[yehaaa]

Exactly, which raises the question why they are interested in buying Arm to begin with. A public statement is in the article:

"With its proposed acquisition of Arm, NVIDIA will be able to turn new AI possibilities into realities much faster."

Which doesn’t make much sense to me, unless they wish to change aspects of Arm that is currently outside of any license.

[ksec]

>which raises the question why they are interested in buying Arm to begin with

Softbank needs to liquidate some of its asset due to the genius work of con-man WeWork losing them tens of billions. The original purchase price for ARM was something like 100 - 120 P/E in 2016. And the current earnings are still 100 -120 P/E with no immediate or short term profits growth. The prospect of the company's fundamentals and future growth hasn't changed since 2016.

Who in the right mind would want to buy a company for P/E 100+ with no visible growth factor?

And since no one wants to buy it, Softbank had to find a buyer. Softbank is one of the largest shareholders in Nvidia. And with its current stock price that was a perfect fit.

Of course that is ignoring Nvidia could have said No. I guess Softbank could decide to liquidate its position on Nvidia instead.

[rrss]

googling "softbank stake in nvidia" gives me "SoftBank Sells Entire Nvidia Stake" from 2019: https://www.wsj.com/articles/softbank-sells-entire-nvidia-st...

Is your information about SoftBank being a major shareholder in nvidia up to date?

[ksec]

Thank You for Fact Checking. Looks like I was wrong about Nvidia ( My memory might have mixed up a few things ). But the ARM P/E point still stands.

Feeling bad for spreading misinformation. :(

I will need to double check on Son, Softbank and Nvidia again.

[sitkack]

I think the outcome should be that Arm should have less control over who has a full license with the ability to add instructions etc.

It is somewhat ironic to me that, it is the software layer made available by gcc and clang that actually makes these billion dollar cpu vendors viable.

[lima]

> It is somewhat ironic to me that, it is the software layer made available by gcc and clang that actually makes these billion dollar cpu vendors viable.

CPU vendor usually contribute to both GCC and Clang these days.

[kevingadd]

NVIDIA has already released customized ARM chips, like Denver: https://en.wikipedia.org/wiki/Project_Denver

"Denver's binary translation layer runs in software, at a lower level than the operating system, and stores commonly accessed, already optimized code sequences in a 128 MB cache stored in main memory"

[stjohnswarts]

When they can just buy ARM licenses and develop the same technology? That seems rather disingenuous.

[IshKebab]

They might want to make changes to the standard ISA which only ARM can do currently.

But it's nVidia. No way they are going to do anything good.

[colejohnson66]

Doesn’t the ARM ISA license already allow for extensions? I know Apple’s chips are full of them.

[wmf]

Apple's extensions are non-standard. Nvidia could make changes to the standard itself, not extensions. How much money this is worth I don't know.

[IshKebab]

Sure but those wouldn't be standard instructions that other people have to implement.

[colejohnson66]

Is a custom core required to support the full standard instruction set? I thought I could license the rights to have my own, say, A53, and then I’d be able to customize it (as long as I don’t call it an A53 core). Is that not the case?

[IshKebab]

I wouldn't be surprised if their licensing terms require your chip to pass some kind of compliance test for the standard ISA, but probably only the licensees know that. I'm sure you can't use the ARM trademarks if it isn't compatible thought - that's pretty standard.