ARM up for sale again?
Pages: 1 2
Doug Webb (190) 1158 posts |
Seems Softbank maybe considering a partial or full sale Anyone got a few pennies in a jar or down the back of the sofa? |
Alan Robertson (52) 420 posts |
Maybe ROOL can open a bounty. |
Jon Abbott (1421) 2641 posts |
I’d be surprised if many companies would be interested in ARM as it doesn’t really have a profitable future, no existing customer could realistically purchase it without challenge. Google/Apple for example would raise questions all round. All the key customers have licenses that allow them to spin their own design, if ARM went bankrupt tomorrow they’d be financially better off and could carry on designing their own chips. Just take a look at the route Apple are taking essentially redesigning ARM from the ground up. The switch from 32bit to ARM64 has proved the instruction set is irrelevant, smartphones have all switched with little to no impact to the market. This has essentially proved a switch to RISC-V is viable, provided the power target can be met. It wouldn’t surprise me if the Chinese purchased it to get their own back for the Huawei 5G debacle. Apple would be royally screwed! |
Rick Murray (539) 13806 posts |
This is something that happens when the core of the operating system is written in a high(er) level language. As long as the processor is capable, and there’s a competent compiler available for it, it doesn’t make any difference to 99.9% of users what the instruction set is. People don’t give a remotest inkling of a f… what actually happens inside. For all anybody cares, it could be a million tiny copies of Zooey Deschannel splatting different coloured ink at the inside of the glass, a hundred times a second, to make it look like moving images. Talk of stuff like processors and memory, these are only statistics, like 2GHz is better than 1GHz, four cores are better than one… and probably the only real reason that anybody even cares what the operating system is is because Apple have made themselves into the brand. They’re the exclusive ones. Android, on the other hand, is available from pricey flagships down to cheap knockoffs. And, yet, still nobody is going to care what the processor actually is, never mind something as utterly nerdy as the instruction set. ARM’s power is in consuming little power, and in having the major players licencing the designs and able to create their own custom versions (as Apple have, and likely Samsung too). This will help keep ARM viable because it’s a lot of expenditure and development to try to perfect those cores. Can ARM be replaced? Not yet. But if there’s something better, faster, cheaper, and more power efficient around, don’t delude yourself into thinking that the major players would not jump ship in a matter of days. Like I said, it just needs a capable processor and a competent compiler. What’s actually running underneath? Pretty much irrelevant in the long scheme of things. Kinda annoyed that Softbank didn’t understand what they were buying when they bought ARM.
We all know the Chinese are spying on us. I’ll leave this space blank so recovering-Steve can insert some other reasons why Cisco kit sucks. Either that, or go read The Register. ;-)
I doubt it. They already have a legally binding licence. There might not be any more ARM designs if China bought ARM, but I would imagine Apple can keep on developing the tech that they currently have as is provided in their current agreement. |
David Feugey (2125) 2709 posts |
While they have an architectural licence, they must pay royalties for each chip sold. If ARM disappears, it will be a nightmare for Apple. On the other hand, most of the business is for core licences. And ARM signs tens of them every quarter. The funny part, is that a lot of companies that have an architectural license (Samsung, Qualcomm, Nvidia, etc.) now just pay for core licences of the latest Cortex-A offers. For 64bit, don’t forget that the high volume markets (IoT, embedded, etc.) are still 8-16-32bit. But, yes, one day, all of this will stop. I don’t really care, because with multi GHz, multi core, modern I/O chips we already have something RISC OS will never be able to use fully. I guess a two core Cortex-A78 @3 GHz with NVME, 8GB RAM and dual 4K will be enough for all our dreams :) |
David Feugey (2125) 2709 posts |
With the Snapdragon 865, we are almost here :) |
Steve Pampling (1551) 8155 posts |
Chilling and not giving work much thought today. I got an appointment for an assessment of the state of my newly good eye and the not yet fixed left one today. August 6th, that’s two days after my current treatment episode is complete. Hope I can get the second op as swiftly as the first.
Good idea. |
Rob Andrews (112) 164 posts |
So no brainer then we need apple to buy ARM they have the money, they want to control there own destiny, would be a logical buy. Or maybe they will just wait until they can pick it up for cents on the dollar. |
Grahame Parish (436) 480 posts |
But what other potential licencees would trust ARM if Apple controlled it. It should never be owned by one of its users for that reason. |
Mike Freestone (2564) 131 posts |
except Acorn, of course! |
David Feugey (2125) 2709 posts |
Correct. |
Rick Murray (539) 13806 posts |
Ah, but then again it should never be owned by a hostile government. I fear the damage might already be done, with SoftBank failing to understand what it is all about (something of an MO for them, it seems) and, well, will it be unloaded on somebody who wants to keep the processors and architecture going, or will it be unloaded on some VCs looking for a quick return on any interesting bits that can be stripped out? Why was ARM sold to SoftBank in the first place? |
Grahame Parish (436) 480 posts |
Money? |
Steffen Huber (91) 1949 posts |
Because (nearly) every shareholder thought that Softbank was offering enough money per share? Over 40% over their valuation back then IIRC.
The indications are that Softbank knew exactly what they bought, and that they are currently in desperate need of cash, so selling ARM seems to be a sensible move. Maybe their only move to avoid bankruptcy? |
Steve Fryatt (216) 2103 posts |
And and the time of the sale, the weak Pound following the 2016 referendum made UK companies attractive targets for foreign buyers. A Brexit Bonus, if you will. |
Jon Abbott (1421) 2641 posts |
I suspect ARM was sold because of a lack of growth potential. Shareholders are primarily large pension funds, who demand an annual growth to increase share price and dividends. ARM does not have much growth potential as they pretty much have a monopoly in the mobile market and can’t expand into many new markets, certainly none of note. It wouldn’t surprise me if the large shareholders outvoted the in-house board members to force a sale. For large share holders, the sale to Softbank was a large return on investment that couldn’t be turned down. Sadly for Softbank, the deal wasn’t so great as they very quickly realised there was no growth potential and started looking at ways to dump it without taking a massive hit. Dumping a majority stake in ARM Chima was their only option as core ARM can’t be broken up. The writing is on the wall for ARM. With such a dominant position in the mobile space they can only lose market share. There are very few companies that would be interested in purchasing ARM as whoever purchases it either pays over market value or causes customers to jump ship. Apple’s problem now is they’ve committed to ARM for the long term, if they attempt to buy it, the sale will be blocked. If they don’t buy it and it’s sold to China, they run the risk of the US government putting ARM on the ban list. |
David Feugey (2125) 2709 posts |
While smartphone processors are about 78% of the ARM market in value, there are only 6,3% in volume. On some markets, ARM is not so present (networking chips, micro-controllers, datacenter/cloud, desktop). And some other have incredible potential (networking, automotive, IoT, consumer electronics). See slides 12 and 13. I’m afraid the situation is more simple: Softbank bought ARM because it was a good deal (thanks Brexit). And because the idea was to sell it to China later. And if you read between the lines, it’s already done:
They can also not lose market shares as Intel did for decades on desktop. Supercomputing is one of these: 152064 processors for just one cluster :) Except with some political issues, ARM is still at the beginning. |
Jon Abbott (1421) 2641 posts |
That financial projection very succinctly makes the point that in 2019 ARM had a 90% monopoly in two markets and has little potential for growth over the next 10 years, outside of markets where shipment numbers aren’t that high in comparison. Take Automotive for example, just how many ARM are needed per vehicle? They’d need a high margin per chip, which would not make them attractive vs alternatives. As for the misbelief that Softbank got a bargain due to brexit, that’s simply not the case. Softbank paid way over market value for ARM. Supercomputing? Just how much profit is there in the cores in a supercomputer for ARM? And how many supercomputers does the world need? Good for marketing perhaps, the fact the #1 supercomputer is current ARM based with 7.3 million cores, not really going to make a mark on the bottom line when they’re licensing tens of billions elsewhere. |
David Feugey (2125) 2709 posts |
They will. We speak here of partially (very common today) or totally automatic cars. The chips used are very pricey compared to traditional desktop chips.
You’re not fair here :) Embeded does not count because of low price and margin, and supercomputing or datacenters do not count because it’s low volume? Just do not forget that the smartphone market IS a low volume one for ARM. But with high margins. In term of value, I would say:
In term of volume, I would say:
Amazon using ARM. Not a mark on the bottom line. Combine all of this and ARM has probably more than 10% of the datacenter market today. The problem is that ARM is nowhere in IDC/Gartner, because these studies count what is sold. How much Amazon servers are Graviton? Good question. Since the price is the key, I would say much less than tomorrow. Let’s face it, the plan was to bought ARM then to make ARM China independent. Now ARM China has all the IP to make ARM processors for half of the world. The best part in term of growth potential (China, Russia and all the countries were patents are unknown or ignored). The ARM of today (the one owned by SoftBank) is already the ARM of yesterday. So they can sell it. Checkmate. |
David J. Ruck (33) 1629 posts |
Those Graviton chips are spectacular, version 1 was competative but version 2 blows the competiton away. It’s definitely getting ARM noticed in the data centre, and those chips don’t come cheap. So even keeping licences as a small percentage is going to increase ARM’s income many fold over smartphone type chips. I really want to play with a large Graviton2 instance to see how my multi-threading stuff scales (pretty sure its badly, going by other architectures), but I don’t think Amazon offer them on the free tier yet. |
David Feugey (2125) 2709 posts |
Yes, and there is a interesting phenomena around this. The only way to compete, is to use an ARM offer. And a lot of companies still want to compete with Amazon :) |
Rick Murray (539) 13806 posts |
There is a heck of a lot of MIPS in networking. My little router (not the Livebox, the other one). the 4 port switch, various (mini)PCI WiFi cards, my 480p webcam, the Vonets… mostly variations of Ralink (MediaTek) and Atheros devices. Old, but cheap and cheerful. MIPS is basically dead now, however 1. It’s history probably ought to serve as a warning, or parable, of ARM’s possible future… 1 Read the Wiki page for MIPS Technologies and marvel at the game of “pass the parcel”. It’s now owned by Wave Computing, a company so notable it doesn’t even have a wiki page. |
Rick Murray (539) 13806 posts |
Depends what you call “embedded”. To be honest, there are really two distinct types of embedded device.
To my mind, it isn’t really valid to refer to “embedded devices” as such, given the schism between the simple ones and the complicated ones. And this isn’t even to consider embedded devices outside of domestic use, such as traffic lights, car ECUs (simple style) and car entertainment systems (complex style), and the many types of embedded device used in the medical world, from a simple blood oxymeter (can be an ARM M0 core) to an MRI (probably uses many numerous devices within) 2. 1 Think that sounds crazy? Back around 2005 or so, there were many little battery powered MP3 players, which may or may not have had a little LCD (128×32 pixels being a common size). Typically with 64MB to 512MB Flash onboard, but some had more. They were based upon the S1 chip, which basically marries an 8 bit Z80 core (clocking around 24½MHz) to a 24 bit DSP. And when I say 8 bit Z80, I mean it literally had an addressing range of 64K; it accessed its 116K of SRAM and 29K of ROM using paging. Google for the ATJ2085 datasheet… 2 Want to have a crack at building your own desktop MRI scanner? This guy never finished, but it’s still fascinating to read. http://www.phys.cwru.edu/undergrad/Senior%20Projects/PreviousProjects/papers/papers2014/Rebec_Griswold.pdf |
Kuemmel (439) 384 posts |
Seems Nvidia is interested now to buy, see here Good or bad, I don’t know…for sure Nvidia’s graphics technology is the most advanced…but will everything be even “more” closed source then…? |
Rick Murray (539) 13806 posts |
How would that work? The ARM core is documented. The MMU is documented. A fair amount of standard peripherals on the SoCs are documented (though where they are in memory might not be?). The graphics stuff probably won’t be documented, but that’s the current situation. I suppose it’s better than “China” buying it, given the Americans are taking protectionism to new levels, and the newly liberated UK is so utterly beholden to America that they simply follow orders… |
Pages: 1 2