Friday, July 31st 2020

NVIDIA in Advanced Talks to Acquire Arm from SoftBank

It was reported last week that NVIDIA is "interested" in acquiring UK chip-design firm Arm from Japan's SoftBank that holds a treasure chest of tech IP. Now Bloomberg reports that things are getting serious between NVIDIA and SoftBank, with the two reportedly engaged in "advanced talks" over the possible acquisition of Arm by NVIDIA. The graphics and scalar compute giant recently surpassed Intel in market capitalization.

With a few quick moves, NVIDIA stands a real chance of displacing Intel as makers of the world's most popular CPU machine architecture, driven mainly by smartphones, tablets, networking infrastructure, wearables, and IoT devices. The Arm architecture is also taking strides into the server space, and Apple recently decided to dump Intel x86 in favor of Arm-powered homebrew SoCs. Arm could cost NVIDIA an arm and a leg. New Street Research LLP estimated Arm's valuation at USD $44 billion if its IPO took off in 2021, and as much as $68 billion by 2025.
Source: Bloomberg
Add your own comment

41 Comments on NVIDIA in Advanced Talks to Acquire Arm from SoftBank

#1
Caring1
And they can claim the China division of Arm as a tax loss.
Posted on Reply
#3
Vya Domus
I fear for a world in which Nvidia practically controls everything in the mobile/embedded space. Although probably their main intention is to bundle their GPUs with ARM CPUs, many have tried to make ARM happen in the HPC space few succeeded, somewhat. Not on large scales anyway.
Arm could cost NVIDIA an arm and a leg
So many jokes could have been made.
Posted on Reply
#4
trparky
I'm really not liking how this could play out. I already don't like nVidia, this would make me dislike them even more so.
Posted on Reply
#5
ilyon
I'm not sure this move is a great one, assumed the history of x86 patenting and the "proprietary vision" of NVIDIA.
Unless it's a "US move" to assume control of future standards.
Posted on Reply
#6
moproblems99
ilyon
Unless it's a "US move" to assume control of future standards.
I suspect this has much to do with it. Or a way to cut China off from it.
Posted on Reply
#7
moproblems99
xkm1948
That would allow Nvidia to reduce reliance on AMD or Intel's CPU and platform when designing their GPU workstation solutions. Good for them.

Meanwhile how the F did another tech article like this turn into another political BS spewing contest? Plenty of that political shithole on reddit you can vent off all your views. No need to pollute TPU with that.
Mine is not political. We got we deserved from China. I don't fault them at all. I just think 32 billion is a lot to spend when the market already said they didn't want their mobile gpus. It is also a lot spend banking on super computers. It makes some sense for AI I suppose but damn that is a lot.
Posted on Reply
#8
xkm1948
moproblems99
Mine is not political. We got we deserved from China. I don't fault them at all. I just think 32 billion is a lot to spend when the market already said they didn't want their mobile gpus. It is also a lot spend banking on super computers. It makes some sense for AI I suppose but damn that is a lot.
Both Intel and AMD are posed to have complete CPU + GPU + I/O integration. Being a bit weak on the GPU front.

Nvidia is at an disadvantage. They are unchallenged in software support and absolute compute power on the GPU front. They have recently acquired some high speed network chip maker I believe to complete their I/O part. Now they really need the CPU side to complete their integration. Yes ARM CPU will be weaker compared to AMD and Intel's offering. But it would gave Nvidia full package control instead of relying on say AMD's EPYC completely.

In an ideal world if Nvidia can get their hands on Cyrix / VIA's x86 it would be amazing. With Nvidia's R&D track record they would make some decent x86 CPUs.

Also AI and GPU computing is not a small market. If you sit through Nvidia's Youtube GDC presentation you will see there are loads of opportunities here.

I hope Nvidia's acquisition of ARM materializes.
Posted on Reply
#9
moproblems99
xkm1948
I hope Nvidia's acquisition of ARM materializes.
That may be. Let's hope this doesn't turn into NV's ATI moment. Though they seem more fiscally capable even though the ARM purchase is 3 times their yearly revenue.
Posted on Reply
#10
dicktracy
This is a monumental moment. x86 is dying and the only sanctuary is desktop PC and data center. Won’t be long till they move to Arm as well!
Posted on Reply
#11
PowerPC
I wonder how this will play with Apple moving to ARM. Apple doesn't like Nvidia, I think it clearly showed when they went with AMD for all their MacBook and Mac graphics cards (not entirely sure about Mac Pros). But there was a story that Apple didn't like how closed Nvidias drivers were or something before they ditched them for AMD. I would be interesting to know what Apple thinks about their ARM move now, if Nvidia actually acquires it.

Who knows how this Apple move will go, but right now they are all in on it. Depending on whether this turns out to be a good move, Apple will be dependant on Nvidia, not just on ARM. So it could just be a smart investment to eventually basically shake out Apple for as much as they can. Which, Apple has more than $ 100 billion just sitting there in the bank and that's an older number I heard.
Posted on Reply
#12
xkm1948
PowerPC
I wonder how this will play with Apple moving to ARM. Apple doesn't like Nvidia, I think it clearly showed when they went with AMD for all their MacBook and Mac graphics cards (not entirely sure about Mac Pros). But there was a story that Apple didn't like how closed Nvidias drivers were or something before they ditched them for AMD. I would be interesting to know what Apple thinks about their ARM move now, if Nvidia actually acquires it.

Who knows how this Apple move will go, but right now they are all in on it. Depending on whether this turns out to be a good move, Apple will be dependant on Nvidia, not just on ARM. So it could just be a smart investment to eventually basically shake out Apple for as much as they can. Which, Apple has more than $ 100 billion just sitting there in the bank and that's an older number I heard.
Apple is ditching all AMD GPU for in house silicon too. Basically they want ALL the money to themselves. Now thinking about it, the only time Apple tried to be open and used x86 was when their PowerPC CPU got destroyed by Intel's Core uArc.

Not doubting the usefulness of ARM design. But it is also too early to call this end of x86.

There is no strict RSIC vs CISC as most designs try to learn from what the others are doing better. x86 will it superior abosolute performance and compatibility will still dominate the HPC market for the foreseeable future.
Posted on Reply
#13
fynxer
This is good, the more competition Intel gets more they will have to evolve to counter the threat.

Nvidia buying ARM could be the best thing happening for both Intel and the consumer market.

We get faster and cheaper cpu:s.
Posted on Reply
#14
theeldest
If Nvidia does buy ARM it's fully to support their datacenter work and not consumer devices. That doesn't mean they won't use ARM in consumer, but that's not their focus. It makes more sense when you remember they just bought Mellanox and are currently buying Cumulus. This would give them the ability to do 100% NVIDIA super computers which is a bigger market than most people think. It's not just the research super computers any more, businesses have them for their internal HPC and AI applications
Posted on Reply
#15
PowerPC
xkm1948
Apple is ditching all AMD GPU for in house silicon too. Basically they want ALL the money to themselves. Now thinking about it, the only time Apple tried to be open and used x86 was when their PowerPC CPU got destroyed by Intel's Core uArc.
Apple Silicon GPU for ARM Macs could maybe beat integrated graphics on Intel or AMD side, but I don't think they can leave AMD for dedicated graphics any time soon. I'm willing to bet they will keep working with AMD or maybe even with Nvidia again because customers want dedicated GPUs for obvious reasons. They can't compete with AMD or Nvidia here, or if they don't offer dedicated GPUs anymore, they'll lose a very big segment of their customers.

Don't forget the fastest supercomputer in the world is now officially ARM-based. It beat the now second-best x86 supercomputer by a factor of 2.8x ! The only reason ARM may have problems taking over the HPC market, is simply because of the enormous investments already made into x86 over the years. But customers aren't going to stay loyal to x86, if ARM suddenly is seen as a better choice. Here's a Link to an article.
Posted on Reply
#16
Anymal
moproblems99
The plot thickens....
FTW! NV will know what to do with ARM and still will be opened as regulators wont let it not to be.
Posted on Reply
#17
Prime2515102
xkm1948
In an ideal world if Nvidia can get their hands on Cyrix / VIA's x86 it would be amazing. With Nvidia's R&D track record they would make some decent x86 CPUs.
I'm not so sure than would pan out. For the chip to be competitive it would have to support a good number of modern features (like x64 among countless others) with current patents that would have to be licensed from Intel and AMD and if I were either of those two, the price would be very high.

On top of that, there could still be a cross-licensing deal of some kind between VIA and Intel that could prevent them from transferring their existing Intel-licensed portion of x86 tech in the event of a sale (if they even still have any). I'm speculating on that because otherwise why wouldn't somebody have bought out VIA a long time ago specifically for that purpose? Or bought out AMD for that matter when their stock was at $2.70-ish per share back in 2014(?) and a lot of people were thinking they were about ready to fold?

As far as Cyrix-specific x86 tech, the patents on all that expired long ago so there's really no need to buy anything. Sure, circuits can be copyrighted now and protected even after the patent expires, but change the values of all the resistors and capacitors by tiny percentage that doesn't effect the circuit, move the layout around a bit, and you just circumvented that. But, it's old-ass tech that's not going to do much good without licensing the newer stuff with it.

Bottom line whether any of that is even semi-accurate or not: For whatever reason, I don't think they have any interest in x86 at all:

As much as their corporate practices sickens me, I have to admit that nVida aren't stupid, and they have very talented engineers, and plenty of them, and plenty of resources to hire more. If they wanted x86 tech, they would probably design it themselves just like Cyrix did, except unlike Cyrix, they have the money to not get buried alive in legal costs when Intel (and also AMD this time) sues them.
Posted on Reply
#19
moproblems99
Prime2515102
current patents that would have to be licensed from Intel and AMD and if I were either of those two, the price would be very high.
They can't be unreasonable.
Posted on Reply
#20
theoneandonlymrk
moproblems99
They can't be unreasonable.
They have no reason to even consider it.

Unreasonable, they both developed and cross licenced technology that makes the world run literally, why would they let a third competitor in they're niche, and market's ,do they look that skint, AMD we're but that still didn't happen.

This would push MIPS and RISC options into use IMHO
Posted on Reply
#21
moproblems99
theoneandonlymrk
Unreasonable, they both developed and cross licenced technology that makes the world run literally, why would they let a third competitor in they're niche, and market's ,do they look that skint, AMD we're but that still didn't happen.
Well, they may not, but I am pretty sure US requires business to fairly license their IP when they are already licensing it to other companies. I believe that was true when Samsung and Apple were constantly going at each other so I presume it applies here. It may not, I could be wrong.
Posted on Reply
#22
Fluffmeister
Yeah, Intel and AMD are perfectly happy with their duopoly thank you very much.
Posted on Reply
#23
Prime2515102
moproblems99
Well, they may not, but I am pretty sure US requires business to fairly license their IP when they are already licensing it to other companies. I believe that was true when Samsung and Apple were constantly going at each other so I presume it applies here. It may not, I could be wrong.
I'm not 100% sure, but I doubt it. I think they just worked it out. Being forced to license it, except under some extreme circumstance - like a matter of nation security or something - seems rather purpose defeating. A patent (a utility patent anyway) is essentially a 20-year license to run a monopoly on a new product.

They were probably dealing with patent offices in who knows how many countries also though, with who knows how many different laws, so who knows what went on (off the top of my head, I would say patents rarely get filed in a single country anymore).
Posted on Reply
#24
moproblems99
Prime2515102
I'm not 100% sure, but I doubt it. I think they just worked it out. Being forced to license it, except under some extreme circumstance - like a matter of nation security or something - seems rather purpose defeating. A patent (a utility patent anyway) is essentially a 20-year license to run a monopoly on a new product.

They were probably dealing with patent offices in who knows how many countries also though, with who knows how many different laws, so who knows what went on (off the top of my head, I would say patents rarely get filed in a single country anymore).
While the courts did find Samsung set unreasonable license fees, these cases were more about patent infringement between the two of them so likely doesn't apply here.
The court found that Samsung's fee was unreasonable, but noted that, if the companies cannot make a fair and reasonable licensing fee, Samsung could open a new case against Apple.
en.wikipedia.org/wiki/Apple_Inc._v._Samsung_Electronics_Co.

Yes, I know it is Wikipedia, but good enough for this.
Posted on Reply
#25
Prime2515102
moproblems99
While the courts did find Samsung set unreasonable license fees, these cases were more about patent infringement between the two of them so likely doesn't apply here.



en.wikipedia.org/wiki/Apple_Inc._v._Samsung_Electronics_Co.

Yes, I know it is Wikipedia, but good enough for this.
If you look at the reference for that (#40) that was a decision made in a Dutch court. BUT... it was over 3G and was decided because it's an industry standard. This probably does apply in the US also because it's an ITU standard (International Telecommunication Union) and certain things have to be agreed upon internationally obviously.

Otherwise, they could get everyone onboard and they go into production, trucks go on the road to deliver the products and suddenly, "Oh, sorry, we changed our minds. The license fees just went up thirty fold." The holder of a patent for a big enough tech could potentially destroy an entire nation's economy that way if enough other companies were on board (although, at least in the US, it would never happen because it would be blatant monopolistic practice).

x86 is completely proprietary though so I don't think Intel, AMD, and VIA would have any such obligations except under their own agreements with each other.

I would think anyway. I'm going to do some more reading on this stuff, learning about law is fun! lol I think I missed my calling...
Posted on Reply
Add your own comment