Electronics Weekly Electronics Design & Components Tech News
Received wisdom was that ARM’s market cap was so big compared to its profits that no one would be crazy enough to buy it.
Now someone is.
It was also thought that big ARM customers like Apple and Qualcomm would push up ARM’s share price in the event of a bid to price any bidder for ARM out of the market.
Well, at $32 billion, not even those guys probably have the stomach to keep ARM independent.
So is SoftBank crazy?
Well ARM’s value is based on the fact that it owns the incumbent microprocessor architecture in mobile.
SoftBank says it is buying ARM because of its position in IoT. But there is no incumbent processor architecture in IoT and neither ARM, nor anyone else, has a significant position in IoT.
There’s a case for saying that ARM will become the dominant world architecture in all things electronic but how real is it?
The data centre, the biggest opportunity in the chip business, has so far resisted ARM’s incursions and personal computing won’t switch from x86.
Meanwhile ARM’s customers can, of course, move to a new processor core and ARM-compatible cores allow them to do so.
So it’s a lot of money to pay for a business which is not all that defensible and, if SoftBank slips up with ARM’s customers, Intel will be waiting
Great to hear from you Darce. As I see it ARM’s dominance in mobile was largely accidental, but it wouldn’t have happened if people didn’t trust Saxby. Everyone trusted Saxby. In IoT there is no dominant processor, there are loads of cores aspiring to dominate, but no reason why any particular one should become dominant. So, while any of them could dominate, would IoT players trust a core controlled by a heavily indebted, huge, telecoms company?
Hey Manners!
You seem to be very sceptical about IoT. I agree that it is indeterminate at the moment. It’s something that everybody knows is going to happen but is unsure exactly what shape it is going to take. (A bit like Brexit then 🙂 )
I am reminded of press conferences back in the 1990s when Robin Saxby used to say that his vision was for ARM to dominate in the embedded market and become “the Z80 of the 1990s”. Could that skip a generation and does ARM have the potential to become the “Z80 of the Internet of Things”? If it does, the $32bn price tag could yet be a bargain.
Maybe ExST, but history suggests that it’s very difficult, maybe impossible, to dislodge a dominant architecture.
Personally I wouldn’t call Google’s TPUs a new kind of hardware, ExST, – they’re ASICs optimised to run the Tensorflow programme so it’s the software programme which is defining the hardware not vice versa.
Absolutely right. But it threatens Intel+Nvidia in that space. And many similar things are going on in the embedded AI space that will threaten ARM’s dominance.
I agree that ARM has no advantage in AI, Ex-ST, but IMHO if AI needs new HW then AI won’t happen because the benefits of scaling have diminished and are continuing to diminish. What is improving very quickly, and can be further improved, is the efficiency of SW. It is said that no one notices the introduction of AI because, once it is introduced into products, it ceases to be seen as AI. And this, IMO, is how AI is quite quickly getting introduced into the world e.g. through algorithmic share trading systems, automated accountancy packages, automated legal advice systems, automated medical diagnostic systems etc etc.
AI is already happening. Google is building its data center computers around their TPUs already.
The human brain does not run software. Think about it. The brain is a huge associative memory. It’s bad at calculations but it’s damn good at creating connections, at fuzzy pattern matching and almost instantaneous classification. CPUs are really good at calculations but very bad at everything else. A new architecture is bound to happen if AI is to exist in a serious way. You are right, AI is invisible. But, don’t tell me it’s any different from anything we’ve seen before. The ARM instruction set, the ARM core, the Application processors are totally invisible from the lunatics who play Pokemon Go. People care about the service, not the technology behind them. The fact is that if you want to analyse all data generated by billions of sensors, a very significant architecture change is needed in computers. Today, most of the data goes to the bin. If you want to analyse the data you collect (otherwise, why collect it) you have to do some analytics close to the sensors, otherwise the networks will be congested. Therefore, embedded AI is a must. A robot that needs a Cloud connection to do anything, is potentially useless and potentially dangerous, plus it’s a spy. And it’s a very slow robot. The brain gives you an answer in 100ms. Show me a cloud connected AI system that gives an answer in 100ms flat whether you are at home, at work, in your car, on a beach, in a mine or on an oil rig and that does not cost a fortune per MB of data and I’ll accept that AI can be entirely Cloud based.
AI/ANN’s is currently successfully running on GPU’s.
The brain is simplified some hundred billion interconnected neurons/”shaders” running in parallel. Of course AI will need proper hardware support to run efficiently on embedded/server systems.
Yes indeed Bitter, I agree with you that IoT is smoke and mirrors and that the money is mostly in the datacentre. AI seems to creep into the handset bit by bit as new clever functions and apps are added but IMHO the big AI advances come when you ask Google a question – legal, medical, financial or just how to fix your TV. The answers seem increasingly intelligent and this, to my mind, is where AI is getting into peoples’ lives.
The telcos want more intelligence in the handset. AI is where it’s at. The battle is to be wrought out in the datacenters or perhaps in the handset? IoT is just smoke and mirrors.
Transmeta failed trying to replace HW with SW. Let’s see how this “battle” will go. Is general purpose “realtime” AI possible without an efficient HW implementation/accelerator?
Realtime AI is not possible without new HW. This new HW will be significantly different from the traditional processors to avoid the infamous Von Neumann bottleneck and allow true, scalable, parallelism. ARM’s dominant position in mobile APs and in embedded 32bit MCUs is no guarantee that they will dominate the AI space. This is true both on the datacenter side and on the edge side.
AI is a threat to both Intel and ARM. It is also a threat to Nvidia. GPUs are benefitting from AI right now because they are a bit better than CPUs for deep learning but it’s just a gap filler before something better comes. On the datacenter side, Google has already made a first move with its TPU which is a small step in the right direction. I expect Facebook, Amazon and perhaps Apple will come up with some new hardware as well. On the edge side, we’ll see who comes up with the right answer but I don’t see ARM having any particular advantage.
Spot on, this is all being driven by fees, skenneth, the bankers expect to pick up $120m in fees for this deal of which ARM contributes half. There’s no industrial logic to the deal. It destroys ARM’s value proposition as an honest broker. ARM is putting up no one to justify the deal to the press – probably because it knows it can’t be justified.
Absolutely, accurate, coruscating criticism!! I have always felt that when a suitor brings nothing to the table except cash – no market knowledge, no product knowledge, no synergies – then problems are bound to result! Unfortunately, in a free market economy, I do not think it is up to governments to intervene. One just prays that Chairmen, Boards and shareholders can see a bit further than the £ signs flashing in front of them. What a contrast to ARM’s own recent acquisition of Apical.
Oh dear, Howard, he sounds like one of those tame execs the city boys keep in tow to stick into companies which they want to take control of.
Yes Chris we’ve heard from Hermann but nothing from Robin. I don’t expect he’s chuffed but doubt he’d ever say anything which could hurt ARM.
It all happened so quickly.
Why would ARM want to sell to anyone? I know shareholders will benefit, but if I were an ARM employee I would be concerned for the long term .I understand the deal is going to be financed by debt, which is another reason to be worried.
1600 new hires in Cambridge? Try finding a home there. My daughter runs a relocation service in Cambridge, and the biggest problem is finding houses for her clients like ARM.
I wonder what Sir Robin, who was instrumental in making ARM what it is now, thinks about this?
Don’t Sell, Norming!! You still have a chance to save ARM for the nation.
Long ago, and because I have a tech background, I bought on a whim a few ARM shares at not much over £1. I’ve top sliced several times and still now can walk away at some £17. So a good whim.
I also spent a long time as a pensions trustee and a director of its large investment fund. I was never able to see a way of investing some of our pensioners money into supporting UK high-tech growth.
Now there is no company in the UK with the background and money to buy ARM. Maybe I should use my £17 to buy some Goldman Sachs?
I don’t think it matters who buys ARM, RubyT, it’s simply the fact of having any owner at all that threatens the trust for impartial dealing between customers which is at the heart of ARM’s business model.
OpenRISC, neitherstnorste?
possibly, if the new workmodel can be sold in the proper way
bidding on iot at component level if far more risky than betting on Iot at system level. I suspect that those who can (and I am not sure Softbank is in this club) are placing very expensive bets for the sake of kind of “energizing” the IoT timeline.
<aybe a good strategy, maybe not. In the meantime one thing is clear: there is now room for one or more new independent players in the embedded computing arena. Intel? not so sure. MIPS? too much out of the game. RISC V? no real big one pushing for it. Others? if so, David, please name one
>MIPS? too much out of the game.
Or perhaps not? Qualcomm bought Atheros whose chip sets were based on MIPS. In fact these chip sets still live on and I see there is now a Kickstarter round going on for an Omega2 IoT module
http://hackerboards.com/5-dollar-linux-equipped-omega2-iot-module-launches-on-kickstarter/
It seems there is still a lot of life left in the MIPS platform and it could be a fallback position for the rest of Qualcomm since android is available for MIPS.
Yes indeed Gormenghast, the tech community is depressed by this deal. Herman Hauser has condemned it and so have many others. The only people gloating are the MIPS and RISC-V guys and Intel who now see an opportunity to dominate the IoT space. ARM, stripped of its impartiality, cannot prosper. The City have stitched up ARM – getting a non-techie in as Chairman and getting him to sell it out. And people ask why the UK has no big tech firms!!
All Our Base Are Belong To Them 🙁
It seems a perfect irony that an important British enterprise losing its independence, self-determination and, indeed, Britishness is being hailed as a fine example of our post-Brexit destiny when the whole point of that exercise was to restore those very things.
Despite everything that has happened in the last few weeks, I found this story probably the most depressing of all. I might express my despair by ceremonially smashing the STM32 Cortex-M eval board that’s sitting on my windowsill (no great loss – ST’s useless, buggy, development framework renders the potentially great hardware all but useless anyway).
This is a very strange deal.
Softbank has always been daring with its acquisitions and straining the definition of a standard telco. The operator has a history as an Internet player and was an early investor in Alibaba before it grew into a hugely successful Internet content provider.
More recently the company bought US mobile operator Sprint and expected to go after the big two Verizon and AT&T. This has proved a difficult acquisition and Sprint is not challenging the big two and is indeed the 4th largest mobile operator behind T-Mobile.
A few months ago Softbank announced it was streamlining the company into a domestic unit and an overseas one and the overseas one was the one it saw as expanding the company’s revenues since the Japanese market is largely saturated. The ARM acquisition is its first big overseas move since the streamlining.
Softbank also sells the Pepper robot that reads emotions and is designed to look after the elderly. Each month it makes 1000 available and each month they are sold within minutes. It has since launched a Pepper variant for businesses.
Do you need to own a chip IP company for IoT or for robots? If anything, telcos are trying to be more a combination of Internet players and broadcasters, as far away in the foodchain as you can get from chips and hardware.
I see ARM as an investment but as mentioned in other comments is Softbank really the best owner for the company’s future?
Meanwhile, EW should start speculating as to which chip company Deutsche Telekom should buy. Or BT? Or Facebook, Amazon, and Google for that matter.
And that perhaps is the point: would people be more accepting if the buyer of ARM was a mega data centre player like Facebook or Amazon? Softbank as mentioned is certainly not a conventional telco.
The markets certainly think it’s crazy – Softbank’s share price dropped by over 10% as a result!
The problem will be, SEPAM, what SoftBank wants ARM for. Does it really want to keep ARM exactly as it is or does it have an agenda? And, if so, how does that agenda affect ARM’s impartiality between customers?
Yes indeed Dr Bob, Ed would have loved this one
Wise and illuminating words, Alxlr8, I too believed that a large part of ARM’s value resided in its independence. Softbank seems to have been suckered and ARM gutted. Hermann Hauser says he regrets the deal. Everyone I’ve spoken to thinks it’s awful. So who benefits? Only Goldman Sachs and a bunch of advisors so far as I can see. It’s a sad reward for all the superlative efforts of those ARM design engineers over the years. And Yes there’s a definite whiff about it – maybe a rat maybe fishy.
Almost worthy of Ed
this is a sour deal as it is being presented here. I worked at ARM up until June 2014, and at the time the management were talking about how ARM had to stay independent in the interests of the industry. No company had to buy the company in order to access their technology – that was the beauty of the business model. Now, softbank are not buying ARM to access their technology, and they profess to have bought ARM in order to invest in it and grow its global presence while preserving everything about it. However, their strategy is ludicrous without taking ARM completely off on a tangent away from its power base within the semi IP business, and out into the long grass where they have little expertise.
On the one hand, they claim they will find another 1600 employees to bolster heads at the company in the UK alone. For what ends? If it’s operating within the UK semi industry, then where are these heads coming from, seeing how exceptionally difficult it is to find ANY home grown talent. This is an exceptionally hard sell, especially given how BREXIT is looming over everyone – a large proportion of talent comes from Europe and also overseas.
Basing future operations solely on the Semi IP industry, ARM already has the market saturated. A few more dots on the processor and GPU roadmap to satisfy a couple more nodes in the marketplace isn’t going to keep 1600 new employees busy, and isn’t going to bring in enough extra revenue to justify the ludicrous markup paid for a company already well out of its sensible valuation.
I smell a rat.
Yes it’s certainly better than a PE company, Jamo, and certainly no semi company could have invested much in ARM after paying such an enormous purchase price for it. It’s not inconceivable that Softbank will manage it well, but why oh why, did it have to be bought by anyone? It has done such a good job in supplying the industry with inter-operable processors this last quarter century, why risk screwing things up?
It’s preferable to a semiconductor or PE company.
Well I don’t know why the present owners would want to sell unless it’s for the cash, SEPAM, which is as good a reason as any. On the other hand it’s unpatriotic, detrimental to the UK technology base, problematical for ARM’s customers and staff and the end of a very effective example of the old inter-dependent electronics industry business model.
OK well maybe 2% is worth it, SEPAM, so long as it lasts – but will it? Will ARM’s major customers want to rely on a major component which is controlled by a major company? The last time this happened in the electronics industry was when x86 dominated the PC industry. Does anybody really want a repeat of that?
ARM has, at least so far, seemed like the paragon of stability. While I remain unhappy about the Piccolo story they have nevertheless built it brick by brick. In a world with negative interest rates, anything this stable would be a good deal.
Sure, there are competitors, like MIPS (with android support) and RISCV (with Linux and seL4 support) so alternatives exist. So they cannot be complacent.
The last CEO, Warren East, is now running Rolls-Royce, AnotherDavid, but ARM’s current chairman is someone I’ve never heard of. I see Goldman Sachs is behind the deal – another marriage made in hell.
Yes indeed Djonne, this is clearly an awful move for ARM’s customers, staff and the worldwide technology industry – bad for everyone except ARM’s shareholders, senior management and Goldman Sachs. ARM and Softbank seem to recognise this by announcing these silly guarantees – how can any responsible management guarantee to double the headcount in 5 years? Adding staff depends on expanding the business – and there’s no guarantee of that.
Absolutely Keith it’s a staggeringly stupid statement which only a politician grasping at any straw to bolster the case for British economic prospects, would make.
Bowers & Wilkins bought by a productless californian startup 2 month ago, and now ARM bought by a japanese telephone company named like a bank…
What’s next, BBC bought by Newscorp, or Monsanto maybe?
Isn’t the chairman also running Rolls Royce now?
I think HMG may have a golden share, but with the overdraft we have….
So sad, last one out switch off the lights and all that.
Not Rolls thank god, but he was ‘Group Chief Executive’ at Pilkington plc when it was acquired by Nippon Sheet Glass in 2006 – so he’s got form…
> Is it sensible to pay $32 billion for a company which makes a profit of $660 million?
This is an interest of 2%. That is not bad in this financial climate. ARM is 26 years old, too old for any dramatic changes which translates to an even more secure investment. From a purely financial investment point this is not too bad. With IoT (where quite likely only the chip makers will make money) to upside is even greater.
Softbank probably did a good deal here. Probably better than the present owners. And who were they??
I can’t believe the comment that Hammond made: “Britain is open for business – and open to foreign investment. Softbank’s decision confirms that Britain remains one of the most attractive destinations globally for investors to create jobs and wealth.”
What planet does he live on? This is not so much being open to investment as selling off the family silver…
This is deja vu from the time Japan seemed invincible. There are to my mind two problems here: one small problem in that a foreign company wants to buy ARM but a second bigger problem in that the present owners now want to sell.
The pound plunge should have alerted the present owners that ARM, exporting licenses, now was even more competitive. This nobody is talking about.
Also interesting to note Softbank states they are going to invest more. Why didn’t the present owners do this too?
My reaction too, Dr Bob, very sad
when i heard this on the news this morning i thought it was a strange joke
then the ‘oh shit’ moment hit
yet another uk company going!
An excellent point, mam, ARM would have cost Softbank $4 billion more at the pre-Brexit exchange rate.
isn’t it the first impact of the brexit – low GBP??