? Anyone want to take a side bet on how long Intel stays out of the mobile processor market? Hell...there are still billions of decedents of the Intel 8051 shipped every year.
MIPS : The "not-ARM" choice in the embedded/mobile market, especially in networked devices. MIPS ceded the workstation/server/super market a 15 years ago, but there's some life in that area with folks like Longsoon. There have been some recent painful losses (Cavium moving from MIPS to ARM and the loss of the exclusive deal with Broadcom (also to ARM)) but it's awfully premature to declare MIPS 'dead' (it's been done before). STMicro seems to be doing great with STM32, for example, and between Russian and Chinese MIPS processors (as the "not-Intel" or perhaps "not-Western" options) as well as RISC-V there's a good bit of R&D life left there. RISC-V holds out the promise of a really open source platform.
SPARC: Fujitsu is keeping the SPARC64 torch alive as a general purpose processor, but they're bit-players. Oracle is grudgingly keeping the line going, but Niagara and follow on designs seem to be evolving towards custom Oracle/MySQL processors that happen sometimes be good at other highly threaded workloads. SPARC has ceded the embedded market and never played in the mobile market. If any arch is dead, it's SPARC.
PPC/POWER: Still a lot of life in PowerPC in the embedded market, though a no-show in mobile. PPC ceded the other markets a long time ago to POWER. POWER is delivering really, really powerful POWER7 & 8 processors in the workstation/server/super space, but no idea how many they're selling. IBM is now licensing POWER, which is exciting, but the ecosystem doesn't seem to be evolving that much.
ARM: ARM does seem to winning on all fronts. We'll see how long it lasts. FWIW...ARMs business model in no way limits their ability comply with any Snooper mandate. They just have to make the surveillance IP part of the required core ARM definition and the licenses will have to include it to meet the license terms. Hopefully they won't, but you know...
And since we're (sorta) on the topic, one of my sentimental favorite chips, the 32-bit Hitachi SuperH, had all it's patents expire. This is allowing an open source project to reimplement the architecture (details: http://0pf.org/j-core.html). They want an end-to-end open source processor ecosystem and are making good progress. The SH4 is their goal, and it's a great little chip.
posted by kjs3 at 3:40 PM on July 19, 2016 [5 favorites]
« Older She's been working on the railroad | “Is she going to survive?” “Unfortunately, yes.” Newer »
This thread has been archived and is closed to new comments
posted by mhoye at 6:26 AM on July 18, 2016 [12 favorites]