News Arm Allegedly in Talks with Nvidia and Intel to Become Anchor Investors

Oh boy, is ARM swimming with the sharks covered in blood on this one.
Well... Nvidia is a major ARM customer. So, they have deep vested interests in keeping ARM well-resourced and continuing to produce the IP Nvidia needs.

Intel is the oddball, here. The only interest I can see Intel having in ARM (given that I no longer think it likely Intel will design its own ARM ISA cores) is in ARM as a fab customer. Is that enough? Maybe, if they expect to take a significant chunk of the mobile SoC market from TSMC fabs.

Maybe they should ask their buddies at Apple how nice it is to work with ngreedia and intel.
There's no equivalence between the investing in a company vs. being its supplier.
 
  • Like
Reactions: eYe_Surf
RISC-V is coming for ARM's market, and it knows that it's days are numbered.
It's a long-term threat. Right now, ARM enjoys a lock on the phone market, deep penetration into self-driving and robotics SoCs, and is the only viable cloud alternative to x86. That's not a bad position to be in.

ARM's main problem is that its business model hasn't supported the kind of revenue levels that traditional chip companies have garnered. That's why it's suing Qualcomm over Nuvia's architectural license, in order to try and extract royalties from its downstream customers. It's a "Hail Mary", but if ARM were to prevail, it'd be huge for them (in the short/medium term, anyway).

Its second problem is longer-term: the way its patents have been shown to be vulnerable to exploitation by governments engaged in trade disputes and security concerns. That was a really bad precedent, and I wish ARM had sued the US to get it overturned. I'm sure that put China and many others off the idea of using ARM-based CPUs for their critical industries and infrastructure.

Unfortunately, the Qualcomm/Nuvia lawsuit has poured cold water on any US-based firms introducing new ARM ISA CPUs (chiefly Intel & AMD), after the patent/trade dispute already turned off foreign firms and governments from jumping on the ARM bandwagon.

So, I agree that the brightest days of the ARM ISA are possibly behind it, but not because RISC-V is actually better. Mostly due to business and trade reasons. I've seen no technical analysis which supported the idea that RISC-V is better than AArch64.

BTW, ARM can always pivot and get into the RISC-V IP market. It probably wouldn't be as lucrative for them, but as long as they have the expertise and there's money to be made, it should remain an option.
 
So, I agree that the brightest days of the ARM ISA are possibly behind it, but not because RISC-V is actually better. Mostly due to business and trade reasons. I've seen no technical analysis which supported the idea that RISC-V is better than AArch64.

Jim Keller on How Risc-V Will Change the World

I think the Free & open nature of RISC-V will eat into the ARM market very soon, especially in server.

Why pay ARM when you can do the same for free after you re-factor your code to work with RISC-V and not be "Beholden" to ARM in any way financially or technologically.
 
Why pay ARM when you can do the same for free after you re-factor your code to work with RISC-V and not be "Beholden" to ARM in any way financially or technologically.
Those fall squarely in the category of the "business and trade reasons" I cited.

The downside of RISC-V is a more splintered ecosystem. And we've already seen some chip designers try to push their nonstandard extensions into opensource key toolchains and libraries. I'm not saying it's enough to stall out RISC-V, just that there's no free lunch.
 
Those fall squarely in the category of the "business and trade reasons" I cited.

The downside of RISC-V is a more splintered ecosystem. And we've already seen some chip designers try to push their nonstandard extensions into opensource key toolchains and libraries. I'm not saying it's enough to stall out RISC-V, just that there's no free lunch.
Alot of Upper Management C-Suite "Bean Counters" in those businesses are going to look at:

RISC-V = Free
ARM = Pay them royalties

They're going to push their devs to go with the "Free" option.
 
Alot of Upper Management C-Suite "Bean Counters" in those businesses are going to look at:

RISC-V = Free
ARM = Pay them royalties

They're going to push their devs to go with the "Free" option.
Nothing is free, in that sense. You always have to pay somebody for IP. It's just a question of who and how much.

If you license RISC-V IP, you're still paying someone like SiFive or Tenstorrent for it. You can design your own (as Western Digital did - one of RIISC-V's early adopters), but then you're paying a staff to design your IP.

The point about Western Digital isn't a small one. Royalties are most painful on high-volume, low-margin products. That's why RISC-V has dominated the microcontroller and embedded core market, first. As you move up the value chain, ISA licensing fees potentially become a much smaller piece of the pie. That's why ARM seemed to have a lock on the phone and server markets, until the two game-changing events I mentioned above.
 
Last edited:
Intel is the oddball, here. The only interest I can see Intel having in ARM (given that I no longer think it likely Intel will design its own ARM ISA cores) is in ARM as a fab customer. Is that enough? Maybe, if they expect to take a significant chunk of the mobile SoC market from TSMC fabs.
ARM doesn't make any CPUs itself or did that change?!
If intel makes sure that ARM keeps going strong then everybody that actually makes ARM CPUs could become their FAB customer and that's a lot of customers.
If intel can leverage a deal into being able to make their own ARM designs without paying royalties every time they would be able to attract even more customers.

Intel and ARM already announced a collaboration towards that goal.
 
ARM doesn't make any CPUs itself or did that change?!
If you want to make a SoC using IP from ARM, then ARM needs to have a version of it for the fab node you want to use. If you want to fab a mobile SoC on Intel 3 that includes cores (and probably other blocks) which ARM designed, then they need to have done the layout and verification of those IP blocks on Intel 3.

Each fab node has specific cell libraries and design rules. These are contained in a fab node's PDK.

What's more interesting is the rumor going around that ARM might start selling actual chiplets. Still not entire CPUs, but if someone is building like an AI chip, then they only have to worry about making & verifying a chiplet with their own special components and ARM will provide you a chiplet with the CPU cores. It's a simpler integration model and relieves some of the costs and headaches from their customers. In this case, ARM could even become a direct customer of IFS.

If intel makes sure that ARM keeps going strong then everybody that actually makes ARM CPUs could become their FAB customer and that's a lot of customers.
Yes, if ARM has enough funding to port its IP to Intel's various nodes, then it makes an easy pathway for IFS to get more business.

If intel can leverage a deal into being able to make their own ARM designs without paying royalties every time they would be able to attract even more customers.
It's just an investment, rather than a controlling stake. So, Intel won't be able to make ARM do anything not in ARM's own self-interest (as foregoing royalties would surely would be). Were Intel to pursue a controlling stake, that would get shot down by regulators in a heartbeat.
 
With an IPO for ~20 pc ownership of ARM, it may be possible for companies such as Broadcom to take a 'small' position [under 2 pc ownership] now in order to cozy up to ARM later. And, the public need not know which companies do this -- Texas Instruments? a previously unknown offshore subsidiary of Samsung? Even mighty TSMC?