Nvidia has approached Softbank and is considering buying ARM Holdings

Page 8 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

uzzi38

Platinum Member
Oct 16, 2019
2,565
5,575
146
Will update with articles as they come out (they haven't yet but news is public). For now just a couple of Tweets:



This is a real significant move for Nvidia, having a huge role in ARM core designs could have major implementations down the line. Most importantly, it secures them a platform and ecosystem down the line.

Oh and I guess Nvidia becomes the defacto standard for GPU IP for mobile instead of Mali. That too.

EDIT: Bloomberg article here: https://www.bloomberg.com/news/arti...ompany-arm-is-said-to-attract-nvidia-interest
 
Last edited:

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
That's where ARM is headed, if it can get traction. And it's headed for the server room, too. Or at least it's trying. Those are the markets where NV needs ARM the most - not RISC-V. NV is not looking at buying ARM to take over the smartphone market.

Maybe. Other than Apple which is its own world, it remains to be seen if ARM will go anywhere on laptop/desktop. And aside from cloud servers, I see no evidence that ARM is making it in general purpose servers. Marvell's move just confirms that.

Anyway, this thread is explicitly about Nvidia possibly buying ARM, and people's worries about that. That means worries on mobile where it has a big presence, not PC/server where ARM designed cores have no presence (i.e. they are all closed systems and architectural licenses)
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Anyway, this thread is explicitly about Nvidia possibly buying ARM, and people's worries about that. That means worries on mobile where it has a big presence, not PC/server where ARM designed cores have no presence (i.e. they are all closed systems and architectural licenses)

As I stated, NV has presence in desktop/workstation and HPC. That's where NV needs a platform so they can't get kicked out by Intel and AMD. If you were nVidia, which would you rely on for your vertically-integrated HPC platform (and beyond)? RISC-V or ARM?
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
If you were nVidia, which would you rely on for your vertically-integrated HPC platform (and beyond)? RISC-V or ARM?
RISC-V.

Permissive license - free
Product use - free

The decode to native via binary translator actually allows superscalar RISC-V instruction flow to convert into a native VLIW/EPIC RISC-V instruction flow. Nvidia's control with RISC-V would let them have a less bugged implementation.

They could also use RISC-V on their GPUs/Multimedia/etc units.

Convert CPU Vector-workloads to GPU Vector-workloads could help minimize certain aspects of HPC for better CPU&GPU scaling.

Instead of fixed function being stuck with the codecs, they could push fixed-performance software-controlled function units. Where newer more open and efficient codecs would be added down the road.
 
Last edited:

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,482
14,434
136

Vattila

Senior member
Oct 22, 2004
799
1,351
136
For those that do not have subscriptions to paywalled sites, here is a freely available article on the matter, released a couple of hours ago, by Engadget:

"SoftBank might be close to finding a buyer for ARM, and it won’t surprise you who the bidder might be. Wall Street Journal sources say SoftBank is close to a deal to sell ARM Holdings to NVIDIA for “more than” $40 billion. The two have reportedly been in exclusive talks for several weeks, and an agreement is near enough that they could finalize the sale by early next week. [...] However, this is also assuming any deal clears regulatory hurdles. NVIDIA clearly wouldn’t be a neutral owner. It might face stiff opposition from officials concerned that it might withhold technology from rival chipmakers or otherwise give itself an unfair advantage. It could effectively dictate the futures of companies that design ARM-based chips, like Apple and Samsung, and both the government and competitors would likely want reassurances that ARM’s owner will play fair."


So, contrary to my disbelief, it seems Nvidia is attempting, and is well on its way, to take over ARM wholesale. I just cannot image how ARM will continue to function as an open and fair (if not free) licensing house to the industry. What about competitors in the fledgling ARM server space, such as Ampere? How will Apple like to be dependent on Nvidia for their CPU platforms into the future? AMD's embedded security processors inside their CPUs are all ARM. Et cetera.

Either there are some clever conditions and arrangements in the deal, to ensure ARM continues to operate fairly independently, or this will be staunchly opposed from many corners.

Bloomberg also makes this point:

"A deal for Arm could be the largest ever in the semiconductor industry, which has been consolidating in recent years as companies seek to diversify and add scale. Still, any agreement is likely to attract fierce regulatory scrutiny, and potential complaints from rivals of Arm, such as Samsung Electronics Co., Apple Inc., Qualcomm Inc., Broadcom Inc., Intel Corp. and Huawei Technologies Co., which license its designs or code for their own products. While Arm has carved out a successful niche for itself by being independent, an acquisition by Nvidia, also a licensee, would be a challenge to that neutrality. SoftBank’s purchase went ahead largely uncontested because the Japanese company wasn’t a competitor to any of Arm’s customers."


PS. Unhappy voices in the UK:

"Miliband’s criticisms of the potential deal echo those of Hermann Hauser, an investor who was one of Arm’s co-founders. Hauser said one of the key conditions on Nvidia should be a guarantee that Arm is allowed to maintain its business model, which allows the British company to work with any chip manufacturer, including Nvidia’s rivals. “It is very much in Nvidia’s interest to kill Arm,” Hauser told the Guardian."


"ARM’s co-founder Tudor Brown, who worked for the company for 29 years, has previously spoken out against ARM being sold to a semiconductor firm. It should instead remain a neutral supplier to tech companies, he suggested."

 
Last edited:

Tabalan

Member
Feb 23, 2020
41
25
91
"Miliband’s criticisms of the potential deal echo those of Hermann Hauser, an investor who was one of Arm’s co-founders. Hauser said one of the key conditions on Nvidia should be a guarantee that Arm is allowed to maintain its business model, which allows the British company to work with any chip manufacturer, including Nvidia’s rivals. “It is very much in Nvidia’s interest to kill Arm,” Hauser told the Guardian."
Why would Nvidia spend >40 mld/bn $ to hurt... um ... who exactly? Intel and AMD will be pleased if ARM CPUs will fail, Mellanox rather doesn't care about ARM. So who (from Nvidia competitors) would really loose when Nvidia acquires ARM?
 
  • Like
Reactions: Tlh97 and Vattila

Thibsie

Senior member
Apr 25, 2017
727
752
136
Why would Nvidia spend >40 mld/bn $ to hurt... um ... who exactly? Intel and AMD will be pleased if ARM CPUs will fail, Mellanox rather doesn't care about ARM. So who (from Nvidia competitors) would really loose when Nvidia acquires ARM?

All the others.
Besides, Nvidia would get GPU+CPU capabilities and middle/long term it could hurt the other two big time. In theory.
 
  • Like
Reactions: Tlh97 and Vattila

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
How will Apple like to be dependent on Nvidia for their CPU platforms into the future?

They won't be. The author of that article doesn't understand the ARM market, and that Apple designs their own CPU cores and therefore doesn't depend on ARM for anything. Oh sure, Nvidia could screw up future ISA additions, but Apple would be totally fine if they ended up 'stuck' on ARMv8.6 for the next decade.

Qualcomm and Samsung both have architectural licenses as well, so if Nvidia screws up the licensing of future cores they can go back to designing their own. It might hurt them in the short run as they'd be stuck on A78 for a few years while they spun that process back up but they'd still be able to shrink it so it wouldn't be too big of an issue for them.

Nvidia can't compete with them selling smartphone SoCs because they don't have a modem, and somehow I doubt Qualcomm would make it cheap or easy for an OEM who dared to buy an Nvidia smartphone SoC to buy the same discrete Qualcomm modems Apple is going back to buying.

I will go back to my Wall Street rule of thumb. Acquisitions of over $10 billion almost never pay off. For every one you can name that was successful, I can name ten that weren't.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
They won't be. The author of that article doesn't understand the ARM market, and that Apple designs their own CPU cores and therefore doesn't depend on ARM for anything. Oh sure, Nvidia could screw up future ISA additions, but Apple would be totally fine if they ended up 'stuck' on ARMv8.6 for the next decade.

Apple hates nVidia enough that they will ditch ARM if they have to deal with nVidia in any fashion, including paying the standard licensing fees. I would think not being able to update the instruction set would be an issue. I will say that Apple probally has a plan for this, since they didn't show any interest in buying ARM. Either they are moving to their own instruction set or forking ARM and will just stiff nVidia.
 
  • Like
Reactions: Tlh97 and Vattila

Tabalan

Member
Feb 23, 2020
41
25
91
All the others.
Besides, Nvidia would get GPU+CPU capabilities and middle/long term it could hurt the other two big time. In theory.
Who is "all others" that Nvidia competes withs? In their main GPU market they compete with Intel and AMD. Both of these companies would love ARM to fail as it's their future competitor. I'm sure AMD would love to trade failing ARM for ditching security ARM core in their CPU.
 

Tabalan

Member
Feb 23, 2020
41
25
91
Apple hates nVidia enough that they will ditch ARM if they have to deal with nVidia in any fashion, including paying the standard licensing fees. I would think not being able to update the instruction set would be an issue. I will say that Apple probally has a plan for this, since they didn't show any interest in buying ARM. Either they are moving to their own instruction set or forking ARM and will just stiff nVidia.
And what would Apple do? Go back to x86?
Start from the scratch RISC-V core? It would take a lot of time and cash.
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
Also the power of producing a heavy-lifter for a lot of HPC/AI work. Something Intel has been desperately trying (and failing) to take over via acquisitions.
TRUE!

But why I am pointing out the fabs vs fabs-less is this stock market price and they prize not total units sold in a revenue perspective, or total income after expenses but also a rough idea of the upsides and downsides of things like how expensive it is to build and maintain fabs (Capital Expenditures, CAPEX.)

Intel's Revenue numbers for 2019 are $71.97 Billion dollars, and Intel made a profit of $41.94 billion dollars off that income.
Nvidia's Revenue numbers for 2019 are $11.72 Billion dollars, and Nvidia made a profit of $7.21 billion dollars off that income.

58.2% vs 61.5% margins do not explain that insane stock price difference when Intel does 5 to 6x the business of what Nvidia does. It is projections into the future and fear that current CAPEX will not deliver the future profits in the same way it currently does.

So instead of writting 150 words it is much easier to say Fabs vs Fab-less is what the stock market is rewarding.
 
Last edited:
  • Like
Reactions: Tlh97 and Saylick
Mar 11, 2004
23,031
5,495
146
Apple hates nVidia enough that they will ditch ARM if they have to deal with nVidia in any fashion, including paying the standard licensing fees. I would think not being able to update the instruction set would be an issue. I will say that Apple probally has a plan for this, since they didn't show any interest in buying ARM. Either they are moving to their own instruction set or forking ARM and will just stiff nVidia.

My guess is Apple figures that if Nvidia tries to squeeze them they'll just go to DoJ and get Nvidia slammed. Nvidia tried patent litigation a little while back and got denied pretty soundly. I can't imagine they'd fare much better on this.
 

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
Apple hates nVidia enough that they will ditch ARM if they have to deal with nVidia in any fashion, including paying the standard licensing fees. I would think not being able to update the instruction set would be an issue. I will say that Apple probally has a plan for this, since they didn't show any interest in buying ARM. Either they are moving to their own instruction set or forking ARM and will just stiff nVidia.

Bullcrap. Companies aren't people, they don't have irrational hatred of other companies to the point that they will uproot all their plans just to spite them. That's the sort of thing a person will do, but not a corporation the size of Apple.

They don't care who gets the yearly check for their architectural license fees. Their "plan" for Nvidia buying ARM is to do nothing, and keep using ARM. They can't "fork" it, but worst case they could continue using ARMv8.6 or whatever the latest version their architectural license covers forever. New additions to the ARM ISA are tinkering at the edges, they won't make any real difference in performance or capability if Apple ignored them.
 
  • Like
Reactions: xpea

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
Bullcrap. Companies aren't people, they don't have irrational hatred of other companies to the point that they will uproot all their plans just to spite them. That's the sort of thing a person will do, but not a corporation the size of Apple.

My point is that Apple would have stepped in and either bought ARM themselves, or would find a way to disrupt the acquisition once nVidia was linked... if they didn't have a way out.
 

Vattila

Senior member
Oct 22, 2004
799
1,351
136
My point is that Apple would have stepped in and either bought ARM themselves, or would find a way to disrupt the acquisition once nVidia was linked

They may still do. This deal is far from settled, I guess. It wouldn't surprise me if ARM's licensees are joining up on video conferences these days, discussing how to protect their interests.

Some thoughts:

I find it hard to believe Jensen spends $40B+ to get into a low-margin licensing business. I can see the argument that he may be spooked by AMD's "Radeon Everywhere" drive; in particular the deal with Samsung around Radeon IP in tablets and mobile phones. Getting control over the ARM ecosystem, as some here have suggested, may let him shoehorn GeForce into the mobile space. But spending $40B+ to achieve that seems to be an expensive way to go about it, with no clear route to big profits and payback.

My take on the motivation for the takeover is that Jensen wants to strengthen Nvidia's position in the data centre. Nvidia's lead in AI/GPGPU and rapid revenue growth in this space are the main reasons for the lofty market valuation (recently passing Intel's market capitalisation). There is a lot of hype driving the stock, about AI and data processing from the data-centre to the edge, where shareholders bet that Nvidia will lead, and thus generate large profits in short order.

However, Nvidia needs world-class server CPU capability. Intel has stumbled badly recently, and Nvidia was forced to switch to AMD EPYC in their flagship DGX A100 server. That is is a brittle and embarrassing situation. To end the dependence on a competitor that has "A+A" (AMD EPYC CPU + Radeon GPU) as its competitive goal, and with Intel now also having comprehensive AI and GPGPU ambitions, Nvidia needs server CPU capability urgently, to stand up to its data centre competitors long term.

ARM is now the immediate challenger to x86, and they have an aggressive and promising roadmap for attacking the server space (a few years ago, when Jim Keller was enthusiastically working on K12, AMD's high-performance ARM processor, ARM stood on stage with AMD proclaiming 20% data centre share for ARM in 2020, which reflects the scale of their ambition, if not execution). To leapfrog Ampere, Nuvia and other promising ARM server CPU upstarts and efforts, taking over ARM makes sense. Jensen may be willing to spend big to jump ahead, and getting ARM's excellent CPU team and server roadmap will let him hit the ground running. This can clearly be worth $40B+, if it takes Nvidia into the lead in the data centre and keeps AMD and Intel at bay, while — if panning out as promised — the AI revolution ushers in a new era of high-performance computing.

But then there are all the competition issues with the ARM licensees. As concerned ARM founders express, there is a real threat that Nvidia may undermine competitors and completely change ARM's current open and fair business model. But, unless Jensen wants to spend years in litigation, I guess he has a solution to this conundrum; perhaps a licensing division operating independently, perhaps with a stake for the biggest licensees; perhaps a consortium of sorts, that can protect the interests of licensees and cooperate on shaping the future of the ISA.

PS. Note that ARM licenses more that just CPU ISA and cores, such as interconnect and other SoC technologies (e.g. see their System IP offerings).
 
Last edited:
  • Like
Reactions: Schmide and Tlh97

NTMBK

Lifer
Nov 14, 2011
10,208
4,940
136
They may still do. This deal is far from settled, I guess. It wouldn't surprise me if ARM's licensees are joining up on video conferences these days, discussing how to protect their interests.

Some thoughts:

I find it hard to believe Jensen spends $40B+ to get into a low-margin licensing business. I can see the argument that he may be spooked by AMD's "Radeon Everywhere" drive; in particular the deal with Samsung around Radeon IP in tablets and mobile phones. Getting control over the ARM ecosystem, as some here have suggested, may let him shoehorn GeForce into the mobile space. But spending $40B+ to achieve that seems to be an expensive way to go about it, with no clear route to big profits and payback.

My take on the motivation for the takeover is that Jensen wants to strengthen Nvidia's position in the data centre. Nvidia's lead in AI/GPGPU and rapid revenue growth in this space are the main reasons for the lofty market valuation (recently passing Intel's market capitalisation to become the largest semi-conductor company in the world). There is a lot of hype driving the stock, about AI and data processing from the data-centre to the edge (last hop on the Internet before the consumer), where shareholders bet that Nvidia will lead, and thus produce large profits in short order.

However, Nvidia needs world-class server CPU capability. Intel has stumbled badly recently, and Nvidia was forced to switch to AMD EPYC in their flagship DGX A100 server. That is is a brittle and embarrassing situation. To end the dependence on a competitor that has "A+A" (AMD EPYC CPU + Radeon GPU) as its competitive goal, and with Intel now also having comprehensive AI and GPGPU ambitions, Nvidia needs server CPU capability urgently, to stand up to its data centre competitors long term.

ARM is now the immediate challenger to x86, and they have an aggressive and promising roadmap for attacking the server space (a few years ago, when Jim Keller was enthusiastically working on K12, AMD's high-performance ARM processor, ARM stood on stage with AMD proclaiming 20% data centre share for ARM in 2020, which reflects the scale of their ambition, if not execution). To leapfrog Ampere, Nuvia and other promising ARM server CPU upstarts and efforts, taking over ARM makes sense. Jensen may be willing to spend big to jump ahead, and getting ARMs excellent CPU team and server roadmap will let him hit the ground running. This can clearly be worth $40B+, if it takes Nvidia into the lead in the data centre and keeps AMD and Intel at bay, while — if panning out as promised — the AI revolution ushers in a new era of high-performance computing.

But then there is all the competitive issues with the ARM licensees. As concerned ARM founders express, there is a real threat that Nvidia may undermine competitors and completely change ARM's current open and fair business model. But, unless Jensen wants to spend years in litigation, I guess he has a solution to this conundrum; perhaps a licensing division operating independently, perhaps with a stake for the biggest licensees; perhaps a consortium of sorts, that can protect the interests of licensees and cooperate on shaping the future of the ISA.

PS. Note that ARM licenses more that just an ISA, such as interconnect and other SoC IP (e.g. see their System IP).

Nvidia already has an ARM architectural license (for their Denver custom CPU), and they could surely have acquired Nuvia for much less than $40bn. Hell, they could have bought Vulcan when Broadcom were shopping it around, or XGene from AppliedMicro. Or the laid off Samsung CPU team. There are a lot of options short of buying ARM that would let them build a credible server chip.
 
  • Like
Reactions: Vattila

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
My point is that Apple would have stepped in and either bought ARM themselves, or would find a way to disrupt the acquisition once nVidia was linked... if they didn't have a way out.

My point is that they don't need a way out, because who owns ARM doesn't really matter that much to them.
 

scannall

Golden Member
Jan 1, 2012
1,944
1,638
136
Apple hates nVidia enough that they will ditch ARM if they have to deal with nVidia in any fashion, including paying the standard licensing fees. I would think not being able to update the instruction set would be an issue. I will say that Apple probally has a plan for this, since they didn't show any interest in buying ARM. Either they are moving to their own instruction set or forking ARM and will just stiff nVidia.
They have a perpetual v8.x license. They don't actually *need* to stay ARM compliant if nVidia were to buy them, since their processors only go into their products and not anywhere else. Plus they write the entire software and software development stack.
 

Vattila

Senior member
Oct 22, 2004
799
1,351
136
There are a lot of options short of buying ARM that would let them build a credible server chip.

Participating is easy. Winning is hard.

So the details matter; the particular resources, IP, technology, roadmap and people needed to, not only participate, but to win. I presume Jensen has evaluated all of this with much more information than is generally available, giving him the insight to see this take-over as the best solution.
 
Last edited: