Report: ARM is for sale and Nvidia’s interested, Apple isn’t
arstechnica.com
arstechnica.com
Apple can design AND manufacture chips based on ARM architecture, essentially without any interference from ARM the company. Just pay a (relatively small) license fee and you're all set.
The article says it well:
> Arm’s licensing operation would fit poorly with Apple’s hardware-focused business model.
> There may also be regulatory concerns about Apple owning a key licensee that supplies so many rivals.
I think it's important not to confuse the success of a specific architecture, with the economic success of the company owning and licensing IP related to that architecture. These two things might not necessarily go hand in hand.
Edit: two very relevant comments in the parent thread, I'll just mention them here for completeness:
> the revenue of ARM before Softbank acquisition was less than the Net profits of Intel. (from https://news.ycombinator.com/item?id=23930895)
True. As mentioned above, architecture success != economic success
> Risc-V (from https://news.ycombinator.com/item?id=23930814)
Absolutely relevant. I am bullish on Risc-V and think there's a good chance we will all eventually converge to that.
Or in memespeak: scenes when nVidia controls Apple's CPU roadmap.
Historically I think they have just licensed the ISA and built their own thing on top. Maybe they needed the Big.Little stuff in the past but for the future they seem to be confident enough not to need anything from ARM or do it themselves.
You'd expect that ARM would be pretty responsive to the needs of one of its highest profile customers especially given the history and there is no reason why that would change with any sale.
Sure, Apple has escape routes if things go wrong but it's very very unlikely they will be needed.
So they probably could but whilst the relationship is good the best of all worlds is a strong ARM team plus a strong team in-house.
Even Qualcomm?
Why would they? With Intel, AMD, IBM with PowerPC, etc. CPU ISAs are not a monopoly.
Whoever buys ARM might want to do that to start cranking up licensing fees, and might not give a damn if this results in customer loses.
ARM was founded as a joint venture between Apple and Acorn, and Apple funded, helped design, and built their first joint ARM chip (the 610), and the ARMv3 architecture. I'd say they have a special relationship. They're the first licensee and like all the big names (Intel, Nvidia, Qualcomm, Samsung, etc.) have a far more permissive architectural license that allows them to extend the ISA or modify ARM cores as long as they comply fully with the ARM architecture. They also hold a lot of related patents.
So Apple is the last company that should worry about licensing (nothing to lose), or be interested in buying ARM (nothing to gain).
It doesn't seem to have hurt Apple in any of their past transitions.
The fact that Apple could improve their vertical integration by switching to their own silicon, doesn't mean that they need to control every single aspect of it. That rarely happens.
Also, any move by Nvidia to specifically hurt Apple would be not only detrimental to their own business, but likely an easy target for a lawsuit.
But due to regulations around monopolies they likely will have problems if they end up with this after buying ARM.
It also isn't as profitable as Apple, so Apple stack holders might not like it. Even more if Risc-V takes over things might not be that good for ARM so it's another liability for Apple with like not that much benefits.
ARM is enormously economically successful, they're just not a manufacturer and thus not remotely a comparable business to the entirety of Intel, fabs and all.
Apple has a perpetual license from ARM, so it doesn't matter who buys ARM, Apple is all set license wise.
There is a risk that a new buyer might under-invest in the evolution of the architecture and future core designs. This is a small but real risk for Apple because they do base their cores on ARM designs. I've seen the argument before that the only thing Apple uses is the ISA, but that's flat out not true. They customise and tune the design sure, but the basic architecture comes from ARM. The secure enclave, for example, is an ARM design. Lack of investment in that core ARM architecture would be a problem.
On the other hand, it would equally be a problem for all the other ARM licensees, many of which are current direct Apple competitors. For this reason such a risk seems manageable from an Apple perspective. They are in a much better position to compensate for under-investment by the owner of ARM that most of their current competitors.
The exception in Intel, because Apple is soon be a direct competitor of theirs in the desktop and laptop CPU space. Hence my use of the work 'current' in several places above. Ultimately though, Apple seems to be on top of their CPU design game. They're already beating the living pants off their competitors on mobile, no threat there whatsoever, and very well placed to beat Intel on desktop/laptop. Underinvestment in ARM would be in inconvenience, but not an imminent threat.
RISC-V is interesting, but I think there's a lot of wishful thinking happening around it. Most of the really interesting RISC-V chips include proprietary enhancements. Theres a serious tragedy of the commons issue going on there that is already a problem, and that issue isn't going away. The sheer amount of talent and money ARM and their licensee ecosystem, including Apple, Qualcomm, Amazon, etc are throwing at the problem simply can't be matched by the ecosystem around RISC-V.
100% of ARM chips on the market include proprietary enhancements. I don't think that has any bearing on the adoption of the core ISA, unless you can point to cases where a decision was made to avoid RISC-V over licensing an ISA extension.
I'm very hopeful for the mill. I think they have a good 15-20 years left to get their stuff together before we can definitively say they've succeeded or failed. Barring that, though, riscv is quite nice. 'Specially the vector instructions—so much nicer than simd.
Looking at moves by Chinese vendors and others, I'm going to hypothesize that ARM said to itself:
"We can't compete with 'free' and RISC-V is 'good enough' that its going to wipe out the commodity micro-contoller licensing stream that is over 50% of our revenue. That means we won't have the resources to invest in designing higher end features in the high margin parts and we're going to die. We should sell now, take our cash, and let the buyer eat the future losses."
I don't know of course. I'm an outsider like most people. But I have been part of, and then watching the semiconductor industry since about 1984.I was personally on the "inside" at the development of the SPARC architecture, and watched it grow, and die, as it was mismanaged by Sun. And I've been associated with American high tech companies for more years than I care to remember :-) and have seen a bit about how such companies survive, grow, and fail to grow. A common theme is that there is a "cash cow", a product that has exceptionally high margins relative to its "class" and that revenue funds the R&D that does the "innovation" work. These businesses live or die on finding "the next thing" before their cash cow dies. Failing to find that thing, as SGI did, as Sun did, and as I think Google is in the process of doing, starves the company of the cash to pay the innovators as they focus more and more on keeping the cash cow alive for longer and longer. The company essentially morphs from a mission of doing cool new things, to one that is preserving the 'one thing' at all cost. That thing eventually dies in spite of their efforts and what is left of the company is bought by a healthier company for its assets and maybe a couple of its projects that were cool, but not good enough to replace the cash cow.
If I am correct, this is the right move for ARM, and a really risky buy for anyone who would buy them now. A smart buyer, knowing how this will evolve, will wait for the desperation to set in so that they have maximum leverage in the acquisition.
And another thing that will be true (if I am guessing correctly here) is that RISC-V is going to really take off and have lots of people making them for cheap.
And MIPS never went anywhere because it never cracked the cellphone market (which is where the real volume is).
When someone puts out a RISC-V phone, you've got a battle on your hands. Until then, you're fighting over table scraps.
Obviously ARM is much further along in the practical implementation of these features.
RISC-V is a serious threat to their business model. In a way this is good, CPUs and the surrounding IP should be a (free) commodity, just like operating systems and utilities.
I'm with you if this means that CPUs should be a collaborative effort with shared overheads and common standards but ARM is pretty close to this already.
Which led me to speculate the other day that mobile phones should be programmed in lower level languages for an immediate jump in battery life. Nothing my phone does requires gigahertz processors.
Chinese firms may rush to RISC-V but given the geopolitics (e.g. Huawei) I can easily see barriers being put up in the US and elsewhere to cheap chinese RISC-V chips being used everywhere.
Western Digital adopts RISC-V [1]
Microchip's new push [2]
Analysts covering the microprocessor marking making some pretty big projections [3]
[1] https://riscv.org/2017/11/ee-times-article-risc-v-spins-driv...
[2] http://ww1.microchip.com/downloads/en/DeviceDoc/00002911A.pd...
[3] Semico Research estimated that there will be 62.4 billion RISC-V cores shipping in 2025. It’s not just for microcontrollers, Asanovic said. There is demand for every performance level. -- https://venturebeat.com/2019/12/11/risc-v-grows-globally-as-...
That doesn't mean that ARM 'is going to die'. RISC-V is not 'free' in the sense that someone has to design an implementation that works for you - and then support it. The economics work for WD but not necessarily for everyone - not everyone has WD's scale. Plus price isn't the only factor in choosing an ISA / implementation.
Licensing IP to put on SoCs has been common for over 25 years - in fact its the whole basis of ARM's and others business models.
Don't fear change, it's a cool thing.
* https://www.forbes.com/sites/greatspeculations/2019/09/24/am... https://wccftech.com/amd-radeon-and-nvidia-geforce-discrete-...
I understand that has historically been the case, but AMD is in a pretty good negotiating position right now, being effectively the only company that can provide what they bring to the table.
* a SoC with integrated CPU and GPU
* with high performance and decent power consumption
* and great synergy with the desktop x86 platform, in terms of developer knowledge, available tooling, and ISA compatibility with older games
I wouldn't be surprised if AMD is making a decent respectable margin this time around. And if not, what they most certainly did get, was a bunch of free R&D money for developing ray-tracing technology for the consoles, which they can now leverage through their entire portfolio.
As dralley mentioned. AMD is actually one of the few businesses with both strong CPU and GPU expertise. Building a console with any other company would require buying the CPU and GPU from two different companies. Intel would refuse and ARM CPUs including the ones Nvidia designed are not powerful enough, so really AMD CPU+Nvidia GPU is the only alternative to AMD CPU+AMD GPU but there is absolutely no reason for AMD to collaborate with Nvidia. The truth is that AMD is the only one that can serve this market.
And it looks like big tech is developing much the same rich-and-out-of-touch reputation that bankers and stock traders used to have.
I don’t know who, but I am assuming one of the FAANGs will be forcibly broken up in the next decade.
Privatizing the gains and socializing the losses is a playbook the tech giants are all too happy to keep pushing as long as the government allows them to.
"ARM is so widespread that buying it will be a regulatory nightmare, and even the most lenient rubber-stamp regulators around the world must shudder at the idea of an existing ARM licensee buying ARM. "
Practically all of Apple's product line will be powered by ARM in the near future. Buying ARM would put them at a huge advantage while heavily influencing the chip IP of competitors. I can't imagine regulators in the US and EU not raising an eyebrow at this.
Purchase by any ARM customer (including Nvidia) potentially reduces the value as it will alienate other ARM customers who compete with the buyer (e.g. AMD holds ARM licenses so how would it feel about using Nvidia controlled technology). Yet who else will be able to generate synergies from the acquistion?
Of course Nvidia may feel that the synergies are big enough to offset the loss of business plus its share price gives it a fantastic acquisition currency.
Plus, the line when Softbank bought ARM was that ARM as a listed company was constrained by lack of cashflow, so a deep pocketed owner would be able to take up opportunities not available to ARM as as stand-alone company.
I know there aren't tons of news articles about new high performance chips being offered but where I work we are seeing a ridiculous level of interest from other companies particularly given our industry...
Currently it looks like SiFive was borne out of a solution for the organizational issues that plague innovator's dilemma companies. I am envious of SiFive's position right now. I hope they don't get swallowed by a dinosaur.
The only reason NVidia wants ARM is because they can't have an x86 license.
Apple doesn't care because they probably have an escape/hedge clause on their ARM license and they will soon be shipping an IR. They have ISA hopped so many times it is old hat to them.
Currently though, RISC-V already is ingrained heavily in the embedded and IoT space and growing astronomically. That's where ARM is losing heavily, although in my opinion, the embedded space wasn't the money maker anyways for ARM
My understanding is, they design the specification of the ARM assembly language, and the design of the hardware (not individual chip fab but like... abstractly, somehow?)
But if they specify the design of the hardware, why would one chipmaker have faster/better CPUs if they are using the same spec as another chipmaker with slower/worse CPUs?
What's the business model? Do they get paid per ARM chip sold, or yearly for the rights to sell ARM chips, etc?
- The ARM ISA (the "specification")
- A series of chip designs implementing such ISA (the Cortex series)
Customers can either license the Cortex design to implement in their own silicon or license the ISA and design their own CPU (like Apple did). There isn't one single Cortex core, but many of them at different performance/price points
What's the IP that Apple would otherwise be infringing upon if they didn't license? The "ARM" trademark?
Intel & AMD have a cross license, x86-64 was actually created by AMD. And before then, (ie, in the 70s & 80s) Intel was forced to license to third parties to qualify for defense & government contracts.
But re: patents, in the case of Cyrix:
> Focused on removing potential competitors, Intel spent many years in legal battles with Cyrix, consuming Cyrix financial resources, claiming that the Cyrix 486 violated Intel's patents, when in reality the design was proven independent.
[And this is despite Cyrix having white-box reverse-engineered Intel's chips to figure out how to be software- and socket-compatible with them! That seemingly didn't matter, as long as in the end the design they actually put in their own chip wasn't encumbered by Intel's patents.]
> Intel lost the Cyrix case, which included multiple lawsuits in both federal and state courts in Texas. Some of the matters were settled out-of-court and some of the matters were settled by the court. In the end after all appeals, the courts ruled that Cyrix had the right to produce their own x86 designs in any foundry that held an Intel license. Cyrix was found to never have infringed any patent held by Intel.
With that case-law in place, it sounds like anyone demanding that third-parties license their ISA these days is effectively bluffing.
AMD and Intel license IA-32/AMD64 and extensions from eachother, and they have agreements which make this mutual licensing apply to basically any ISA extension. The original x86 patents are long since expired, but every year they add extensions and cross-license them to eachother.
Both. Basically at the top end You have licensing that big companies like Apple pay a large one-off sum ( comparatively large from ARM's perspective , but tiny in Apple's view ) for the use of ARMv8. ( i.e You will need to pay again for ARMv9 or ARMv7 ), and do custom design with the Instruction Set as along as it remain compatible with ARMv8. At the bottom end you are simply buying a blueprint from ARM, add some other IP you want and pay per ARM chip.
Anandtech has an article [1] about it and most of it are still relevant.
Since we are taking about business It is also worth mentioning the revenue of ARM before Softbank acquisition was less than the Net profits of Intel.
[1] https://www.anandtech.com/show/7112/the-arm-diaries-part-1-h...
Keep in mind that I am NOT well-informed about this and I may say some wrong things.
https://www.arm.com/why-arm/how-licensing-works
From what I understand, ARM will give you various different implementations of different ARM cores, depending on the licensing terms. My guess is that this includes:
- RTL optimized for simulation
- RTL optimized for synthesis, maybe with variants for FPGA vs ASIC
- Various lower-level implementations, complete with layout. Just a “block” that you drop onto your chip design. I imagine that these are something like “here’s a block that you can give to TSMC for their 16nm process”
> What's the business model? Do they get paid per ARM chip sold, or yearly for the rights to sell ARM chips, etc?
Either or both.
That's not the interesting thing, really. ARM licensees don't compete on how fast the ARM itself is, they compete on what else is on the same chip that they've designed themself. Radios, high-speed serial interfaces, analog-digital, etc.
(the big exception is Apple, who have their own silicon team improving the ARM implementation)
ARM have a very wide range of price/performance/area/power options. You don't always want the biggest and best.
The IP itself can be placed into semiconductor fabric and combined with many other parts such as peripherals, flash memory, sram, etc.
However, if you look at stock as its own currency, basically dollars*managementskill, putting a dollar into Nvidia or Tesla basically means "if I give this money to someone else, they can do better with it than if I hold onto it. It would also be a way of "investing" in a management team, helping support them financially so they can go out and make acquisitions.
Then Nvidia and Teslas stock price becomes less "a bajillion times their next five years of earnings" and more "what could these great teams buy or build if we pump them full of cash." Youre investing more than just the future of the current company, as it exists right now.
I don't see Softbank wanting Nvidia stock though, unless they intend to liquidate it quickly.
Cheap RISC-V CPUs seem to have a much brighter future than ARM for consumer appliances like washing machines than have slim profit margins and don't need compatibility with other devices.
Do you think ARM will fortify it's position in the smartphone/pc/server space and surrender it's traditional stronghold of embedded to RISC-V?
ARM should fit well with them, plus it would give them a way to enter the mobile space again, this time by owning the IP and not making hardware.
But many reasons Apple buying ARM is not a good fit for Apple.
Nvidia may well be good fit and a better fit than Apple by far, though would it be best if ARM is owned by some company or publicly listed?
If I was Softbank, I'd go for IPO route as I do feel out of all the IPOs, this is a mature tech firm. That makes this one that will be popular on many levels, let alone the number of Geeks who will want an ARM share certificate upon their wall.
ARM is definitely mature, it originally IPO'd in 1998 and remained public until it was bought by SoftBank in 2016.
Whoever controls that company can decide who's allowed to make processors that run consumer facing software (sure, RISC V might present a threat in the long term, but it's a good decade out).
Of course that would be a nightmare for every other licensee (which is why I'm hoping for the IPO), but the potential gain is immense, isn't it? Or are the licenses so nice they're safe regardless of who owns the company?
reminds me the old Eric Schmidt quote about how YouTube was cheap because it was expensive - had a ton of legal liability and a ton of engineering work required to meet the needs of rights owners
Getting a new home from non-neutral buyer might cause licensees to start looking for alternatives.
I realize all the tech companies in the running and ARM are global brands. Recently T-Mobile and Sprint were allowed to merge in the US. Is it really that certain that Apple or another giant purchasing ARM wouldn’t be allowed?
The Mellanox deal was delayed for months by the Chinese competition authorities, and ARM is a far larger and more important player in the space.
You can imagine what might happen if an anti-competitive player controlled the major competing architecture to x86.
"He who controls the spice controls the universe." etc, etc...
Even your toaster and fridge probably have ARM licensed designs in them. Your Smartwatch certainly does, and the billions of non-Apple smartphones too.
> Chinese lead by Huawei are forced by US sanctions also to create their own design
Pretty sure they were heading that way even without US sanctions. Steal the IP, then make a competing product while paying zero licensing fees to ARM.
Huawei would be in a better long-term position if they chose RISC-V versus IP theft.
It's not trivial to build a CPU. Particularly if there is no existing industry with expertise in this field within your country.
The core instruction set is from the 80es. Surely you can come up with some better today were you to design it from scratch?
Ya, but then you have to develop and support the entire ecosystem around just the instruction set. That means compilers, languages, runtimes, and more - which is not a trivial undertaking.
Perhaps Apple could pull it off, but it would take a very long time to get something entirely new stable enough, and performant enough, to compete with anything being offered today.
Plus, it's not Apple's focus and doesn't really benefit them much either. Nobody builds iOS apps in assembler - but they expect it to run fast. That takes a very well tuned compiler, and a very well refined CPU architecture.
It isn't "mostly the same instructions but with 64-bit registers" the way 64-bit x86 is.
(Also, it's not like ARM was a single instruction set before either. For quite some time ARM had two totally different instruction encodings, the "usual one", and Thumb2, with microcontrollers usually only supporting the latter.)
Why not use ARM if the alternative is to build an entire software stack from scratch. ARMs licensing fees are very reasonable, and then you get most of the free software ecosystem for free.
I am surprised that Intel in not in talks as a bidder.
I'm surprised too, but can't help thinking Intel would be the worst possible steward of the ARM ecosystem. Back to the same-old anti-competitive Intel practices of the 2000's...
For what it's worth, I kind of have the same feeling about Nvidia too. I think the best stewards of ARM IP would be an organization that is fully invested in seeing ARM dominate... not an organization that's just looking to add ARM patents to it's warchest.
Intel would a) not be allowed to buy for very clear antitrust reasons b) as owner would be very likely to be a disaster for both Intel and ARM. Intel has had ARM licenses in the past and has sold the relevant businesses.
On top of that, their mobile business lines are and have always been weak.
No wonder everyone is jumping the ship over to RISCV.
Soft Bank did not do its due diligence, did anyone at SoftBank hear of RISCV before buying ARM in 2016?
The writing was on the wall since RISCV ISA was made open source.