Qualcomm wants to buy ARM and it may need Samsung's help | Hacker News Reader

Qualcomm wants to buy ARM and it may need Samsung's help

sammobile.com

46 pointswalterbell59 comments

Title is pretty out of context. From TFA

> Qualcomm would be interested in forming a consortium of sorts to acquire ARM. “We’re an interested party in investing,”

Relatedly, the CEO has mentioned that ARM going public would make him buy a bunch of shares, given his faith in the continued success of ARM. As is the case of any "bull case".

Forming a consortium or buying public shares is not, as the title implies, a singular company buying ARM.

Depending on the participants and mechanics, a consortium could actually be the best long-term outcome, in regards to ARM.

(I'd still rather not have Qualcomm and their "business™ practices" involved…)

Why dont they go public? after the nvidia fell through i cant imagine qualcomm acquiring them. It seems like ARM is in a really strong position for the next decade or two. Would be an instant BUY rating by most brokers IMO.

Please no. Publicly trading companies creates one of the worst incentives for poor long-term business decisions. ARM is the kind of company that we can't as a society afford to become short-term profit oriented.

>It seems like ARM is in a really strong position for the next decade or two.

What? They just had a massive layoff which culled 15% of their staff?

In this climate, no way in hell they would be rated as an instant buy.

Their revenue growth and revenues in general are tiny what $2.8 billion, and mostly restricted to licensing/royalties.

They were gonna be bought by nvidia for what 40-60 billion dollars?

They are also competing in the space with massive players a la Nvidia, AMD, Intel who are looking to eat their lunch.

The landscape is shifting by adopting the Big Little structure, adding more core counts, and a much bigger prioritization towards dedicated accelerator offload cards. Not to mention an increase in prioritization to RISC V.

I wouldn't put them as an instant BUY.

> Their revenue growth and revenues in general are tiny what $2.8 billion, and mostly restricted to licensing/royalties.

Their business IS licensing and royalties

> The landscape is shifting by adopting the Big Little structure, adding more core counts ...

Arm invented Big/Little in 2011 [1]

Arm SoCs win on core counts [2]

[1] https://en.wikipedia.org/wiki/ARM_big.LITTLE

[2] "Ampere's flagship 128-core Altra Max M128-30 ... packs an unprecedented number of general-purpose 64-bit cores"

Not sure why you're being downvoted for asking a reasonable question but yes, I do think going public would probably be a bad end and shift focus toward immediate returns for shareholders.

Definitely for the next few years, but who knows - RISC-V is catching up scarily fast from what I can tell

Where is RISC-V being used? Do they have some notable recent design wins?

For the most part, RISC-V is being crammed into SOCs for low-end functionality where you'd have to pay ARM a license fee. There's a massive technical alliance with thousands of members and many such SOCs being made.

Wins in SOCs are things like Google's Titan M2 in Pixel 6, or inside various Seagate storage ASICs-- or NVIDIA's stated intent to replace the Falcon controller CPU in all of their products with RISC-V.

Yes, it's just a few percent of the embedded processor market, and all the current microcontroller use is very low end right now.

Isn't NVIDIA's new firmware controller chip on their cards (the one that allows you to use the new open source linux drivers) RISC-V as well?

Yes-- that's the Falcon / now NV-RISCV. https://www.techpowerup.com/291088/nvidia-unlocks-gpu-system...

Of course, if NVIDIA ever does actually acquire ARM, that might change architectures again.

Don’t think this will ever fly. According to [^0] Qualcomm has 38% of the Mobile AP (application processor) market including apple, which is mostly if not all ARM. Exclude apple (they don’t sell chips), and it is 50%.

[^0] https://news.strategyanalytics.com/press-releases/press-rele...

ARM's IP is much broader than just Mobile AP.

It is also worth pointing out the percentage you quote is Revenue Share, not unit volume. Qualcomm's AP business isn't exactly in as good position as many think it is.

The current antitrust doctrine is immediate harm to consumers. Apparently only 1 quarter of lower prices is enough to decide no harm. I don't think antitrust prevents this.

I'd prefer Nvidia, were I forced to pick a torture to endure.

See: "Qualcomm shook down the cell phone industry for almost 20 years"[1] (and then the 9th District let em them away with it, the dogs!).

[1] https://arstechnica.com/tech-policy/2019/05/how-qualcomm-sho... https://news.ycombinator.com/item?id=31557324 (251 points, 3 days ago, 125 comments) https://news.ycombinator.com/item?id=20055517 (528points, 3 years ago, 300 comments).

Does it matter which company, though, given that they are run by humans who might change jobs or retire?

And isn't greed what drives most non-technical upper-management folks, because "greed" is what they are hired for?

> Does it matter which company, though, given that they are run by humans who might change jobs or retire?

Eh, if you have an existing big organizational apparatus built around IP extortion... you may be more of a danger than someone else who'd need to tool up.

Yes, because companies can "own" property. The fact that humans change jobs or retire inside one has no impact on this.

Greed may be what they are selected for, but I don't think it's necessarily obvious that's what drives them. It's also not necessarily obvious that this is the most robust mechanism to achieving sustained success.

"All companies are greedy so all choices are pointless" is a pretty nihilistic perspective. Do you really think every company engages in all the same practices that are identically detrimental to the commons? I'm almost as far from a fan of capitalism as you can get and even I don't think that.

You somehow believe that having only one corporation will yield better results for the society, instead of many companies competing in a free market? (Where you can also compete if you think your better)

I'm having trouble seeing how you got from my post to this question, so I'm really not sure how to answer it.

You said you are far from being fan of capitalism. Free market means as many companies as needed. The opposite is one company only controlling everything, most commonly irl in form of communism.

The opposite of a vacuum is the center of a black hole, and yet we manage to live in neither.

At any rate a deeper conversation about this would go pretty far off topic, so I'm going to leave your strawmen and conflation of several different ideas and axes into one binary choice alone.

Yeah that’s why i said believe in. Whether you tend to believe if fewer is better, or more is better. To me it seemed like you belong to the former, and thats why i made a specific question which you are actively avoiding answering.

His point is rather that given competition allows certain for profit enterprises to do more right than others hence giving us choice we better make use of to let/get the greedy fringe (or majority of them) out.

I have no problem with that. My response was in his statement that he does not believe in free markets.

Not a he or him, thanks.

Sure thing. You are just as welcome.

FWIW, greed is what drives ARM right now anyways. People seem to forget that ARM is not a community-owned endeavor.

There's different business models behind greed, though.

You can be an innovator running a license business, with a "friendly" stance towards licensees as you work to grow a massive market. ARM's mostly played this game.

Or, you can be a rent-seeking, litigation-happy shop that seeks to acquire many types of IP to force all market participants to license from you; unilaterally determine payments based on ability to pay; and threaten to tie up indefinitely in litigation anyone who doesn't pay the extortion. This has been Qualcomm's bread and butter.

Or, anywhere inbetween.

I don't really think there's much of a difference, contextually. ARM still collects rent, and most of their "innovation" is contributed by their licencees. Pretty much the only thing ARM can peddle at this point is their core designs and ISA, and both of those things are starting to show their age now. Maybe 10 years ago ARM had a "friendly" stance towards their customer base, but nowadays they're frankly no better than Qualcomm. If you want to make a usable ARM computer today, you need to have ISA extensions out-of-the-box and probably design your own cores unless you want to saddle your customers with 8x M53's.

I agree that not all of these companies are equal, but ARM right now reminds me of Intel in 2016. You can only rest on your laurels for so long, sufficiently motivated competitors will be sure to remind you again and again.

> Maybe 10 years ago ARM had a "friendly" stance towards their customer base, but nowadays they're frankly no better than Qualcomm.

I'm sick of this kind of comparison, like I detail here: https://news.ycombinator.com/item?id=31578523

Qualcomm systematically shook down everyone in entire industries with a web of technologies, threatening "well, even if you didn't use ____, you'll also violate ____ and ____. And, we'll cut off your access to important chips in ___, ___ ,and ___."

Between the two, ARM is an absolute pleasure to work with.

> If you want to make a usable ARM computer today, you need to have ISA extensions out-of-the-box and probably design your own cores

Samsung just moved from proprietary cores in flagships to Cortex-X + A710 + A510. Ditto for Snapdragon 8. Graviton is using ARM Neoverse-N1/N2. Actual cores from ARM proper are a bigger share of usage than almost anytime before, with Apple being the main remaining exception.

I think we'll have to agree to disagree, then. I don't think your other comment was right either: people typically give Apple more scrutiny because

A. they are the larger company

and

B. they have much more of our personal data than Meta does

In this case, I think the comparison is apt because the only thing that matters is if the current IP owners are abusing their power. In this case, I think ARM and Qualcomm would wind up to be about equally as tyrannical. If you think ARM's slate is clean, you should read into the neverending drama that is ARM China. Nobody is necessarily good here.

> If you think ARM's slate is clean, you should read into the neverending drama that is ARM China. Nobody is necessarily good here.

I don't understand how having a regional subsidiary going off the rails on their own has anything to do with the typical experience of a licensee.

> Pretty much the only thing ARM can peddle at this point is their core designs and ISA, and both of those things are starting to show their age now.

Not sure what else you expect Arm to be ‘peddling’ given this is and basically always has been their business.

If you were ARM's competitors, and or otherwise wanted to see ARM weakened to open up supposed market opportunities, one of the best ways to do it would be to form a consortium ownership, in which ARM will suffocate. The added bureaucracy and distractions will damage ARM, it stands zero chance of coming back out of that arrangement as good of a company as it was going in (the group will endlessly bicker, as it erodes operationally they'll begin discussing changing the ownership arrangement, as some of the owners will want it to actually succeed - that move will come too late to matter much).

I hope not, Qualcomm's strategy is to stagnate with inefficient chips

I was hoping Apple would buy them or AMD

With Qualcomm, ARM will die

> I was hoping Apple would buy them

God I hope not. Apple would try to prevent everyone else in the industry from using ARM IP and would lead to either a schism in ARM architectures or everyone jumping on something else (maybe that's what would get RISC-V to take off?).

> maybe that's what would get RISC-V to take off?

Export controls to China alone will do it.

Why would they? they would face backslash and probably an investigation

I'd hope they contribute for a better ARM ecosystem for everyone using their experience crafting M1

Apple have a perpetual architecture license from ARM, and having significantly expanded from it they have no reason to purchase ARM (other than to be anticompetitive)

> Qualcomm's strategy is to stagnate with inefficient chips

What do you mean? Seems like they're constantly trying to catch up with Apple's chips.

countless number of chips, no significant improvements over the generations and expensive af

and since Apple M1 came out 2 years ago, they are still unable to offer equivalent

> and since Apple M1 came out 2 years ago, they are still unable to offer equivalent

The fact that they can't keep up with their big competitor doesn't seem like evidence that they deliberately want things to stagnate.

IIRC Apple is an architectural licensee but Qualcomm is not (though they were ~6 years ago). Back in the day, Qualcomm's designs were probably more competitive than the reference ones are today.

Disclosure: I'm a Qualcomm employee.

Well they just purchased Nuvia and are expected to ship Nuvia based cores in 2023 so we should see if they've caught up.

Never attribute to malice that which is adequately explained by stupidity.

To be fair, only Apple could pay more cost for chip.

ELI5: After the Nvidia deal fell through, why does Qualcomm think they have a shot? What makes them different in the eyes of US regulators? I'd argue they're business is more monopolistic than Nvidia.

So is this going to be like the Java purchase where nobody sees the value in the thing until someone comes and turns it into a cash cow?

Is Java a cash cow for oracle?

It won't last long but yes. Recent anecdote, we purchases 1 or 2 years of licenses for the oracle JVM across thousands of hosts to give us time to migrate "properly and safely" to openJDK. Migration has very high man power and risks costs, Oracle knew it very well before their purchase of Sun.

Can someone explain why instruction sets are not open and free to use?

Feels like paying for a programming language (from a software guy’s perspective)

once noone replies, i'll share my understanding, but i didn't have any experience with that i think because they can (demand money from anyone due to intellectual property).

i'm not sure if anyone have challenged it. or i may be wrong and basically they can enforce their demands based on usage of their brand only. or noone really has needed that, because everyone is happy to collaborate with the arm company in some other areas besides licensing, so this very fee is nothing for them.

I am not an expert, but here is the extend of my understanding. What your paying for is not necessarily the instruction set, your paying for other things. So, lets take ARM as an example. When you pay ARM, you are buying a license, which I believe comes with some pre-made stuff you can use for your own ARM chips you design and essentially support. If you and I decide to make a new fancy ARM chip that we will put in thermostats, we pay ARM and they will give us a lot of building blocks as apart of that license. We can choose to not use them, but it is almost silly not to. As another example, Apple and the M1 and A-series chips. Apple did not design every aspect of those processors. They almost certainly used building blocks that ARM gave them and then essentially added on their own little twist. I don't want to diminish Apple's work, regardless, what Apple has shown that ARM chips are capable of with the resources it uses is still amazing. With the license, you also get engineering support from ARM or ARM will help you design and implement some of your features.

Because instruction set designers don't seem to have the same urge to work for free as software developers.

Seems like a good candidate for an international consortium of industry and government to acquire ARM .

Qualcomm is a glorified patent troll, so this would not be great.

Why ARM company is this important?

You can't sell processors compatible with the ARM instruction set without a license from them. They also design popular processor cores.

Reply on news.ycombinator.com