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

 

[NTMBK]

I'm still trying to wrap my head around what the long-term vision is for Nvidia to own ARM...

If they wanted to re-enter the mobile space, they already have an ARM license and they currently made chips that have ARM processors with Nvidia GPUs. If they wanted to go into datacenters with their own CPU, again they already own an ARM license and can develop their own products already. ARM is an IP house which makes money from licensing, not from selling products, so what's in it for Nvidia? Is it because they eventually want to license out an ARM core that has NVLink built-in?

My theory is that Nvidia wants to sell their GPU IP to mobile vendors. They tried this back in 2013, but as far as I can tell nobody took them up on it: https://www.fool.com/investing/general/2013/12/29/nvidias-gpu-licensing-plan-is-clever.aspx

If they can sell it as a bundle deal with an ARM CPU, that puts them in a better position.

 

[blckgrffn]

It looks like ARM is likely in midst or renegotiating the ARM v9 architectural licenses (long term income time!) as the publicized ones are all ARM v8 based, near as I can tell.

Nvidia may also see super computing as a long term way to stay viable in the Data center space (margins++) and being able to control the ISA for ARM based designs is probably useful there.

I mean, there is some long term thoughts about how nvidia can remain viable as dedicated ASICs/custom CPU designs encroach on their GPU compute markets while the rise of the APU on the desktop side might make them less and less relevant as a necessary component there. Like sound cards... you can still buy them if you want, but for 99% of the people out there is no need. I am talking ~5-10 years out. AMD's current APUs are actually pretty darn capable for everything, and DDR5 will no doubt help them be able to play basically everything at minimum/low details @ 1080p at decent frame rates. That's going to further reduce the necessity of video cards sub $200 or so (IMO). Remember when there was an actual market for new cards in the $50 to $125 space because all PCs needed basic video cards for functionality?

Being in the drivers seat for ARM would give nvidia more equal footing with Intel and AMD from a platform perspective (especially as Windows is increasingly ARM portable over the next 5-10 years) and thus also help ensure their long term viability.

You could turn it around and say, can nvidia afford *not* to buy ARM? Without them, where will they be in 10 years?

Debt & Dollars are very cheap right now, if there is a time to take a gamble for long term relevance this might very well be it.

 

[VirtualLarry]

You could turn it around and say, can nvidia afford *not* to buy ARM? Without them, where will they be in 10 years?

Debt & Dollars are very cheap right now, if there is a time to take a gamble for long term relevance this might very well be it.

This... entirely!

 

[Thibsie]

Well, given the openness Nvidia has been showing in the industry, I'd fear for Arm and it's licensees.

 

[Roland00Address]

Any chance that Windows will modify the OS to allow mixed architecture cpus and memory space allowing you to have an Intel CPU and a Nvidia GPU?

Imagine having a midsize GPU but also having 8 maybe 16 big ARM cores on GPUs that cost lets say $200 and more? Maybe $300 or more?

 

[Vattila]

I'm still trying to wrap my head around what the long-term vision is for Nvidia to own ARM...

Very strange move indeed. As already pointed out, there is an obvious anti-competitive issue when it comes to ARM's licensing business. However, I recall MIPS being sold to Imagination, and the IP issue was solved by setting up a consortium to guard the majority of the patent portfolio and protect licensees.

"Imagination agreed to acquire the company for about $60 million in cash. The assets include the operating business as well as 82 of a total of 580 patents, and a perpetual license to the remaining 498 patents. [...] 498 patents were acquired by Bridge Crossing, a new entity created by Allied Security Trust (AST) that apparently has the purpose to manage the patent portfolio."

tomshardware.com

By acquiring the the operational side of ARM (the CPU and GPU design teams, and perhaps some lucrative patents and a perpetual license to the rest), Jensen effectively makes ARM's future core developments Nvidia branded, allowing him to get up to speed competing in the nascent ARM HPC market. Yes, this will kill ARM as a developer of cores, i.e. the microarchitecture side of ARM, as it becomes part of Nvidia. He probably will not license the cores, like ARM currently does, but perhaps establish a semi-custom side of the business, learning from AMD.

On the other hand, the ARM ISA will of course persist. If the deal is structured like the MIPS acquisition by Imagination, it and the related patents may end up guarded by a consortium of licensees — or perhaps SoftBank will retain that side of the business, or sell it to someone else.

PS. In the long term, the ARM ISA will probably be overtaken by RISC-V. If this deal goes through, that transition will probably accelerate.

 

[NTMBK]

By acquiring the the operational side of ARM (the CPU and GPU design teams, and perhaps some lucrative patents and a perpetual license to the rest), Jensen effectively makes ARM's future core developments Nvidia branded, allowing him to get up to speed competing in the nascent ARM HPC market. Yes, this will kill ARM as a developer of cores, i.e. the microarchitecture side of ARM, as it becomes part of Nvidia. He probably will not license the cores, like ARM currently does, but perhaps establish a semi-custom side of the business, learning from AMD.

Except that right there is going to run into anti-competition laws. If they stop licensing cores out, then they've basically killed Qualcomm's Snapdragon business and Samsung's Exynos business. I don't know if either of them could restart their entire CPU design pipeline fast enough.

 

[Vattila]

If they stop licensing cores out, then they've basically killed Qualcomm's Snapdragon business and Samsung's Exynos business.

Right, if those businesses rely on ARM's microarchitecture roadmap for future product plans, they will lose out, if the operational business ends up part of Nvidia. Current cores will continue to be licenced I presume, through current agreements, a consortium, or some other arrangement.

 

[moinmoin]

If Arm stops designing cores and iGPU IPs for easy (not necessarily cheap, but all inclusive) licensing (Arm's current bread and butter business and the reason Arm is so widely used across the whole price range) for Nvidia wanting to turn it into some semi-custom like business, you can bet that most of the current licensees will look for another ISA to switch over after the current Arm gen stops being supported. A MIPS like setup wouldn't help preventing that at all.

Nvidia (everything proprietary even when using open interfaces) and Arm (essentially proprietary, but profiting a lot of open software and open hardware efforts) seem a complete mismatch culturally. But who knows, with increasing success in datacenters (bound to surpass their consumer graphics business) and the huge payoff the acquisition of Mellanox brought (which turned out to be for essentially free due to the excellent financial figures it contributes since) Nvidia may actually want to tackle its huge competitive disadvantage in the datacenter business, its lack of openness in a space that is more and more build on openness. *shrugs*

 

[Roland00Address]

Except that right there is going to run into anti-competition laws. If they stop licensing cores out, then they've basically killed Qualcomm's Snapdragon business and Samsung's Exynos business. I don't know if either of them could restart their entire CPU design pipeline fast enough.

I am not a lawyer but could this be solved by Nvidia releasing their final product and the ability to license designs at the same time and thus there is XX month head start?

 

[Tup3x]

Right, if those businesses rely on ARM's microarchitecture roadmap for future product plans, they will lose out, if the operational business ends up part of Nvidia. Current cores will continue to be licenced I presume, through current agreements, a consortium, or some other arrangement.

I'd imagine that the licensing would probably continue more or less like at the moment (they'd scrap Mali in favour of their own GPU ip) and Cortex-X Custom Program might get higher prio. I'd imagine that NVIDIA might start producing ARM CPUs for HPC. They could also try entering ARM laptop market and maybe even desktop at some point. (If that would happen it would be really interesting.)

 

[Doug S]

Except that right there is going to run into anti-competition laws. If they stop licensing cores out, then they've basically killed Qualcomm's Snapdragon business and Samsung's Exynos business. I don't know if either of them could restart their entire CPU design pipeline fast enough.

Given that Nvidia has 0% share of today's mobile market it would be hard to argue that buying ARM is an anti-competitive act, since all current licenses would remain in force.

They could only stop licensing future cores, Qualcomm and Samsung could keep using A78, A79, X1 or whatever the latest core they have a license for. They would still get the benefit of shrinks so there would still be performance increases, and they still have valid ARMv8 architectural licenses so they could go back to designing their own cores after a few years stuck on A78 or whatever.

I don't see how Nvidia could afford to do this though. ARM's only value is from licensing, if they stopped all future licensing and sold its future CPU designs as Nvidia products only I don't see how that could exceed ARM's licensing income - they'd still be competing with Qualcomm and Samsung, and lacking a modem to integrate onto the SoC they would have a very very hard time convincing any Android OEMs to buy their SoCs. If Nvidia wants to burn a few tens of billions this would be a good way to do it.

 

[HurleyBird]

I don't see how Nvidia could afford to do this though. ARM's only value is from licensing, if they stopped all future licensing and sold its future CPU designs as Nvidia products only I don't see how that could exceed ARM's licensing income - they'd still be competing with Qualcomm and Samsung, and lacking a modem to integrate onto the SoC they would have a very very hard time convincing any Android OEMs to buy their SoCs. If Nvidia wants to burn a few tens of billions this would be a good way to do it.

Well, they wouldn't do that. The worst thing that they might do is vastly increase licencing fees and only license tech that is a bit behind what they use for their own products to give themselves a competitive advantage. Actually, that's probably exactly what Nvidia would do.

 

[DrMrLordX]

he worst thing that they might do is vastly increase licencing fees and only license tech that is a bit behind what they use for their own products to give themselves a competitive advantage. Actually, that's probably exactly what Nvidia would do.

nV would probably make more money just licensing things as before. What they want/need to do is use ARM Holdings to prop up their existing business. It makes no sense for them to strangle competition out-of-the-gate, especially when those competitors only compete with Tegra.

The #1 problem facing them is future platform compatibility for their GPU products. They need assured platform compatibility if they still want to sell cards. Stuff like OpenCAPI, CXL, and CCIX may threaten nVidia. With control over reference ARM-based server platforms (Neoverse), nVidia can assure inclusion of NVlink or something simliar in reference designs. They can also build low-level consumer GPU tech into consumer reference designs as a replacement for Mali GPUs.

 

[Doug S]

Well, they wouldn't do that. The worst thing that they might do is vastly increase licencing fees and only license tech that is a bit behind what they use for their own products to give themselves a competitive advantage. Actually, that's probably exactly what Nvidia would do.

Yet that still gives them no market to sell to. They don't have a modem so they can't sell to the Android market. There's no way Qualcomm would be willing to sell a standalone modem to work with an NVidia ARM SoC, why should they help a competitor be successful? They'd tell those Android OEMs "if you buy an NVidia ARM SoC you are on your own as far as a modem, but we'll sell you our ARM SoC which uses an older ARM core but includes our latest 5G modem".

They can't sell ARM PCs because there is essentially no ARM PC market. They don't really want to sell ARM in the HPC market, they already occupy it with their GPUs. They have no experience at all in the server market.

Can you explain to me how this strategy would benefit NVidia, when they would have no market to which to sell their own products?

 

[DrMrLordX]

The sale of ARM Holdings may be imperiled by rebellion at ARM China:

https://ca.finance.yahoo.com/news/arm-accuses-china-venture-ex-022721526.html

Essentially, the CEO of ARM China has been ousted, and he refuses to step down.

 

[Kenmitch]

The sale of ARM Holdings may be imperiled by rebellion at ARM China:

https://ca.finance.yahoo.com/news/arm-accuses-china-venture-ex-022721526.html

Essentially, the CEO of ARM China has been ousted, and he refuses to step down.

With that mess in the works who'd buy ARM then. Seems like it's a disaster waiting to happen. ARM China looks to be a ticking time bomb!

I guess there's always the conspiracy theory....Devaluing the company while rumors of selling it off....Hmm

 

[NTMBK]

If I were Fujitsu I would be PO'd. They don't have to deal with Oracle any more because they ditched SPARC, and now they're going to have to deal with Nvidia instead!

 

[Doug S]

With that mess in the works who'd buy ARM then. Seems like it's a disaster waiting to happen. ARM China looks to be a ticking time bomb!

I guess there's always the conspiracy theory....Devaluing the company while rumors of selling it off....Hmm

Maybe seeing that mess on the horizon is why Softbank decided they might want to unload it.

 

[beginner99]

Maybe seeing that mess on the horizon is why Softbank decided they might want to unload it.

Which would be ironic as i thought softbank wanted that China branch to begin with.

 

[DrMrLordX]

As far as the "undisclosed breaches of conduct" are concerned, I have heard it has something to do with how Wu is investing his money. I don't know anything more.

 

[coercitiv]

As far as the "undisclosed breaches of conduct" are concerned, I have heard it has something to do with how Wu is investing his money. I don't know anything more.

Details in a Bloomberg article.

Arm Ltd. ousted the head of its Chinese venture after discovering the executive had set up an investing firm that would compete with its own business in China, according to people with direct knowledge of the decision.

Arm China called the allegations against Wu’s fund “inaccurate and misleading.”

The rationale for Wu’s dismissal was also outlined in a document, reviewed by Bloomberg, from SoftBank Chairman Masayoshi Son and Arm Ltd. Chief Executive Officer Simon Segars to Hopu Chairman Fang Fenglei. The document cited breach of contract and the decision to set up the fund.

What followed was a public, acrimonious clash between Arm Ltd. and Wu, who refused to budge and used the Chinese venture’s WeChat account to amplify his defiance. That Arm and Hopu have been unable to assert their will reflects the intricacies of Chinese rules that confer an advantage to Wu as the holder of key registration documents. As the legal representative of Arm China, Wu holds the company’s registration documents and the company seal, or stamp. Changing the legal representative requires taking possession of the company stamp -- something Wu has refused to give up. Arm Ltd. and Hopu could go through the courts, but the process could take years.

Following the investigation of Wu by Arm Ltd. and Hopu, the Arm China board voted 7-1 to dismiss Wu. Given Wu’s refusal to vacate his role, Arm Ltd. is growing anxious over the security of Arm China’s intellectual property, assets and finances, according to the people with knowledge of the matter. If it can’t dislodge Wu in a timely manner, Arm Ltd. would consider suspending support to Arm China. Such a step would be a last resort, said the people.

And last but not least...

Wu himself was cited several times in local media pledging to work with Huawei Technologies last year, when Washington first banned the sale of American software and circuitry to the Chinese tech champion.

 

[uzzi38]

[EXCLUSIVE] Samsung mulls buying stakes in Arm

Samsung Electronics is set to join the acquisition race for the British semiconductor company Arm, which has been put up for sale by its current owner SoftBank, but the Korean tech giant will try to participate in the bid in the form of equity investment, obtaining a small stake in the...

www.koreatimes.co.kr

Sounds like this could lead to a consortium instead of just purely Nvidia-owned

I'll laugh if it was caused by all the players in the industry just worried about Nvidia owning ARM.

 

[Qwertilot]

A consortium of a bunch of the large companies would be much the sanest solution.

They've all potentially got a fair bit to lose, but no specific way to win from monopoly ownership. Or they could just return it to the stock market, of course, making sure to each keep a big share.

 

[.vodka]

I'll laugh if it was caused by all the players in the industry just worried about Nvidia owning ARM.

You already know this, but nV is a company that can't be trusted. Too greedy and anti competitive. Not to mention anti consumer.

You don't want anyone with that mindset even thinking of the possibility of owning the CPU IP provider everyone uses, that's a massive shot in the foot for the *entire* industry, on every market where an ARM core is used.

Consortium sounds like the sanest solution, too.