Nvidia has approached Softbank and is considering buying ARM Holdings

[uzzi38]

Will update with articles as they come out (they haven't yet but news is public). For now just a couple of Tweets:

https://twitter.com/x/status/1286003982079533056

https://twitter.com/x/status/1286004990327132161

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

 

[Ajay]

They have to pay Softbank 1.25B USD if the merger does not go through.

Source: https://www.barrons.com/articles/so...ia-pays-750-million-licensing-fee-51600459412

So NV gets something out of this if it fails:

That sum has been prepaid to SoftBank (ticker: SFTBY) as part of a $2 billion cash payment due from Nvidia (NVDA) at signing, according to a securities filing detailing terms of the deal. The rest of the $2 billion reflects $750 million paid to Arm—not its parent, SoftBank—for the licensing of intellectual property.

 

[Heartbreaker]

Why on green earth does nvidia need ARM for? They have a perpetual license, they sell $1000 GPUs, they sell every GPU they make. why would they need the IP rights for ARM CPUs?

A few posts above yours:

Page 17 - Nvidia has approached Softbank and is considering buying ARM Holdings

Page 17 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Roland00Address]

So in a way Nvidia's whole push to buy Arm could be summed up with Jensen Huang wanting to one up Christiano Amon and now not having the decency to let his pride take a step back to recognize the futility of this endeavor? How childish...

This is the same Nvidia CEO who took it real personally when Intel said no more motherboards, and lots more language that I am skipping due to profanity rules.

Of course Nvidia got messed with by Intel but the CEO took it real personal said bad language, and then after a while turned it into hype that this is actually a good thing and how Nvidia's future is ARM socs.

 

[jpiniero]

Why on green earth does nvidia need ARM for? They have a perpetual license, they sell $1000 GPUs, they sell every GPU they make. why would they need the IP rights for ARM CPUs?

They want to get into the CPU business (Servers, desktops, laptops, you name it). Wiping out the ARM competition would make getting a foothold much easier.

 

[Saylick]

This is the same Nvidia CEO who took it real personally when Intel said no more motherboards, and lots more language that I am skipping due to profanity rules.

Of course Nvidia got messed with by Intel but the CEO took it real personal said bad language, and then after a while turned it into hype that this is actually a good thing and how Nvidia's future is ARM socs.

Yep. Kind of crazy to think that was only 10 years ago. JHH strikes me as a super competitive, never-show-weakness kind of fellow. Whenever he loses, he always finds a way to either spin it as a win or to drop all mention of it entirely as if the loss never happened in his eyes, e.g. Tegra and Icera.

 

[Ajay]

They want to get into the CPU business (Servers, desktops, laptops, you name it). Wiping out the ARM competition would make getting a foothold much easier.

Definitely servers. That's there bread and butter. Still, seems like, by now, they should have been able to build a competitve ARM based server CPU if they had really made it a priority. Then again, NV hasn't done so well with it's own internal ARM CPU development.

 

[DrMrLordX]

I'm starting to think that it would be better for NVIDIA to not buy ARM. At this point I'm not entirely sure what they would gain that would be worth the money.

At this point, yes. Anyone looking to dodge a future NV/ARM merger can just get perpetual licenses to ARMv9.x and roll their own. They'll be covered for a decade if ARMv9 lasts as long on the market as ARMv8.x did.

Why on green earth does nvidia need ARM for? They have a perpetual license, they sell $1000 GPUs, they sell every GPU they make. why would they need the IP rights for ARM CPUs?

To force people adopting future NeoVerse iterations to use CUDA and associated hardware.

 

[Ajay]

To force people adopting future NeoVerse iterations to use CUDA and associated hardware.

Interesting point. I don’t think NV would get away with forcing it on customers. But, they can simply have ARM stop development of GPU IP and offer a Nvidia graphics instead. Eventually, the old IP will be so out of date that its usage would stop. NV is really looking down the road on this acquisition.

 

[DrMrLordX]

I don’t think NV would get away with forcing it on customers.

JHH likely wants a hedge against AMD and Intel pushing NV off their platforms.

 

[Heartbreaker]

JHH likely wants a hedge against AMD and Intel pushing NV off their platforms.

Much more than a hedge.

I think he is just correctly betting on Windows eventually switching to ARM. When that happens NVidia, turns a long time deficit (lack of in house x86 CPU) to a massive Strength.

When first asked Jensen indicated, this deal gives them a new distribution channel for IP. They don't have to force anyone, they just have to make it convenient. Look at today. No one really wants ARM GPUs. Qualcomm uses their own Adreno (IP they acquired from AMD), and Samsung struck a deal with AMD. Apple uses Imagination derived in house design. That's pretty much all the top end ARM SoCs, none using ARM GPU. It really looks like no one actually wants ARM Mali GPUs if there is an option.

Also NVidia has decent quality everything else, GPU, AI,connectivity (Mellanox), they really only lack the CPU piece to have everything in house to build world class SoCs that can run everything, phones, tablets, laptops, gaming PCs, game consoles, Servers, Super Computers...

Having ARM would create the complete in house, one stop shopping solution for everything in computing.

You could argue that they could do that now, just licensing ARM. But another thing they don't have is a decent CPU design group to build world class designs like Apple does, so they can't push out x86 beating ARM SoCs like Apple does. But if NVidia owned ARM, and owned the design team, they could increase it's funding with an eye on beating x86 sooner rather than later, and accelerate all of this...

 

[Doug S]

They want to get into the CPU business (Servers, desktops, laptops, you name it). Wiping out the ARM competition would make getting a foothold much easier.

Nvidia is worth 3.5x more than Intel, is being in the CPU business really going to grow them that much? More and more computation is moving off the CPU onto GPU or other specialized units. What's more that's happening across the entire landscape from phones to HPC, so it isn't an isolated trend. Trying to make a big entrance into the CPU market in 2021 might end up looking as smart in a decade as trying to make a big entrance into gasoline powered automobile market in 2021.

Other than Apple there is currently no real competition in the "servers, desktops, laptops" business as far as ARM goes. The serious ones like Apple or Qualcomm (who isn't a real competitor there yet but has a chance if the Nuvia thing works out) have architectural licenses so it is impossible for Nvidia to "wipe them out". A lot of companies who are licensing ARM designed cores like Samsung have architectural licenses, and could go back to designing their own cores Nvidia tried to squeeze them out.

 

[itsmydamnation]

And Intel has 3x the revenue, the stock market is gambling , things don't have to be based off reality. Make cap is a poor way to judge if it's worth entering a market.

With lots of startups and major chip makers entering the ai space how is NV going to continue 50pc yoy growth that gives them said market cap your talking about without entering new markets?

 

[Hitman928]

Nvidia is worth 3.5x more than Intel, is being in the CPU business really going to grow them that much? More and more computation is moving off the CPU onto GPU or other specialized units. What's more that's happening across the entire landscape from phones to HPC, so it isn't an isolated trend. Trying to make a big entrance into the CPU market in 2021 might end up looking as smart in a decade as trying to make a big entrance into gasoline powered automobile market in 2021.

Other than Apple there is currently no real competition in the "servers, desktops, laptops" business as far as ARM goes. The serious ones like Apple or Qualcomm (who isn't a real competitor there yet but has a chance if the Nuvia thing works out) have architectural licenses so it is impossible for Nvidia to "wipe them out". A lot of companies who are licensing ARM designed cores like Samsung have architectural licenses, and could go back to designing their own cores Nvidia tried to squeeze them out.

Nvidia's last reported annual revenue was ~$17B, Intel's was ~$78B. There's quite a bit of room for Nvidia to grow if they take over on the CPU front.

 

[Heartbreaker]

Nvidia is worth 3.5x more than Intel, is being in the CPU business really going to grow them that much?

The stock market is a bet on future performance. Not current earnings.

A lot of people already assume that NVidia + ARM as a better bet going forward, than Intel that is facing increased competition.

NVidia has almost tripled in value since announcing plans to acquire ARM. Some of that increase will be people betting the ARM takeover happens.

Intel has been stagnant for years based on rising competition on all fronts.

 

[jpiniero]

The stock market is a bet on future performance. Not current earnings.

The stock market is a casino.

 

[Roland00Address]

The stock market is a bet on future performance. Not current earnings.

A lot of people already assume that NVidia + ARM as a better bet going forward, than Intel that is facing increased competition.

NVidia has almost tripled in value since announcing plans to acquire ARM. Some of that increase will be people betting the ARM takeover happens.

Intel has been stagnant for years based on rising competition on all fronts.

Language has something called "performative utterances", things you can not prove via empirics or via science at the time the utterances statements are made.

There are many forms of performative utterances, once of which is "I bet you X this will happen." The bet is of course its own form of logic, but there is no way to empirically validate it in the here and now, we merely have to wait.

Was 1000s of people making their own individual bets the best describer of future reality, we will never know until the moment of agency has long past us.

 

[DrMrLordX]

@guidryp

I'm not sure NV is really betting on WARM. Not saying they aren't, but right now, the majority of their revenue comes from large discrete accelerator cards: dGPUs, etc. With the advent of CXL and a drive to push developers off CUDA exclusivity, the CUDA/NVLink lockin is somewhat imperiled. And it would get worse if AMD and Intel just straight-up broke functionality with NV's cards on enterprise-class systems. NV needs their own platform where they can continue to promote exclusive lock-in to their product stack since that's how they make their money now.

 

[beginner99]

@guidryp

I'm not sure NV is really betting on WARM. Not saying they aren't, but right now, the majority of their revenue comes from large discrete accelerator cards: dGPUs, etc. With the advent of CXL and a drive to push developers off CUDA exclusivity, the CUDA/NVLink lockin is somewhat imperiled. And it would get worse if AMD and Intel just straight-up broke functionality with NV's cards on enterprise-class systems. NV needs their own platform where they can continue to promote exclusive lock-in to their product stack since that's how they make their money now.

Fully agree and thew push towards cloud makes it more commercially viable for the cloud provider to support not yet very common APIs eg. RocM or oneapi. For local use on local laptop/server CUDA is unbeatable. But in the cloud you don't care anymore about what pytorch or tensorflow uses underneath. You only care about performance and not having to bother with setting up the whole stack yourself.
if it suddenly becomes feasible to use AMD and Intel GPUs for basic data sciene / ML /DL tasks for the avergae Joe user on the cloud, then it puts a lot of pressure on NVs CUDA ecosystem. While using CUDA is relativley easy it's even easier not having to deal with this level at all. "It just works".

 

[Heartbreaker]

@guidryp

I'm not sure NV is really betting on WARM. Not saying they aren't, but right now, the majority of their revenue comes from large discrete accelerator cards: dGPUs, etc. With the advent of CXL and a drive to push developers off CUDA exclusivity, the CUDA/NVLink lockin is somewhat imperiled. And it would get worse if AMD and Intel just straight-up broke functionality with NV's cards on enterprise-class systems. NV needs their own platform where they can continue to promote exclusive lock-in to their product stack since that's how they make their money now.

Intel or AMD breaking competitor functionality would be subject to regulatory oversight. So this seems like a very far fetched argument.

It seems obvious that it's only a question of when, not if, x86 dominance fades in it's last strongholds (Windows and Servers).

NVidia owning ARM, means NVida can both accelerate that by pouring more R&D money into desktop/server class ARM cores, and benefit significantly when it happens.

Regardless of how soon ARM ascends on desktops and Servers, NVidia would be able to benefit from owning the ARM sales channel, as already mentioned, this was their primary motivation when asked. They can do bundle deals with GPU, Mellanox, or other IP into the the ARM sales channel. This pays off even before the the rise of ARM desktop and servers.

That's the short term. Long term it seems inevitable that ARM will takeover in desktops/servers, and there will be additional massive dividends when that happens.

 

[DrMrLordX]

Not sure how ARM acquisition helps ensure a CUDA lockin.

Once NV acquires ARM, they can effectively "poison the well", making it hard for anyone to adopt NeoVerse without also adopting CUDA. NV will be able to sell you the entire HPC/Enterprise package: motherboard, CPU, dGPU/accelerator, and software stack. All in one tidy package.

(that assumes NV will ever finally win approval, which it now seems they never will)

 

[Heartbreaker]

Once NV acquires ARM, they can effectively "poison the well", making it hard for anyone to adopt NeoVerse without also adopting CUDA. NV will be able to sell you the entire HPC/Enterprise package: motherboard, CPU, dGPU/accelerator, and software stack. All in one tidy package.

(that assumes NV will ever finally win approval, which it now seems they never will)

NVidia locking out competitors would be run foul of regulators just like AMD/Intel blocking NVidia would. This is not a strong argument.

 

[DrMrLordX]

NVidia locking out competitors would be run foul of regulators just like AMD/Intel blocking NVidia would. This is not a strong argument.

Apparently it already has!

 

[Saylick]

Speaking of companies moving towards RISC-V....
https://www.anandtech.com/show/17104/imagination-launches-catapult-family-of-riscv-cpu-cores

At present, Imagination has finished their first CPU core design, which is a simple, in-order core for 32-bit and 64-bit systems. The in-order Catapult core is being used for microcontrollers as well as real-time CPUs, and according to the company, Catapult microcontrollers are already shipping in silicon as part of automotive products. Meanwhile the real-time core is available to customers as well, though it’s not yet in any shipping silicon.

​

The current in-order core design supports up to 8 cores in a single cluster. The company didn’t quote any performance figures, though bear in mind this is a simple processor meant for microcontrollers and other very low power devices. Meanwhile, the core is available with ECC across both its L1 and TCM caches, as well as support for some of RISC-V’s brand-new extensions, such as the Vector computing extension, and potentially other extensions should customers ask for them.

​

Following the current in-order core, Imagination has essentially three more core designs on their immediate roadmap. For 2022 the company is planning to release an enhanced version of the in-order core as an application processor-grade design, complete with support for “rich” OSes like Linux. And in 2023 that will be followed by another, even higher performing in-order core for the real-time and application processor markets. Finally, the company is also developing a much more complex out-of-order RISC-V core design as well, which is expected in the 2023-2024 timeframe. The out-of-order Catapult would essentially be their first take on delivering a high-performance RISC-V application processor, and like we currently see with high-performance cores the Arm space, has the potential to become the most visible member of the Catapult family.

Farther out still are the company’s plans for “next generation heterogeneous compute” designs. These would be CPU designs that go beyond current heterogeneous offerings – namely, just placing CPU, GPU, and NPU blocks within a single SoC – by more deeply combining these technologies. At this point Imagination isn’t saying much more, but they are making it clear that they aren’t just going to stop with a fast CPU core.

 

[moinmoin]

Speaking of companies moving towards RISC-V....
https://www.anandtech.com/show/17104/imagination-launches-catapult-family-of-riscv-cpu-cores

May deserve its own thread depending how well Imagination executes on that promised schedule.

 

[Saylick]

Apparently, Nvidia is creating a new R&D group in Israel and hiring hundreds of people to accelerate CPU development in light of the changing HPC market. I'm guessing this is the fallback option given that their attempt to acquire ARM is likely going to be rejected. If this truly is the fallback option, it goes to show that buying ARM was firstly about buying the CPU design team and less about saving ARM.

https://videocardz.com/newz/nvidia-establishes-a-new-cpu-rd-group-in-israel