SoftBank reportedly closing a $40 billion sale of ARM to Nvidia -- now confirmed

[onetheycallEric]

Forward-looking: Nvidia has announced it has reached an agreement with SoftBank to acquire Arm Limited, in a transaction worth $40 billion. While Nvidia will take possession of Arm and its lucrative IP portfolio, the transaction does not include Arm's IoT Services Group, which will remain with SoftBank. Additionally, SoftBank will continue to remain invested in Arm through its shareholder stake in Nvidia, which is expected to come in under 10 percent. More details to follow.

Analysis: Nvidia purchase of Arm completely resets semiconductor landscape

Follow-up: It's official: Nvidia is acquiring Arm

After weeks of speculation about the future of ARM, it seems a deal is about to be struck between SoftBank and Nvidia, allowing the GPU maker to acquire ARM in what would be one of the most expensive tech acquisitions in history. Any such deal involving tech giants is guaranteed to be met with regulatory force, but Nvidia is likely betting it can assuage lawmakers and regulators.

Largest tech deals in history

Company	Acquisition	Price	Year
Dell	EMC	$64 billion	2015
Avago	Broadcom	$37 billion	2015
IBM	Red Hat	$34 billion	2018
Softbank	ARM Holdings	$31.4 billion	2016
Microsoft	LinkedIn	$26.2 billion	2016

According to a report from The Wall Street Journal, SoftBank and Nvidia are nearing a deal for the graphics giant to acquire ARM for $40 billion in a cash and stock transaction. SoftBank acquired Arm for around $31 billion in 2016, and in recent months has been looking to offload the chip design and IP company in an effort to mitigate losses. Nvidia has increasingly emerged as the most interested party.

While the terms of the deal have yet to be revealed, an official announcement could come as early as next week. The deal will undoubtedly come under scrutiny and be subject to regulatory approval. Arm's co-founder has previously said that an Nvidia acquisition would be a disaster, and that Arm should be brought back home to the UK.

One key concern stemming from an Nvidia acquisition is retaining ARM's neutrality in licensing its chip designs and IP. SoftBank isn't a chip maker, and as such, isn't competing with other manufacturers like Samsung, Apple, Intel, Broadcom, and Qualcomm, just to name a few -- all of whom are ARM licensees.

Nvidia controlling access to such an important and broad ecosystem that the industry relies on could be seen as a conflict of interest and is bound to raise concerns.

Nvidia has shown that it wants to continue its vertical integration and forge ahead in high performance compute, AI, data center infrastructure, and supercomputers. They've made several acquisitions in an effort to scale its ambitions across not only its GPUs, but also its software stack and networking. Nvidia recently acquired Mellanox Technologies and Cumulus Networks, and brought them under the Nvidia Networking brand. The end goal is the desire to more tightly couple Nvidia's HPC platforms with networking hardware and software.

Nvidia wants to do the same with GPUs and CPUs -- except it lacks any CPUs of its own. Bringing in ARM under its umbrella may allow them to compete for lucrative exascale supercomputer contracts that Intel and AMD are enjoying. Recently, ARM-based supercomputers have made more of a splash, with Japan's Fugaku supercomputer dethroning IBM's Summit as the world's most powerful supercomputer.

Nvidia has its line of Tegra SoCs, and others, that have historically leveraged ARM's instruction set and Cortex-Axx series of CPUs, but that's not the same has having custom CPU and GPU design under the same roof.

Having its own CPU design in-house would allow Nvidia to better conjoin ARM CPUs to its GPU accelerators, and put it on a more level playing field with Intel and AMD. Nvidia already announced it'll bring Cuda to ARM and enable Nvidia's full stack of HPC software within the ARM ecosystem. Custom ARM CPUs could be the next step, as far as Nvidia is concerned.

We likely won't have to wait long to see how this development plays out, and if it does, it will be one of the biggest tech acquisitions in history and have a serious ripple effect on the semiconductor industry.

Image credit: Daniel Constante

Permalink to story.

https://www.techspot.com/news/86722-softbank-reportedly-closing-40-billion-sale-arm-nvidia.html

 

[Mr Majestyk]

No effing way this should be allowed to stand by the authorities. Kill it off immediately, this will be a disaster.

 

[gamerk2]

No effing way this should be allowed to stand by the authorities. Kill it off immediately, this will be a disaster.

Nah, let the market decide winners and losers. The market is ALWAYS right.

/sarcasm

 

[Irata]

Still don‘t get why nVidia needs to buy ARM in order to make their own custom CPU. Apple and others already do it, so....?

 

[Endymio]

No effing way this should be allowed to stand by the authorities. Kill it off immediately...

I don't see much market overlap between Arm and NVidia. Not sure why you're so alarmed.

Myself, I'm a bit more concerned about the potential for NVidia to acquire a crushing debt load to go along the acquisition, which would reduce its R&D expenditures ... but that's not something with which regulators should be involved obviously.

Still don‘t get why nVidia needs to buy ARM in order to make their own custom CPU. Apple and others already do it, so....?

NVidia tried -- and essentially failed -- already. That was Tegra's original goal.

 

[trparky]

I'd ordinarily say that this is a good thing, however... this is nGreedia (nVidia); the company that already charges insane prices for their highest end GPUs. Now way this is a good thing with them in charge.

 

[Darth Shiv]

I'd ordinarily say that this is a good thing, however... this is nGreedia (nVidia); the company that already charges insane prices for their highest end GPUs. Now way this is a good thing with them in charge.

Their highest end GPUs are massive silicon and boards. It's kind of justified. AMD isn't releasing competition at the top of the tree. Why not? That's not NVIDIA's fault.

 

[VaRmeNsI]

Turing prices were largely a result of zero competition and technically you only need such expensive cards for gaming, you're not going to starve without one.

 

[poohbear]

If this deal goes through, it could very well be the end of x86 CPUs. It's all Intel's fault really, they dropped the ball so many times with their delay after delay and even lost Apple to ARM CPUs... the writing is on the wall for X86 CPUs.

 

[poohbear]

I'd ordinarily say that this is a good thing, however... this is nGreedia (nVidia); the company that already charges insane prices for their highest end GPUs. Now way this is a good thing with them in charge.

Insane prices? Their GTX 3070 founder's edition, which offers performance equal to a 2080 ti (a $1600 card), is only $499. I'm sure OEMs will offer even cheaper options.

I think they've tried different pricing in the past to experiment with what consumers can handle, but they eventually settle their prices to market conditions, especially with a resurgent AMD.

 

[BookofNeat]

If Nvidia leverages the acquisition of Arm to further its hold on the market, I unironically support it. Creating monopolies on non-essential markets could help fix the problems that come with rampant consumerism. If suddenly a gpu or cellphone cost 300% more because of it, I'll just have to keep my current ones a few more years. I can live with that but that's just me.

 

[quadibloc]

If NVIDIA satisfies authorities that it won't license ARM in a discriminatory way to smartphone makers, or make problems for Apple, but it gains from owning ARM a better ability to make ARM desktop CPUs as an alternative to x86, this would be a good thing.
Of course, then they'd be under pressure to license ARM in a non-discriminatory way to Intel and AMD too, I suppose. But I don't expect NVIDIA to make the ARM destop that roaring a success that quickly; AMD and Intel both probably will be better off sticking to x86 for some considerable time to come.
Of course, Apple could mis-step, and end up becoming an acquisition target itself.
Imagine it being bought by NVIDIA and IBM together, with the new Macintosh containing chips that handled both the ARM and PowerPC instruction sets...

 

[JAmie71]

I just think back to 3DFX and how they stopped supporting products just basically IP stripped it.

 

[gamerk2]

If this deal goes through, it could very well be the end of x86 CPUs. It's all Intel's fault really, they dropped the ball so many times with their delay after delay and even lost Apple to ARM CPUs... the writings on the wall for X86 CPUs.

x86 is a horrid architecture; it's a shame IBM didn't go with the 68k.

Also remember Intel made a CPU that was specifically designed for multithreaded programs; it was called Itanium. But no, AMD has to extend x86 instead.

What's going to ultimately happen is PCs are going to be stuck on x86-64 due to legacy software, and everything else is going to go ARM.

 

[poohbear]

Of course, Apple could mis-step, and end up becoming an acquisition target itself.
Imagine it being bought by NVIDIA and IBM together, with the new Macintosh containing chips that handled both the ARM and PowerPC instruction sets...

no one can afford to buy Apple. It was the world's first 2 trillion dollar company. Nvidia is a 300 billion dollar company, IBM a 108 billion dollar company, so not even close

 

[Irata]

Insane prices? Their GTX 3070 founder's edition, which offers performance equal to a 2080 ti (a $1600 card), is only $499. I'm sure OEMs will offer even cheaper options.

I think they've tried different pricing in the past to experiment with what consumers can handle, but they eventually settle their prices to market conditions, especially with a resurgent AMD.

I am curious about the FE‘s availability, I.e. how many and for how long at that price.
Judging by the leaked third party card prices (via Videocardz) it appears that they are going to be quite a bit higher.

 

[HardReset]

x86 is a horrid architecture; it's a shame IBM didn't go with the 68k.

Also remember Intel made a CPU that was specifically designed for multithreaded programs; it was called Itanium. But no, AMD has to extend x86 instead.

What's going to ultimately happen is PCs are going to be stuck on x86-64 due to legacy software, and everything else is going to go ARM.

Itanium was supposed to be architecture where CPU can have simple design because compiler should do most work arranging instructions so that CPU could basically just crunch them. Well, there was never compiler that could do it effectively so Itanium flopped. That has nothing to do with multithreading though.

I have read about death of x86 for over 30 years now. Still waiting.

 

[Endymio]

I have read about death of x86 for over 30 years now. Still waiting.

And it's even less likely today than it was then, as the legacy code base is even larger.

 

[Lounds]

All this will mean is more expensive smartphones... They'll get the data center contract's by being able to undercut AMD and intel and at the same time control the price of smartphone SoC's making smartphones even more expensive. I mean all I want is a new Shield TV lol.

 

[Darth Shiv]

If this deal goes through, it could very well be the end of x86 CPUs. It's all Intel's fault really, they dropped the ball so many times with their delay after delay and even lost Apple to ARM CPUs... the writing is on the wall for X86 CPUs.

Not with Windows ecosystem it won't be - too large to be eaten by ARM for a very very long time. Also AMD is doing fine on iGPU console chips. Desktops still can handle the higher thermal footprint. Is there a single threaded competitor to x86 cores?

 

[gamerk2]

Itanium was supposed to be architecture where CPU can have simple design because compiler should do most work arranging instructions so that CPU could basically just crunch them. Well, there was never compiler that could do it effectively so Itanium flopped. That has nothing to do with multithreading though.

I have read about death of x86 for over 30 years now. Still waiting.

Early Itanium compilers were bad, just like early x86 compilers were bad. The compilers got better with time, and towards the end of it's lifespan I'd argue that Itanium compilers were outperforming the standard x86 ones (MSVC, GCC, and Intels compiler) since they didn't need to rely on as many architectural tricks to extract performance (the same architectural tricks that are now allowing security holes, which I note the Itanium architecture is immune to).

Simply put, if you designed a CPU architecture for todays needs, it would look an awful lot like the Itanium.

 

[Stoly]

Still don‘t get why nVidia needs to buy ARM in order to make their own custom CPU. Apple and others already do it, so....?

I guess you could argue they don't need to, but just the IP is worth billions in research and then there's the licensing. Actually I'm surprised Apple, Amazon or even MS didn't buy it first.

 

[Stoly]

Nvidia has been trying to license its ip to mobile devices since the original Tegra IIRC. Finally they have the chance to incorporate it. I mean its orders of magnitude better than MALI. This alone would be a HUGE deal. Mobile is the only market where nvidia tried to enter and failed miserably, now they will be back with a vengeance.

 

[Stoly]

no one can afford to buy Apple. It was the world's first 2 trillion dollar company. Nvidia is a 300 billion dollar company, IBM a 108 billion dollar company, so not even close

exactly, Apple could buy both plus AMD and Intel and still keep change in its pocket.

 

[HardReset]

Early Itanium compilers were bad, just like early x86 compilers were bad. The compilers got better with time, and towards the end of it's lifespan I'd argue that Itanium compilers were outperforming the standard x86 ones (MSVC, GCC, and Intels compiler) since they didn't need to rely on as many architectural tricks to extract performance (the same architectural tricks that are now allowing security holes, which I note the Itanium architecture is immune to).

Simply put, if you designed a CPU architecture for todays needs, it would look an awful lot like the Itanium.

There was never (and probably never will be) compiler for Itanium that worked well enough. Itanium is immune to many security issues that come from speculative execution because Itanium is in-order CPU and so there is no real need for speculative execution.

And for last paragraph, let's look some latest CPU-designs:

- AMD Zen: Out-of-Order
- ARM Cortex cores: Out-of-Order
- Apple ARM cores: Out-of-Order
- Intel Cove-cores: Out-of-Order
- Intel Tremont (latest Atom): Out-of-Order, first Atom's were in-order.

So basically all modern "high performance" cores seem to be out-of-order so they are all very very far from Itanium design.