The problem with the original SoftBank deal, and now the Nvidia deal, is that these companies will simply hollow ARM out (slash R&D, cut staff etc) to pump up the target’s value, before flogging it off or breaking it up. The stock market doesn’t value synergies, although it is quite happy to punt this aspect when M&A deals are in the offing. Given its strategic position as the defacto standard for mobile comms, that’s a serious threat.
ARM also gives the UK strategic leverage in any trade negotiations. Would we knowingly allow Rolls-Royce Jet Engines to be flogged off? No. Ditto ARM.
So, I think this is the right policy and hopefully the government will see fit to step into other deals like this and block them.
I’ve been through a PE acquisition. We were stripped bare and bits sold off before finally getting our mojo back. I benefited handsomely but we were not the company we were before and the job losses in the process were huge. It will be no different with Nvidia’s proposed take over of ARM.
(disclaimer: I was a semiconductor executive in a global MNC).
That's a very real concern with Softbank (and why I questioned why regulators were allowing the acquisition in the first place). I don't think you can be any further from the mark implying that's Nvidia's plan. Nvidia wants ARM to flourish, they would have no interest in killing it or selling it off once acquired. They already use ARM tech in their Tegra CPUs, and in the "Smart" NICs they acquired with Mellanox. Additionally they pre-announced a datacenter focused ARM chip just last week.
They would have absolutely nothing to gain and a ton to lose if they were to kill off ARM or gut the technical talent.
The ecosystem business model that defines Arm simply doesn't have any space for them being perceived as biased this way or that way.
I would imagine plenty. Take a look at the list of vendors that sell Nvidia GPUs despite the fact that Nvidia sells directly into the market.
>The ecosystem business model that defines Arm simply doesn't have any space for them being perceived as biased this way or that way.
I don't know, Intel's recent announcement that they're going to start licensing their IP directly contradicts that. I don't think Intel would waste what will likely be a fairly large amount of money spinning up a new business unit without first doing some market research and ensuring there's demand for it.
We have people licensing Arm and building interesting CPUs (Ampere).
I'm not sure that the NVIDIA acquisition would be good for the market or for Arm.
> I'm not sure that the NVIDIA acquisition would be good for the market or for Arm.
NVIDIA works with Ampere and just started shipping server nodes with Ampere CPUs.
If that was bad for Ampere, they would not be working with NVIDIA on this.
Not that you are right/wrong (I don't know), but Ampere appears to be an example that shows the opposite of what you are trying to claim.
Intel once was in a similar position to Nvidia and everyone hated them then. The failure of their last few generation has meant they don't have as strong a monopoly position and so have to start taking a more conciliatory position with the market. I don't know if they're succeeding here but it's obvious they've been "humbled".
Now, maybe Nvidia will treat ARM completely differently than it's GPUs. But companies tend to have single cultures and so it's easy to imagine Nvidia squeezing dimes from other ARM customers, where ARM has been at the opposite side of the market, reasonably priced, good-enough tech. Charging a high premium on your capital investment works for a market leader, that uses the money for more investment. But for a market followers, the same strategy results in market share shrinking 'till the whole thing dies (not that this isn't a way to make money but when the thing dies, some people are unhappy).
Adding to that Intel didn't exactly pick this model as their first choice. The combination of a wafer production shortage and Intels CPUs underperforming is what drove Intel to this. It is neither Intels or most feasible clients first choice.
Intel has done lots of things that it turns out people didn't want to do (with them).
I can't tell if you're saying you have insider information or you're just assuming they're getting customers with no evidence.
And Annapurna (owned by Amazon) sells CPUs to all sorts of vendors, including Ubiquiti and Synology.
> which customer would prefer to buy IP from a potential competitor rather than from a neutral provider.
The same customers that buy NVIDIA GPUs, NVIDIA interconnects, support NVIDIA's software, etc.
You seem to be suggesting that this "somehow" matters, but are not explaining "why" do you think this matters.
AFAICT if AMD wants to ship a data-center node, it needs Mellanox interconnect, and has to buy that from NVIDIA, pretty much in the same way that NVIDIA ships AMD CPUs with their DGX boxes.
There are dozens of thousands of pattents that all these companies need to buy from each other on a regular basis to function.
I don't see how this particular change makes the status quo worse.
Not as low power as mobile phones, but not completely far away either.
Until now the competing products of NVIDIA and Qualcomm differed in NVIDIA having worse CPUs and better GPUs and Qualcomm having better (ARM-designed) CPUs and worse GPUs.
I do not find it believable that in the future NVIDIA would ever provide Qualcomm with better CPUs than those of NVIDIA products.
Given NVIDIA's desire of dominating the datacenters, it is also hard to believe that they will give to all other companies who will attempt to design server CPUs new ARM cores at the same time when they will be available for the internal NVIDIA design teams.
ARM does not make CPUs (it makes CPU IP), Qualcomm does.
The fact that NVIDIA builds bad ARM CPUs does not change with the acquisition as far as I can tell.
For that NVIDIA would need to invest a lot of R&D into their CPU team, and again, if it does that, it can compete with Qualcomm, independently of whether the acquisition takes place.
I think the overall trend is still accurate and in either case ARM + Nvidia is (or would be) a compelling provider of hardware but the segment of the market that needs/relies on X86 architecture does seem to be shrinking.
If a chip supplier like Nvidia buys ARM and uses its products, it will ultimately have an outsize say over ARM’s product (actually IP) roadmaps. That’s not going to make ARM’s current customers happy. Nvidia want to diversify- that’s why they’ve targeted ARM. They don’t need the technology- it’s readily available to licence.
Arm is valued at $40 billion on c. $ 2 billion in annual sales while Nvidia is valued at > $300 billion on $ 16 billion in sales. With a valuation of 300 billion, why buy ARM and almost certainly damage the ARM business model which is as an independent processor IP provider?
And how will Huawei (and China), big ARM licensees, react to ARM being owned by a US company given the bad blood in that area.
For ARM, it’s a strategic disaster IMV. Better to stay independent!
That's the 40 billion dollar question, isn't it. Why are they doing this?
I don't know, I guess they know something that I don't.
What do you know that we don't ? :D
> For ARM, it’s a strategic disaster IMV. Better to stay independent!
How do you arrive at this conclusion?
For all we know NVIDIA might (1) invest a lot more money into ARM, (2) improve their ecosystem, (3) start licensing their IP to third parties as well using a similar model to ARM, (4) start contributing more to open source like ARM does, etc.
The real issue here is Nvidia are hugely overvalued thanks to the vagaries of the US tech stock market and the ‘irrational exuberance’ that surrounds it. Just a short while ago, Tesla was valued at over $800 billion, Apple has topped a trillion and now Nvidia on $16 billion sales at >$300 billion. As a result that have a while loymtvof stock value burning a hole in their pocket and are looking for something to buy. Unfortunately, ARM is in the frame.
When you make an acquisition, you have to ask whether it positions the combined business uniquely, will is cannibalise existing business (in this case almost certainly because ARM’s great appeal besides its IP and eco-system is that it is independent), will it move the stock price in the right direction, will it improve EBIT, what are the synergies etc etc. A typical M&A will invoke dozens of considerations like this.
More investment into ARM? They own the mobile phone space - Apple, Samsung and almost every other brand (many in China where I worked for a few years). They’ve seen off Microchip in the general purpose 32 bit uC space and dominate IoT applications, Bluetooth etc.
There’s nothing stopping Nvidia copying the existing ARM model right now. They don’t need to buy ARM to do that.
Nvidia stock price will moderate down (as will all semiconductor companies in the coming months as the chip shortages abate) and with it a more realistic valuation.
(What do I know? Been involved in a few M&A’s. ;) )
If this was a sale to a PE firm I think your predictions would be more likely, but Nvidia seems really unlikely to be a “slash costs and sell it for parts” type of buyer.
From what I can tell, there seems to be increasing consolidation in the industry:
* AMD -- Xilinx -- ATI
* Nvidia -- ARM -- Mellanox
* Intel -- Altera -- Habana/Nervana -- Mobileye
The supply/value chain is broken up into little chunks with specialists focusing on a narrow activity eg chip design or fabrication or assembly or test etc. There are a few specialists like AD and TI that are still vertically integrated to some extent eg design and fabs, but that’s because they have specialty analog processes for things like opamps , references and so forth. However, a lot of their assembly and test is sub contracted out. For digital devices and companies like ARM, very few still run fabs (semiconductor speak for ‘fabrication plant’) - they leave that to giant fab only companies like TSMC, UMC et al. How this works is the fabs run standard processes and companies design products to be produced using those processes - so by standardising, costs are reduced. A fab process takes 2-3 yrs and a few hundred million $ to develop. Samsung and Intel still have fabs doing predominantly digital, but they have the scale and heft to afford the $1-2 billion every 7 yrs to invest in new plants. There are few others though ( mainly in memory).
e.g. TSMC <- ASML, and the raw materials providers
What's the rationale of this claim? (you don't say)
I'd expected that ARM CPU vendors are interested in having "great" CPUs to sell and having as big of an ecosystem as possible.
So my expectation would be that NVIDIA would significantly increas - not cut - ARM's R&D budget, and will significantly invest in the ARM platform.
On top of this, NVIDIA apparently is trying to cut its reliance on Intel and AMD CPUs. I don't see how cutting ARM's R&D budget and stripping it for parts would allow them to reach that goal.
This is the opposite of what you claim.
You keep saying this, but this still does not match the reality that Nvidia has bid a lot of money for ARM.
So there must be something "in" for them that makes it worth it for them to pay 40 billion dollars.
The claim that this deal will be a disaster for ARM is IMO equivalent to saying that "killing ARM is worth 40 billion dollars to NVIDIA".
That makes no sense, at all. The only reason ARM is worth 40 billion dollars is because of its users. If ARM loses its user base, it is worth as much as MIPS or OpenPOWER (aka 0 dollars).
Also, if ARM loses its user base, NVIDIA ARM CPUs become worthless.
I have a hard time believing that, after investing 15 years of R&D into their own ARM CPUs and finally shipping serious ARM CPUs to market, NVIDIA would pay 40 billions to make that all worthless.
If so, then it seems like the right thing to do would be to nationalize. If not, was it just a leadership decision for purposes of expanding the company?
It’s not the company that gets the money - it’s the owners. This transaction is about SoftBank’s balance sheet, not ARMs.
Credit to Wisconsin for at least pretending to hold them somewhat accountable by cutting off the gravy train.
What does "flogging it off" mean in this context?
sell or offer for sale.
"he made a fortune flogging beads to hippies"
Softbank is japanese, NVidia is American, ARM is british.
Also, the UK can probs stop any transfer of ARM IP/Company/Etc. if they wanted (idk the laws tho).
Not quite a flogging off, but the MiG 15 engine was a copy of the Rolls Royce Nene engine which a soviet trade delegation won in a bet on a game of billiards.
Admittedly, that was 1946, before the cold war really began
Given that you were a semiconductor executive in a global MNC, may I ask, How?
Now correct me if I am wrong. The whole ARM business model was built on the foundation they dont and will not / could not squeeze you out once they dominate. They are British, not acting like some Giant Tech companies you have across the pond. You also get the option to buy perpetual ARM license in various form, including ISA and Design in case you are not comfortable. i.e There is very little they could do to pump up their value. Cutting Opex doesn't do that.
The whole business model is so well designed to the point it doesn't make sense not to sell to SoftBank when they put in that ridiculously high offer. (Purely from a Financial perspective)
As with the case of TSMC, both ARM and TSMC were not supposed to be in the spotlight. They are the backend enabler of the industry. And they want their customer to get the attention and focus, while they are doing the old boring work. All of that changed when Apple jumped into the picture.
This seems dramatically more likely from a generic investment consortium like SoftBank versus an actual chip-focused R&D organization like nvidia. I don't see how this is possibly a counterargument for sale to nvidia.
Doesn't this/shouldn't this basically encourage countries like the United States to completely ignore intellectual property laws? If not, why not?
A better argument would be that Nvidia gains serious power over companies who are currently ARM customer and Nvidia competitors.
Perhaps the UK want to be able to use ARM designs for domestic chips in the future, without being dependent on an American company, but yeah, then they should have objected when a Japanese company bought ARM as well.
Also, as others have noted, it's quite possible that GHCQ has backdoors into ARM designs that would be jeopardized.
But despite being owned by Softbank, Arm Ltd. is still incorporated in the UK and has their headquarters there. If chip design is seen as a essential national capability, the national security argument is valid regardless, in my opinion.
Seems to be the strongest/easiest route to go to justify blocking it. The whole national security angle and all.
> then they should have objected when a Japanese company bought ARM as well
I have to imagine this was a bit of sleeping at the wheel. I'm not sure though.
Would love to hear your thoughts on the topic.
We've seen an enormous hollowing out of institutional competence since the rise of Free Trade At All Costs.
What makes you think so?
So Intel - US / TSMC - Taiwan can be considered as national security, but ARM not?
I hope that makes sense - at least where I was trying to go with that comment. Maybe it's incorrect or not well thought out but I'd at least like to make my point clear :)
If Nvidia were to relocate the brainiest parts of ARM to the US then the UK would lose a skillset that is arguably relevant for national security.
So I don't think this intervention is a definitive no to the acquisition. Rather, it appears to be a way of entering into negotiations.
It's probably something where the government wanted to stop this deal for a number of reasons, and "national security" was the easiest justification where they already had a legal apparatus in place for doing so.
Trump did a brilliant job :)
I think we are on transition slogan number 3, quite the compensation to being cut off from your customers for many businesses I'm sure.
Edit: Is the HN hive telling me Brexit is a good thing then?
Among articles from the time, https://www.forbes.com/sites/parmyolson/2016/07/18/arm-softb..., complete with government spin.
It's hard not to make something about Brexit, when the government explicitly chose to make the thing about Brexit. The Softbank deal was disastrous for UK tech, and if the government really was interested in national security grounds (which is what they're claiming this time around), it would have intervened in that deal. NVIDIA isn't likely much more of a risk from a national security perspective than SoftBank. But it had to sell Brexit, and a massive foreign investment in a UK company was an opportunity for spin too enticing to give up.
On the face of it the difference with Softbank is that Nvidia is a big semiconductors player and thus that this acquisition can have a material impact on competition in the industry and the UK, which was not the case with Softbank.
In addition, since the UK is now "properly independent" the government is probably keen to show it by doing what the US and EU competition regulators do.
The horse has literally bolted. Shutting the door now does literally nothing other than show you cocked up by not locking it in the first place.
You can't prevent someone to sell something if it doesn't want to keep it.
I believe HMG didn't do anything to stop Astra-Zeneca being sold out of the UK a few years ago, and it's just luck it didn't go through. They probably see the need now.
> The Competition and Markets Authority (CMA) will prepare a report on the competition and national security aspects of the proposed transaction.
So it's being framed as a national security threat, interesting.
But such is policy making. You have to justify your policy with what is popular, not with the truth. For example: lax covid restrictions justified because we can't be sure it'll work, vs. because we know that it'll hurt the economy.
Having access to the fastest chips is a matter of national security, at least for the US. The AI race (even if it's just faster NN processing) has massive implications for modern warfare and national security.
"In the UK, there are 13 national infrastructure sectors: Chemicals, Civil Nuclear, Communications, Defence, Emergency Services, Energy, Finance, Food, Government, Health, Space, Transport and Water. Several sectors have defined ‘sub-sectors’; Emergency Services for example can be split into Police, Ambulance, Fire Services and Coast Guard.
...the loss or compromise of which could result in:
a) Major detrimental impact on the availability, integrity or delivery of essential services – including those services whose integrity, if compromised, could result in significant loss of life or casualties – taking into account significant economic or social impacts; and/or
b) Significant impact on national security, national defence, or the functioning of the state.’
Secondly, Nvidia will probably restructure / absorb ARM. At the moment its corporate structure is mostly intact
Before the Arm acquisition materialized, Nvidia was quite present in the RISC-V space. All current Nvidia cards have at least one RISC-V chip on it.
So while their intended high performance use cases are not feasible yet, this could lead to Nvidia focusing on RISC-V again the medium term, which would definitely be beneficial.
Of course Nvidia can still pursue ARM based designs without owning the company.
But the British government has experience, from the Kraft purchase of Cadbury, that even if the buyers promise to keep jobs in the UK and product quality high, once the deal goes through they can renege on those promises instantly with no negative consequences.
Also nvidia will build a new system in switzerland with 20 Exaflops (albeit fp16) which would be the fastest ML/AI system so far.
FTA: With the help of tight coupling between NVIDIA CPUs and GPUs, Alps is expected to train GPT-3, the world’s largest natural language processing model, in mere two days.
'National security grounds' is interesting. It could simply be that the UK is no longer part of the EU and this is belatedly considered a strategic national asset (i.e. a foot in the game of chip design), vs. GCHQ and ARM working together somehow.
I can see a future where Nvidia buys ARM, becomes more hostile to licensees, and the door is left open for something like RISC-V to gain more market relevance.
I'm guessing though that Arm collaborates with UK intelligence agencies and that that collaboration might be jeopardised with the Nvidia takeover (in a way that it was not with the Softbank takeover). Does anyone have any further insights?
I think it's a much more specific security issue.
I'm not sure about the second point but it seems premised on withholding IP from US which seems unlikely.
Much more likely that there is intense collaboration between say GCHQ and Arm (maybe both ways -> GCHQ gets to understand modern CPU security issues from Arm and Arm benefits from GCHQ input on threats). If that relationship swaps to NSA then there would be real damage to UK security apparatus.
Infineon (German owned) and Nexperia (Dutch owned) both own fabrication plants in the UK, so it's possible to scale that up if we were at war, the knowledge is here.
If you needed them for defence I'm sure you could set up a fab on older process nodes fairly fast. A tank consumes and generates so much power for moving that it doesn't matter if the CPUs in it will last only 2 hours in a cell phone...
(Speaking as a Canadian who has watched 30 years of continental economic integration and heard lots of friendly talk since the campaign for the 1988 Free Trade Agreement but now back of the line waiting for vaccines imported from Europe because the US won't export to us and our domestic manufacturing was sold off decades ago because of said continental integration.)
Also the US itself has used the "national security" excuse for trade issues multiple times over the last few years where it was seemingly not the case (steel & aluminum trade disputes, etc). So the precedent has been set. And there's little sign of Biden really doing much to reverse that.
That said, what Trump pulled with steel and aluminum is BS.
Seem to recall some very agitated US commentators bashing Canada when we chose not to enter the second war in Iraq. How's that working out for you, BTW?
ARM is a piece of infrastructure now, and dangerous in the hands of NVidia for other reasons than spying. Mostly economical. And yes, that's also national security.
From now on, every ARM processor costs $40 per chip to license. Except, of course, those made by NVidia.
From now on, all improvements to the reference ARM architecture will only be available in NVidia chips for the first two years. Afterwards they'll be available for licensing at the current costs.
Those are monopoly issues. And now a national security issue:
From now on, $RANDOM_EUROPEAN_COUNTRY is under embargo wrt to US intellectual property. NVidia being an US company, that includes ARM reference designs. Basically what happened to Huawei with the OS, except now it includes the CPUs too.
And I could go on...
I think you're overdoing the last point: there is lots of key technology licensed from US firms and no-one in the UK thinks that this is a major national security issue.
Much more likely that there is a specific issue that UK government has identified that poses a realistic problem for UK security services.
"says" yes... "thinks"... you don't know. I would think it if i were them.
Insofar as ISAs matter for performance, RISCV is not as good as ARMv8 since the designers are too academic, but it might be good enough.
given that, it's entirely reasonable that the UK doesn't want ARM gutted and the IP shipped out of the country.
> The proposed transaction is subject to customary closing conditions, including the receipt of regulatory approvals for the U.K., China, the European Union and the United States. Completion of the transaction is expected to take place in approximately 18 months.
Any one of these 4 regulators can still stop the sale.
Nvidia don't want to buy a company if they can't sell to Chinese market.
NVIDIA however clearly know that ARM represent market domination and would be a national strategic asset.
the creation of a relevant merger situation as defined in section 23 of the Enterprise Act 2002 (“the Act”), in that:
(a) two or more enterprises will cease to be distinct;
(b) in the course of the enterprises ceasing to be distinct, a person or group of persons will bring a relevant enterprise (as defined in section 23A of the Act) under their ownership or control; and
(c) the value of the turnover in the United Kingdom of the enterprise being taken over exceeds £1 million.
They said they're going to release material soon to justify their decision so we'll have to see what that all means.
Really hard to see this as anything other than a shakedown attempt by people who don’t understand that you can’t keep an ip licensing company inside national borders for long.
The correct move would be to nationalise ARM and build a fab, but the Tories are very unlikely to do that.
I'm not trying to say that ARM is incompetent, but it takes years for an institution to learn how to do this. ARM doesn't currently operate fabs.
You're correct that neither main party would do this, partly because of the inadequacies of the liberal election cycle. Only direct worker democracy could reliably centrally plan for use rather than profit.
- EU drops plans to license ARM and start to obviously fail to buid their own arch
- Softbank announces it want to get rid of ARM
- Nvidia want, UK agree
- Apple announces M1
- Nvidia announces Grace
- LG drop making phones
- China announces cpu independence
- UK oh noes, our treasure ! - suddenly "own" arch is good ? LOL
If it is just money for Nvidia deal need to be preserved.
But Apple and Nvidia are Intel killers now and noone saw potential in ARM ? More likely ARM wanted to ditch UK. Or rather top managers was talking with US buyers earlier... Like Altavista execs moved to Google after headshotting own company.
Anyway, whatever - let just x86 die. And pleas, pleas kill MS Windows too !! There is astonishing amount of money behind this madness. Not to mention "spyware" element in both.