Excellent and concise summary; if I may the one thing that I think could've helped a little with my attention during the bits where you read from the filing; if you could have a highlight pulled from the exact text you're reading (like a highlighter plugin or a crop/zoom plugin, my eyes wouldn't have to wander to the right portion each time you cut back to it. Also, hopefully this can accelerate some RISC-V development, but I won't get my hopes up too much... it seems like that arch is still a long way towards mainstream viability.
I would love to see more RISC-V development and honestly I'm surprised big Chinese companies have not been investing into it. I'm no expert my any means but I think RISC-V was supposed to be an open source ISA, meaning anyone can use it and improve it without paying for licensing/royalties. Is this because it is a huge investment to go into or is there some sort of legal restriction on RISC-V? Also I hope you get better soon Jeff
@@paulrebellion9548 I think it isn't about legal issue. I think it's more on how to actually implement the ISA into silicon that still is the big challenge for everyone. Even in ARM, you can see how Apple vs Qualcomm competes in performance. They both design their chip using ARMv8 ISA, and yet Apple M1 vs Snapdragon 8cx performance gap is so massive. SiFive also discontinued their HiFive Unmatched (and I guess the Freedom U470 SOC too) and now teaming up with Intel Foundry to design a new "Horse Creek" chip based on Intel 4. I think that shows that it's not the ISA, it's more on everyone still trying to gain some expertise in how to actually laying them into silicon. I bet everyone are tinkering with RISC-V ISA in their lab and has some sort of prototype there, be it the Chinese, India, Russia, Japanese or even western players like AMD/TI/Infineon. And I guess they just haven't found the best recipe yet to make the chip performant enough against their own x86/ARM design.
@@paulrebellion9548 I have a guess, maybe others know better and can correct me: Architectures usually aren't as 'simple' as they say they are on the label. For example, x86 CPUs haven't been x86 CPUs since the AMD K5 and one of the pentiums in the sense that they decode instructions from x86 machine code and convert them in to instructions that run in an internal RISC like processor. In intel speak this 'real instruction set' is 'micro ops'. Strange people will tell you this is why x86 uses more energy but really that's a lie as even in the pentium 4 it was only 2-3% of die for that extra logic. Also, we can look at the architecture of the CPU and say, OK, 32 bit x86 has the EAX register, the EBX register the ECX register, buuuut really those are called 'architectural registers'. The real CPUs can have an array of registers much larger than the architectural registers that allows the CPU itself to do stuff like re-write your crappy code on the fly via register renaming and other tricks. ARM surely has much of this stuff going on too, all of it behind some very impressive patents and IP protections. Meanwhile, RISC-V is an open source ISA meaning you can take what's there, but what's there is in some respects less advanced than 1990's AMD or Intel designs, making RISC-V perfect for 'high end microcontroller' work, but not a lot more. If you want to implement a super advanced design that will respect the RISC-V ISA but under the hood do all the stuff that is needed to reach current performance standards then it's allowable to build on the ISA, but I'd expect lawsuits can be expected if you infringe on patents from the established players. TL:DR; I expect RISC-V to replace ARM, MIPS, and x86 CPUs in embedded devices that aren't performance critical, but not a lot more.
@@paulrebellion9548 they have, it goes slow. Espressif has and is investing in riscV and their chips can be found all over IoT market which is going to dominate everything in the years to come. All the stupid knick-knacks that shouldn't be on the internet but some reason people want on the internet :^)
If Arm prevails and they still don't get to an agreement with Qualcomm, then I wonder how quickly all the Arm ALA holders that are startups will pivot to another architecture. Sure Apple would stay, and the big ones who are using their ALAs substantially, but the smaller players looking at IPO or Buy-out are going to be watching with lots of interest.
That’s exactly my point that ARM could be 💩Ing in their own mess kit as the saying goes. The value add ARM can provide in supporting customers should have been the approach.
To what you said about doing mostly hardware reviews, i for one find your analysis and Op-Eds on another level and the content that lets me think of STH as the #1 source for industry news for many years now, Patrick! No knock on your great reviews, but your analysis is what huge companys pay millions of dollars for, and we get it for free! There can never be too much STH analysis content!
RISC-V won't get anywhere unless they start releasing open source peripheral IP and drivers, and even then, competing silicon will probably still be cheaper
If everyone modifies the CPU, it is much less likely that any of the variants will slide far enough down the cost curve far enough to be a real challenger to volume merchant parts. Besides, we are already seeing the first working silicon for disaggregating the multicomputer so that what we now call a datacenter will actually be just one gigantic machine with many millions of processors. Transistors are cheap enough to put 50 BILLION of them into a desktop and that’s without counting storage. So how hard are we willing to work to save or share them? Non-volatile memory is coming that reads a little slower than DRAM, writes about 30% slower, doesn’t need refresh or sustaining power, and doesn’t wear out. How is that going to change processor and system architecture? We have bigger things to think about than everyone creating new instructions. I can see it now: GITFAB - where a build actually carves-up a wafer…
If you were to make a list of how not to make your suppliers happy, pretty high on the list would be to lobby to block their merger agreement, and then to suggest you want to control your supplier by proposing you and other large customers should take significant equity stakes in the supplier. Qualcomm is being pretty aggressive towards Arm and the last thing Arm wants is to be a small, independent company forced to get almost all its revenue from a few giant customers, especially if those customers carve it up for their own purposes.
It wasn't a merger, it was a "Nvidia buys ARM". Also the only ones displeased by the above are the ARM owners, which is SoftBank or some other japanese investment/pension fund
@@marcogenovesi8570 That's a merger. It's irrelevant if Arm was or wasn't "the only ones displeased", because it's Arm that's looking out for Arm and the Arm ecosystem. Not Qualcomm, not Google, not Microsoft. Those companies are all looking out for themselves. And they most certainly don't want strong suppliers, they want weak and compliant suppliers. How do you know if small Arm customers were pleased or not by the merger? When did you hear from them? What about the potential Arm customers that haven't formed yet? But you mentioned a very strong opposition to the dead merger. It's off topic, but I think it's misplaced. As far as popular opinion of tech enthusiasts, so many always want everything for free, unless they are supplying rather than receiving the service, of course. They thought Nvidia buying Arm would make Arm more expensive and now it seems so many of them are chanting risc-v when Arm defends itself. Let me ask you this. Samsung has 35% of the Android OS market share like Qualcomm has 30% of the market share of Arm mobile SOCs. Would Android OS be better off as the "Switzerland of OSes" instead of under Google? Wouldn't you be afraid of Samsung twisting, warping, or splitting the Android ecosystem? Because Samsung does not have the Android ecosystem's interests in mind as they have no stake in it. They only care about using it. But Google is strong enough to keep Android together. And note that Google competes with Samsung and makes Pixel phones. Why doesn't that fall apart? Which companies would be happy or not for Android to be a small independent company instead of under Google? Then note that Android doesn't need the same ongoing investment that Arm does. Qualcomm would love nothing more than Arm to be an IP shell company that doesn't have the money to build effective cores which allow small companies to compete with Qualcomm's Nuvia cores. So when you say what others outside Arm wanted, ask yourself who they were and why they wanted it.
@@mattmexor2882 That's not a merger, it's an acquisition. A change of ownership where the buyer retains its brand and company. The current owner is SoftBank and those are the only ones that are relevant in the argument. I don't understand why would ARM care about who the end owner is, it's a publicly traded company, they answer to shareholders anyway regardless of who the shareholders are. Also, NVIDIA is a (big) client of ARM, they use ARM cores in their embedded and automotive designs. So your statements that ARM should be specially butthurt about Qualcomm "trying to control them" by opposing a deal where a client is buying their supplier outright? Bich please. ARM sued Qualcomm because of other far more valid reasons, probably they got fed up of Qualcomm's shenanigans with their licensed stuff which is a valid reason, not because they somehow are pissy because Qualcomm offered to make a industry-wide consortium to have a shared control over ARM instead of letting them get bought by a single company.
@@marcogenovesi8570 Bich please? Are you 3? So what if Nvidia is now a customer of Arm? How does that refute my point? If Nvidia owns Arm they then have an interest in Arm for Arm's sake. A customer does not have that interest. Google is not going to disrupt Android OS to benefit their Pixel phones. But Samsung would. A small supplier needs to be protected from their customers. Look at the data center. Who else besides the hyperscalers can make much money from their business? Big companies like Intel, AMD, Nvidia, and Broadcom. What is the profit margin Apple's small suppliers are able to maintain? Now consider what happens to Arm's TLA business if Qualcomm can inherent an ALA chip and expand it across their entire lineup and cancel then the TLA purchases. Do you think Arm will have the money for more TLA R&D? Do you think they will be able to prevent complete consolidation of their customers into the giant haves and the little have-nots? Without access to capable TLA's startups won't be able to emergeto compete with the giants. Regarding "merger", Softbank was to take a significant stake in the combined company. It's a merger. The distinction you're trying to make is irrelevant, though. You're trying to use it pejoratively to say that Arm was somehow being destroyed or corrupted in some way. But it was just fear-mongering. I have a feeling among popular commentators those who worked to spread that fear are the same ones who now call for risc-v when a soon-to-be independent Arm defends its IP. I find it odd that you seem to argue that Arm management has no stake in Arm's future. I never said that Arm sued Qualcomm because Qualcomm worked to block the merger. I never remotely implied that. I turned around the quote from the servethehome video on how a supplier usually wants to avoid treating a customer by highlighting how the relationship was already strained to help explain Arm's decision. Qualcomm's aggression towards Arm goes beyond the effort to block the merger. That effort is simply one example of the aggression. Other examples are Qualcomm's "solution" to a cash-strapped independent Arm and Qualcomm's pursuit of the Nuvia merger even after being warned by Arm. Arm's decision to sue Qualcomm did not depend on Qualcomm's attitude or disposition and I never suggested it did. In fact I explicitly said that Arm's decision was made in order to secure Arm's own future by allowing it to continue to service and grow its ecosystem. Qualcomm, on the other hand, has interest in the opposite. I own Qualcomm stock, by the way, and have no financial interest in Arm, lest you think I am somehow being biased here even though my reasoning isquite clearly laid out. Arm's lawsuit hascost me money. But for Arm it is something that they must do and it should be clear what Qualcomm's end game is here. Qualcomm has always been very aggressive in this way and they've gotten away with it because of their geoolitical importance regarding China. That's one good reason to own Qualcomm stock.
Happy Birthday dude! I hope that you have/had a great day, make sure to take the opportunities to spend time with family & friends because those moments are precious and you never know when you may never get the opportunity to do it again! I love the work that you do and I want you to know that your content is invaluable to us Technology Fanatic's who don't have the time or opportunity to work with the kind of hardware that you do, I love your videos! Take care dude and all the best to you :-)
On whether Arm would know if Qualcomm was developing new cores: as I understand it, all custom Arm cores must still be submitted & verified by Arm first to ensure they satisfy the ISA standards (so that all ARM-ISA CPUs work right with ARM-ISA compilers & software). And so, if Qualcomm had submitted no designs, then Arm would've known that QC's ALA was being unused for new core development. This licensing problem seems weird: Apple, surely one to be litigious and fearful of their designs, had no problem sending its vaunted IP to Arm, but Qualcomm chickened out?
Apple was part of the consortium that formed ARM. At one time it was its major shareholder. This is the reason ARM is called Advanced RISC Machines instead of the original working group name of Acorn Replacement Microprocessor. Apple sold its share of ARM when it nearly went bankrupt and needed working capital to buy enough time to regroup.
0:18 Never steal or kill another person´s dog. You never know if it is John Wick, who will kill your gang and ruin your business. Very good explanation. I like the break down of the different parts of the complaint.
Either way this case goes, Arm claiming ownership over Nuvia's IP and suing is a major red flag. Anyone considering Arm's ALA should seriously consider switching to RISC-V after this, especially startups. Then next Nuvia that gets acquired won't go for as high since the acquiring company will have to pay Arm a cut too
I think it's a good video, but you could have put a context in the beginning (as you did in description). Qualcomm bought Nuvia and is trying to use Nuvia's IP, based on Arm IP together with it's own (Qualcomm's) IP that is based on Arm IP to create new cores and Arm has lots of clauses that prevents that. Because, since I for some reason didn't hear about Nuvia being bought I was wondering what the hell was going on for first 9-10 minutes. But it's a good video and as a person that visits lawtube way too often, I would say that you did a good job of explaining things (which is not easy. IP, Trademarks, Patent Law's etc. those are among the most complicated). Just for future reference, you might start with the context. If of course you plan to do more videos like this - I think you could at least follow this case.
Who could have thought that arm would fight tooth and nail for future earnings? Especially after the nvidia deal broke down Honestly if I'm starting a cpu startup I'm going RISC V - no license bs at all (which some insiders dissatisfied with the nuvia acquisition did)
RISC V could be worse if you are a startup, RISC is base, math, graphics, network, and usb/storage controllers aren't there, do you want a separate chipset that handles a a bunch of stuff like the old north/south bridge controllers, how do you talk to them? none of that is included. There was talk of using ISA since the patents have run out on that, but the industry tools are copyrighted not patented, so you are still going to need to licence those instead. its all a mess if you haven't got IP in the game already.
@@Duncan_Campbell except that there are many existing designs that use RV with all sorts of controllers via pcie and bus protocols (mostly AMBA which is open source from ARM)
@@eladdvash3052 Yes, 100% this. The RISC-V ISA has absolutely nothing to do with the surrounding IP/peripherals like graphics accelerators, network MACs, storage controllers etc. Most hardware designs have all of those peripherals sitting on a generic, often-times open standard bus. It's entirely possible to just swap an ARM CPU core out for a RISC-V one without changing any other part of your design and have it "just work", so long as both of your CPU core designs support the bus standard you're using. For example, if I had an entire SoC built around an ARM core, utilising an AMBA bus, I could very easily just replace the ARM core with any other core that interfaces with an AMBA bus. Now sure, the memory and cache controllers are usually core specific, and likely couldn't be reused, but those components are par for the course when you're designing SoC's, and aren't a major blocker when it comes to developing RISC-V based hardware.
This sounds like basic companies arguing over contracts. The court will ask one simple question "Nuvia, would you have been able to create your IP without Arm's IP" They would answer "No" and lose the case.
Except, the ARM IP that Nuvia used to create their own IP Qualcomm already have a license to, remember, Qualcomm already have a (likely larger in scope) ALA and TLA license agreement with ARM, so the assumption being made by Qualcomm here is that Nuvia's IP falls into a subset of licenses they already had from ARM.
@@joshdau well, if thats the case then theirs not much else to it. but usually contracts are independent of each other. even if they say the same thing. qualcomm would have to renegiate a new contract.
Another component that may go into licensing fees beyond the ALA and TLA is the total addressable market that each player was going for and expected sales revenue. So when Nuvia was negotiating their licensing they might have said like ok we are going after the data center, we expect say $3b in revenue for this product over the next 5 years and ARM does the license fees accordingly. Qualcomm can't use that design with that projected revenue across their product stack for a different revenue target without owing ARM money.
Ha. But it's criminal defense! Which I might add is something that is tragically lacking from the highest court in USA, save for a single exceptional new Justice.
THANK YOU for giving an overview of this whole cluster-fudge.... when I heard about this the other day I had NO CLUE what any of it meant or what was going on... PLUS I had no idea about ANY of these companies or that ARM even licensed their designs the way they did... You have educated me SO MUCH just with this video (and in MANY of your other videos too) - Thank you!!!!
Excellent synopsis of the ARM & Qualcomm dispute. It is a great shame that ARM and Qualcomm couldn't resolve this through negotiation. I was looking forward to seeing Nuvia designs appearing not only in data centres but also Windows / Linux laptops that could compete with Apple Silicon, but once the courts get involved this has the potential to add significant delay and further strain relations between ARM and Qualcomm. I suspect the only immediate beneficiary from this is Apple.
13:50 It's actually opposite (I have information from around 2015) it was cheaper to have finished design, but cost increase with more freedom to change core, most expensive was architecture license.
I am not the biggest fan of ARMs outreaching licencing model , but Qualcom is not somebody to feel sorry for. This company is known for shadier bussines practices than intel, way shadier. So i do not wonder ARM sues them
How could ARM make a profit without very strict licensing? Plus the more complicated the customisation, the harder it may be for them to verify compatibility. So without knowing the specifics, I don't think we can judge if their licensing is bad or not. It certainly seems to have worked in allowing the huge explosion in smartphones over the years.
I agree, Qualcomm can be shady and they’ve used/licensed some IP I’ve been a part of in the past. I level near some of the founding board members and friends, but their legal team is shady. The board member I dealt with was a very early founder.
Very good article. It would be very interesting to hear your take on qualcomms filing in response to ARM suit. There are quite a lot interesting bits in there too and your take on it will be very good
Put a bit on Twitter this morning on it. The "broadly overlapping" I was very confused why they would leave that in unless Nuvia licensed something that is not covered by Qualcomm's license
Arm, in their filing, seem to imply that royalties are set differently for server CPUs vs PC / tablet / phone CPUs. So that NUVIA perhaps got a great bargain rate because it was only for servers. But now that the target market has expanded, it's violated the contract (not to mention change of control). Notably, Arm claims Qualcomm admitted its license was terminated and then sent a brand-new design for Arm Ltd.'s verification.
However, it seems the assumption being made on Qualcomm's side is that they already had ALA and TLA's signed for the broader scope of products they typically ship (ie. mobile SoCs, server processors, DPUs etc). So by Qualcomm's reasoning, it seems like they believe ARM has already given them permission to develop their own custom cores in those target markets, and so the Nuvia's IP it acquired fell into a subset of the ALA/TLA licenses it already held. Personally, I feel like ARM is grasping at a technicality in some contract somewhere, and they've seen this as an opportunity to either protect their interests (if Qualcomm move to using Nuvia-based cores, then Qualcomm will no longer need to pay the TLA license fee on every Snapdragon SoC they sell), or to extract additional money from a bigger player.
Considering your back-ground and the impact of futur ARM CPU' 'incarnation'', it was a great idea to produce this feed. Thanks ! You made a great documentory... ;)
Thanks for explaining all of this Patrick. I followed you for mini-PC reviews which i enjoyed (I went with the Beelink GTR5 and it rocks ;). i've stayed because i also enjoy learning about the networking devices and i found this very interesting. I don't think this news would have made as much sense to me if i hadn't watched this video of yours explaining both the technical and legal aspects so well.
Qualcomm played their hand poorly. They should have considered the valuation of Nuvia with regard to ARM's IP. The acquisition negotiations should have included consent with ARM from the beginning. Qualcomm knowingly chose to take the backdoor to ARM's IP without consent first. ARM chose to slap Qualcomm's PEEPEE very hard because of this.
It's a bit sad that we might never see the nuvia design in a product because of this legal battle. On the other hand, while Nuvia was very promising for the future of high performace arm products, by the wording of the license, it doesn't look like they were going to make desktop products anytime soon, so their impact as an "Apple Silicon Competitor" might have been greatly overstated.
Qualcomm won't lose. The worst case for Qualcomm will be to pay ARM a good amount. But Qualcomm has probably the best legal team in silicon valley. And the fact they let the situation get to this point means that they're pretty confident that they can win.
I think ARM is in the right, without their IP, there is no Nuvia IP, Nuvia chose to license ARM as a base to work off of and under that agreement ARM was going to make a profit of sales of the combined product IP, that’s how ARM makes its money, is it not?
Thanks for the breakdown, and happy birthday 🥳 Not sure what to make of this so far. My gut feeling is Qualcomm is in the wrong. They don’t exactly have a clean history, and I don’t think Arm would take legal action on a customer lightly 😅
The first time absorbing a competitor does not translate you extending your reach? Also where are gov't agencies involved in anti-trust and competition laws on this? Is buying Nuvia killing competition in this segment?
What I don't understand is what is it that ARM has that everyone else chooses to license from them vs. build their own from scratch? That was the question I was hoping you'd answer the entire time I was watching the video.
I am not an IP lawyer, so I don't know anything about change-in-control clauses and their limits, but from my limited understandig I don't see a basis for this lawsuit as both Qualcomm and Nuvia had the needed architecture license. I don't see why it should make a difference if a core was designed by an acquired company or in-house. I assume ARM's payments are structured in a way that they might get a fixed sum up front but more heavily rely on a cut from every sold chip. They could simply charge the rate intended for Qualcomm instead of Nuvia which should be a plus for ARM as Qualcomm's market reach is much bigger. If not, that was simply a negotiation gone wrong for them with long-term consequences alienating their customers.
I don't think that ALA licensees are allowed to sell their cores to others without explicit permission from ARM. Qualcomm had to deal with this when they licensed the Centriq core to the JV in China. I cannot disclose how this was resolved with ARM.
The details of each agreement matter. I wouldn't normally sign an agreement that restricts where I can apply my developed product and no partner should expect such with significant value add at little cost. It seems like ARM is trying to grab money and claim ownership of the acquired companies development and double charge QUALCOMM but I can't be certain of that. This should absolutely make people reconsider basing development investment and return using ARM as a platform.
This is interesting, but it also sounds like a 2nd year standard contract law case study for law school students where you would have a mock trial amongst the students to be able to present your arguments, in class, in a competition for who can present and argue their case better. As far as contract law goes, the thesis of the case is actually relatively academic, but the theory is applied in practice. As to who's "right" -- I can definitely unless there's a suddenly fluke judge, I could see Arm winning this tort/contract law case. Qualcomm needs to be able to enforce both it's IP and their IP license agreements. If they can't uphold their IP and license agreements, then their value as a company would basically plummet/evaporate. To me, the case is fairly cut and dry. I came up with an idea. You can license the idea from me and/or you can license the ability to use the *principle* that the idea is built off of and then go off and do your own thing. But you can't make a product with that principle and then sell yourself as a company to someone else so that they can get away with NOT having to pay the licensing fee for licensing the *principle* of the idea. Makes sense to me. Qualcomm should've known better. It's an anti-competitive, anti-capitalist behaviour, but that IS your capitalism at work for ya. Simple, easy, straightforward.
Except Qualcomm have already payed for a license to ARM IP, they had one before they bought Nuvia. ARM are essentially trying to charge them twice for something they already bought just because they bought another company which also bought the same thing. The only way I can see this suit making sense is if Qualcomm's existing license only covered servers and not desktop/laptops or something like that. I could understand if the law was written this way for some reason but in that case the law needs to change, not Qualcomm's behavior. Not that I really believe in our current laws or capitalist system to begin with of course. I don't necessarily like Qualcomm as a company but this area I think they are in the right; I also want this project to succeed for my own reasons.
@@harryhall4001 Different license though. That would be the difference, for example, between having the license for the right to manufacture vs. making a knock-off from the original design. (cf. Amazon making a knockoff of the "nice design" bag) You can have the license to use Arm's intellectual property to implement Arm's designs which is separate from the license where you can use the Arm SPECIFICATION and then come up with your OWN design. Suppose that I designed a recipe for a cookie that meets a very specific taste profile that you can ONLY get with MY specific recipe -- I can sell you a license that you can use MY recipe so that you can make your own cookies and they might be in different shapes, etc., but fundamentally, it uses my recipe. But if you were to then use how I DEVELOPED that recipe to make your OWN recipe for your OWN cookie that has either identical or nearly identical taste profile (such that the customers won't be able to qualitatively tell the difference between my original recipe vs. your new recipe) -- but if you didn't license the right to make your own recipe (as stipulated in our contract when you wanted to license my recipe to get you going/get you started), whilst yes, you did pay for *A* license, but you didn't pay for *THE* license (to make your own). And that's the difference here. Nuvia licensed the right to make their own recipe from Arm. Qualcomm, as far as I can tell, only licensed the right to use Arm's recipe. And in both contracts/license agreements, I am sure that Arm would have a clause in there about the transferrability of the licenses in the event of an ownership change (i.e. the change-in-control or change-of-control provision which as Patrick noted, IS quite standard for these types of contracts). You can't have another company pay for the license to make their own recipe and then you buy them up and call that design yours when you didn't come up with that recipe yourself because if the other company was say a much smaller startup and therefore; and let's make up some numbers where let's say if Qualcomm were to license the right to make their own recipe and it costs $100M because Qualcomm is a bigger company with deeper pockets. Whereas with Nuvia, which is more like a startup, doesn't have that kind of money, so Arm agrees to license the right for Nuvia to make their own recipe for say $10M instead. Well, now, Qualcomm looks at Nuvia and goes "I like what you've got there, so I'm going to buy you up, which includes the design that you came up with for $50M (yes, I'm making up numbers)". Then Qualcomm, instead of paying $100M for the license to make their own recipe, just saved $40M by buying up Nuvia for $50M where Nuvia spent $10M for the license which allows them to make their own recipe. If Qualcomm were to let this happen, then ANY company that wants a new Arm design would spin off their design division into a "startup company", claim that they don't have the kind of money that Qualcomm has, and argue for a lower licensing cost, develop the recipe, and they get acquired by the original, parent company, arguing that they should've never spun off that division in the first place. If they EXPLICITLY said that they wanted to do that, I am pretty sure that that's actually illegal (fraud). But if they filed it under "normal business operations" or "an oopsie", and this wasn't some grand master plan all along, even though it could very well have been, then this move could be legal and Arm would lose out on revenue as a result of allowing this to happen just ONE time. (It's just like how with some rental agreements, you can't sublet without the explicit written approval/consent/authorisation from the party who owns the rental unit that you're renting from. So if you rent from Some Housing Corp. and you want to sublet your apartment, and your rental agreement says that you can't do that unless Some Housing Corp. grants you the permission for you to do so (why would they? They can lose out on the income that *YOU* would be making), and you decide to sublet anyways, then you just violated the terms of your rental agreement and Some Housing Corp. can kick you and the person to whom you subletted, out from their property/apartment. This is the same jist.) This is why I think that as a tort/contract law case, it's quite simple/easy/straightforward prima face.
@@ewenchan1239 I didn't even have to read all of that to tell you have gotten something very wrong. Qualcomm didn't just have the license to manufacture existing ARM CPU designs. They also have the Architecture license that lets them make new designs. Similar to the license Nuvia had to make their own core. It tells you they have this because Qualcomm also tried to design their own core not based on ARM's design for the server market but failed wheras Nuvia apparently succeeded. You apparently didn't watch the video properly before commenting any of that wall of text.
@@harryhall4001 "I didn't even have to read all of that to tell you have gotten something very wrong." "If I don't write it, that's my fault. If you don't read it, it's yours." "They also have the Architecture license that lets them make new designs." You clearly didn't read Qualcomm filed complaint. You might want to read it before commenting. "26. Even though Qualcomm has an Arm ALA, its prior attempts to design custom processors have failed. Qualcomm invested in the development of a custom Arm-based processor for data center servers until 2018, when it cancelled the project and laid off hundreds of employees. 27. Qualcomm's commercial products thus have relied on process designs prepared by Arm's engineers and licensed to Qualcomm under Arm TLAs. Discovery is likely to show that as of early 2021, Qualcomm had no custom processors in its development pipeline for the foreseeable future. To fill this gap, Qualcomm sought to improperly purchase and use Nuvia's custom designs without obtaining Arm's consent." (Source: images.law.com/contrib/content/uploads/documents/394/11978/Arm-v-Qualcomm.Arm-complaint-1.pdf, p.7) "You apparently didn't watch the video properly before commenting any of that wall of text." You KNOW that Patrick EXPLICITLY talked about this at timestamp 7'11" (cf. here: th-cam.com/video/_3uJPWBtxkw/w-d-xo.html). You MIGHT want to review that part of the video and the complaint PDF itself before writing such an idiotic remark which shows your poor grasp of the nature of the subject matter. Patrick also EXPLICITLY mentions the failure of Qualcomm's Centriq 2400 custom Arm processor that Qualcomm DID design under its own ALA, and that's what Qualcomm's ALA was used for, and NOT for the ACQUISITION of Nuvia's design, nor Nuvia's ALA. (Remember: the license agreements ARE NOT transferrable under a change-in-control or change-of-control of the company that licensed the license agreement occurs. When a change in/of control occurs (which happens with mergers and acquisitions (M&As), said merger/acquisition will trigger the change-in-control (or change-of-control) clause as stipulated in the license agreement/contract whereby the custom design that Nuvia produced via it's ALA with Arm -- that license agreement cannot be transferred. Further, Patrick also EXPLICITLY tells you in the video (if you had bothered to watch it at all), that the acquisition itself was subject to approval/authorisation by Arm, also as stipulated in the contracts/agreements. (cf. here: th-cam.com/video/_3uJPWBtxkw/w-d-xo.html) (Which you will also find on paragraph 37, on pages 10-11 of the complaint PDF.) You might want to watch the video again more carefully next time before making such idiotic comments. And whilst you're at it, also read the ACTUAL complaint PDF yourself as well (given that it's only a mere 48 pages which you should be able to polish in half an hour or less).
@@ewenchan1239 I read your other comment just to make sure but that's not at all what you argued. You are trying to claim ARM needed Nuvia's license to make a custom ARM core. Then in the second reply you correctly state they already have a license from when they were designing their own core. You're contradicting literally contradicting yourself. My understanding is this: ARM have an architecture and IP blocks. They licensed both to Qualcomm, the IP blocks for phones and tablets while the architecture was licensed to make their own server CPU design, these were probably separate deals that happend at different times. Nuvia also got the architecture license to make their own core, albeit at a discounted rate because they were smaller. Nuvia succeeded in doing that while Qualcomm failed their own design. Qualcomm buys Nuvia and it's core design. ARM rightly removes Nuvia's license as they are entitled to do so but Qualcomm already has their own license they tried and failed to use earlier. Sincs Qualcomm already paid for their license at the full price before they even purchased Nuvia so how can they be breaking any rules?
Happy birthday! What is arm seeking? More fees? Cease development? Are they planning on bringing this ip to market with another partner? As u suggest in the intro, hard to see a productive partnership after this
Belated happy birthday!! 🎉👏🎂😁 Hope you had a great day! Great explanation and summary of this case and I hope you do follow-ups. I probably would not mind hearing about similar cases or news.
Loved the breakdown! To us mere mortals it does seem like a tiny distinction, but I guess if you're an IP-only company, that tiny distinction means big $$$. 👍
Thank you so much for this video, this was fascinating to watch. I’ve always been curious about the way that ARM licenses it’s IP but would never grok the articles I was happening upon. You put it in a clear and concise way and in a way that actually helps the layperson in tech figure out which product is ALA or TLA (aside from your mention of RPi and Apple)
what a mess. But I guess ARM can define (limited to thier contract) thier license. Is ARM architecture really that superior? Why not pour RnD into RiscV.
Sounds like ARM gave Nuvia a discounted ALA license and they would charge much more for Qualcomm to use the ALA and the license contract was with Nuvia as a startup in the data center market. Not with Qualcomm. Therefore Qualcomm must re-license the ALA with ARM and the terms and cost of the ALA license would be significantly different. Licensing fees and terms of use likely vary from licensee to licensee. This is all ironic when you look at Qualcomm's own predatory licensing model to it's customers.
Qualcomm must have been told that they would have a loophole when they bought Nuvia. They think they can threat ARM financially so they will got better term than play straight. Now it's depend on ARM. Will they go the furthest, destroy Qualcomm and take colateral damage? For Qualcomm, it can be only one way: - you've got greed and a bad lawyer - or you've got greed and a good loophole lawyer. My prediction? Wouldn't it be more interesting if we see dimensity in flagships instead of snap?
I hope they work it out because I thought Nuvia was going to provide stiff competition to intel/amd x86_64 in the pc space. Kinda odd that arm would do this since Nuvia stood to make arm look good.
@@ServeTheHomeVideo yeah but don’t cut off your nose to spite your face. What if Qualcomm and other licensors abandon ship to risc-v. Then less revenue for arm. This whole thing is odd. Qualcomm is probably the largest or one of the largest licensors of arm… snapdragon is everywhere in android devices.
Happy birthday Patrick! I don’t think this is the first time this (ARM ISA licensee getting bought out) has happened, didn’t Cavium already have a custom ARM core when Marvell took it over? And these core design startups (Nuvia and Rivos) are probably formed with the goal of getting acquired by a major silicon vendor, ala P. A. Semi. I’d be surprised if Nuvia didn’t try to get favorable license transfer terms from ARM. p.s. Thanks for the detailed Hot Chips coverage on your site!
Happy birthday Patrick! Just to be clear, do you fully expect this lawsuit to go through? Reading through the complaint, I had formed an opinion that this is ARM last phase to force QC to fold back into negotiation.
The court will usually attempt to do mediation first, which will involve a court appointed mediator to assist in further negotiations. If those negotiations go nowhere, then the case is likely to be heard before a judge.
If the comments in paragraph 22 translate to specific clauses about nontransferable licenses in the Nuvia ALA, then Qualcomm's due diligence team was not very diligent (and hence Qualcomm is going to have a very bad time when they get to court).
It sounds like Arm alerted QC before they acquired Nuvia and they proceed regardless. Anything can happen, but it seems QC will either be paying Arm a new license fee to utilize Nuvia's IP or just dead in the water. Attempting to leverage and aggregate their license with the Nuvia one also screams bad faith which is probably why this has gotten to the point it has.
On TLA vs ALA cost structure my research suggests that TLA is much cheaper generally up front to get rights to use the designs, but has high royalty rate once going out. The ALA is other way around and is quite expensive up front, but then royalties are lower or maybe can even be negotiated to not exist in some market segments since the upfront cost and development costs are so high it need amortized into products for them to be viable sometimes...but seemingly all are unique per customer. Generally I think ARMs real issue here is I don't get the impression they have the right limitations clear on Qualcomms and probably others who might buy out smaller firms in the larger firm ALAs to prevent them from rolling unlicensed by termination 3rd party designs into the still kosher after the deal ALAs and Qualcomm likely has very favorable royalty stuff for their Smartphone and other ALAs they have used in past until they got behind and now ARM wants some higher royalty rates to make up for the TLA stuff Qualcomm will be dropping. So I am sure it creates a bit of animosity generally when you pay however much up front to get different ALA deals with clear cost structures and such and no limitations on whether the designs need to be 100% original or somehow acquired if you go and spend $1.8B also and then are told no, no your ALA and all the money and agreements you made with it are off the table and only this little firms CiC provisions rule all. Those CiCs only invalidate the smaller firms licenses in my mind and so it all then revolves around what Qualcomms licenses say and if there are no provisions preventing qualcomm from buying 3rd party ARM designs and rolling them into their own ALA, that's all tough nuggy for ARM for not writing a better license with qualcomm to prevent them from shopping startup designs to use with their ALA.
Soooo, should qualcomm sue the old Nuvia rights holders for false representation and evaluation maybe get the SEC involved? If the one thing your company does, no longer functions/is not the property of that business any more after an acquisition, is that company worth anything more than the physical properties/building leases.
This should have been seen during diligence. The action you describe would be only if Nuvia actively hid the ALA. Even if they did, then it would not change Qualcomm's obligations to Arm or Nuvia's obligations to Arm.
It seems Arm would make much more of it's money from the TLA licences via "Royalties" ($$ / unit sold), and the bulk of it's money from ALAs via "consulting" on the custom design work of the licencee. If that's the case, I don't see why the "designs" can't be transferrable, even if the licences are not. Now, it all depends on the wording of the licence agreements, and that's likely to be the focus of the case... Or, if it is explicitly worded that the designs are also not transferrable, Qualcomm would argue "fair use", as the bulk of the work was done by Nuvia employees while paying Arm consulting fees.
If my assumptions are correct, and it's a matter of who has the rights to Nuvia's designs, I usually look at who paid for them, and who got paid for them. Nuvia paid it's own employees AND Arm for "consulting" on those designs, so they should have the rights to the designs, and those designs should be transferrable via purchase of the company. Now, those designs are worthless without a licence, but Qualcomm already had TLA & ALA licences of their own, so they should be covered there. If anything, the "tailoring" of licences is the only thing to stand on for Arm, and was obviously the sticking point.
If ARM didn't invalidate the License either prior to or at time of purchase, Qualcomm should hold them liable for the price they paid for Nuvia, since ONLY ARM is benefitting from Nuvia's work and expansion into the Datacenter market (as ARM stated, with preferential treatment due to EXPANDING ARM itself).
The ARM licenses are based on the potential market for the resultant core. Nuvia's market was for servers, Qualcomm's is for everything under the sun (much much larger). So if Qualcomm wants to think it can buy a small company to get an ARM license on the cheap, I think they are sadly mistaken. Basically, Qualcomm is trying to screw ARM out of their slice of the pie.... and get the ARM license in effect for much less than it would have cost if it had been sold to Qualcomm directly in the first place.
Makes me curious what Amazon and Apple think about all this... Both companies are big enough to acquire the entire Softbank group if they had too, but Apple in particular might have quite a lot at stake here.
I think part of the issue is that Qualcomm is famous for playing real hardball in patent and other IP matters. It has been asserted that Qualcomm negotiated very attractive terms when they first went with Arm and I suspect Arm got tired of feeling abused. I was told that the ALAs are for a specified marker segment and specifically address no presumed transferability between market segments by the same licensee, and certainly not in a change of control. I suspect Q thought they could muscle through the issue like they have been able to do in other disputes. Arm’s negotiating position is now very different and Arm saw no reason to acquiesce. I suspect there are a number of folks privately cheering for Arm. I was actually surprised when Softbank bought Arm instead of Apple. However, that “pass” may have been a shrewd decision on Apple’s part to keep Arm in neutral hands. Nvidia’s attempt was worrying to some as it could be too easy for Nvidia to be a proxy for another party.
I don't think it's certain Qualcomm wasn't working with a pretty broad ALA with Scorpion, Krait and one gen of Kyro. That takes you all the way back to 2007 or even 2006 they would have had to get an ALA or some ALA and TLA mix to do those. But generally wouldn't be surprised if in that early era of ARM catching on as a mobile thing with Apple and everyone else taking phone level computing mainstream and there being extreme competition driving massive ARM growth that didn't give out some generally broad and do what you want ALAs. Now sure as they started to target markets outside of their early growth that was predominantly in more consumer electronics style stuff with more very application specific OSes and such they maybe started to see opportunity in the ALA where they wanted it to be less generic to have better royalties from an ALA where there was margin to do so, but really somehow doubting they were thinking that far ahead when they were still running on mostly consumer electronics style devices and very niche narrow market mobiles and still nowhere near the sorta designs needed for servers or any other higher margin stuff at the time. To me all these realities mean that Qualcomm probably has a VERY favorable ALA to do with as they please so long as they pay whatever menial royalties they had promised. I'm sure they have to do some new stuff with like ARMv8 and v9 and such, but even at v8 back in 2011 I doubt they had to start making special per segment stuff like a lot of the more recent efforts to get out of mobile and CE space to higher margin spaces have caused. But, to me that means ARM has very little leverage on ancient ALA holders like Apple/Qualcomm until someone wants an ARMv9 license. It's even possible ARM was seeing the better margin potential by the time someone like Apple purchased PA Semi and wanted to use the team for ARM and got more lucrative ALA royalty agreements than they have with a bit older ALA holder like Qualcomm. And I am sure the TLA heavy royalty nature of qualcomm since they gave up on Kyro custom efforts has been a boon for cashflow. Doesn't mean they can just try to demand more from the standing ALA if Qualcomm uses a 3rd party to drop designs in an ALA without any terms that disallow that.
You could hazard a guess that Qualcomm is trying to buy itself out of the jam of not being able to create a true competitor to Apple by buying Nuvia. A short cut that is understandable given that for the past few years, it has been having problems putting out quality chips. Heck, Mediatek is coming out with better designs now. While Qualcomm seems to be all over the place. Mid range chips that are actually slower than the last year model. Overheating flagships.
I'm inclined to agree with ARM since they are not using this as a method of extortion due to Nuvia being sold to another party, but because of the other party looking to take advantage of the merger.
"because the other side wants to take advantage of the merge" what? Why, then, does the "merge" exist as a commercial procedure between companies? For the sake of a joke?
@@arround1 Hans, the Nuvia license only covers designs for Nuvia products sold for datacenter use and Qualcomm is trying to use it to produce their own cell phone products and sidestep their agreement with Arm. It's quite a clear case. Qualcomm violated the terms of the Nuvia contract with Arm.
Qualcomm does not own all of the ALA IPs to be able to make use of the Nuvia IPs, so should they opt to not to purchase the correct licensing the Nuvia IPs are likely tainted as how can one separate out any of the work derived using the IP from ARM, even if it all there was were simple interfaces between ARM's IP to Nuvia's IPs, who's to say that engineers from Nuvia looking at ARMs IP was not inspired by ARM's work and elements unintentionally made it into Nuvia's IP. I dont see how Qualcomm can get out of this without paying for those licenses. Also, Happy Birthday!
It makes sense. Qualcomm would be skipping out on loads of licensing fees, especially when the owners of ARM are upside down in their investment. Since the acquisition is already completed, I don't see how it gets undone.
Qualcomm tried to do an end-run around ARM's IP licensing model by purchasing Nuvia and then announcing that it intended to go beyond Nuvia's ALA agreement on datacenter servers to other areas WITHOUT ARM's agreement. What ARM has now done is basically make Qualcomm's purchase of Nuvia worth nothing by cancelling the licensing agreements AND cancelled Qualcomm's TLAs as well. That IS a big deal. Qualcomm called ARM's bluff that it wouldn't sue its largest customer and it could get messy from here. Qualcomm's dependence on ARM's technology IP means if that it cannot release any more new processors on ARM's tech IP, which could allow competitors to do so undermining Qualcomm's market share in key areas. How it could get undone is simple: Qualcomm agrees to pay a hefty fee in return for a) Nuvia being allowed to continue developing for datacenter hosts contingent on b) Qualcomm disallowing itself from using those processors in any other device or market without ARM's consent and allowing c) Qualcomm to get its TLA license back. Anything else will simply fill attorney's pockets with money while Qualcomm loses market share waiting for a judicial ruling and appeals process to take place.
I can't imagine how would a huge company like Qualcomm made an acquisition without sorting out the IP and licenses. I saw arguments that Qualcomm might already had an ALA license from Arm. First of all, Arm is now claiming Qualcomm does not have one allowing them to do what Nuvia is doing. Secondly, in my opinion, if Qualcomm does have one, they would have used their resources to design their own CPU cores, instead having market share eaten by MediaTek and the poor performance compared the Apple's Arm CPUs.
As far as I understand, the problem is that ALA licences aren't general licenses but specific to particular product development. Qualcom's ALA license was specific to the development of Centriq 2400, so it can't not be used for Nuvia's IP. Qualcom should have negotiated a new ALA specifically for the development of new cores based on Nuvia's IP.
this is getting very risc-y for arm. i wouldnt be amazed if apple is behind this wrench being thrown into themerger. apple is known for doing shady stuff like this.
So apple is risking here for doing ARM cores design for desktop and laptop? Or their ALA and TLA were laid for that use in the first place but only used for mobile devices in the begining? I'm not sure if Apple being the company it is would be eager to pay for ARM's blessing for desktop and laptop since the beginning or their relationship with ARM when they only made mobile SOCs. I think this move from ARM will make some companies to review their TLAs and ALAs to not being caught between a rock and a hard place.
To say don't sue your customers does not and should not apply to a company like Arm. I believe Qualcomm needs Arm, more that Arm needs Qualcomm. This is the core business of Arm and no one is allowed to 'Bend It Like Beckham'.
ARM is going to find themselves of being accused of “Working for Apple and Tencent.” This could be a publicity nightmare for ARM. It creates incentives to create non-Arm processors. ARM is killing success.
ARM wants a piece of that 1.5 billion and they don't produce their own chips, so of course they were not going to allow transferring (a non transferable) license without a cut. It also seems like they would have to defend their licensing otherwise it could be used against them in other cases..? It'll be interesting to see QCs legal reply. I also wonder if Arm gave Nuvia a sweetheart deal as a startup.
IMHO ARM is cutting a branch it's sitting on. I suspect a fair amount of startups in this space will consider going with license-free ISA like Risc-V instead of ARM. Since most startups are created with a specific aim of selling them to "big boys", this licensing nonsense can make those startups un-sellable.
Overall, none of the rhetoric used by ARM in the complaint matters. The only thing that matters is what was actually in the licensing contract that Nuvia had. Until we see that, no informed opinions are really possible.
"you are not allowed to use any of the technology created by Nuvia" What? How can you defend that? ARM can't require people forget what they've learned. ARM can't compel a company to delete their own data. Nuvia could created a patch file that contains only their changes to the ARM source materials, call it 100% original work. Then apply the patch to Qualcomm's source materials. ARM is in the wrong here. They keep hanging on the "licenses do not transfer" aspect of the contract. Qualcomm doesn't need to transfer the license because they have their own. Qualcomm purchased knowledge and employee contracts. Qualcomm used their knowledge and their workers to create a chip design with their license. I've heard nothing to suggest any different.
Imagine if Qualcomm had to halt all ARM processor production..... So many companies (even huge ones) just use others 'stuff', and have no idea how to do much of anything.
Excellent and concise summary; if I may the one thing that I think could've helped a little with my attention during the bits where you read from the filing; if you could have a highlight pulled from the exact text you're reading (like a highlighter plugin or a crop/zoom plugin, my eyes wouldn't have to wander to the right portion each time you cut back to it.
Also, hopefully this can accelerate some RISC-V development, but I won't get my hopes up too much... it seems like that arch is still a long way towards mainstream viability.
I had that thought about 30 minutes after this went live. Good feedback Jeff. I hope you are feeling better.
I would love to see more RISC-V development and honestly I'm surprised big Chinese companies have not been investing into it. I'm no expert my any means but I think RISC-V was supposed to be an open source ISA, meaning anyone can use it and improve it without paying for licensing/royalties. Is this because it is a huge investment to go into or is there some sort of legal restriction on RISC-V?
Also I hope you get better soon Jeff
@@paulrebellion9548 I think it isn't about legal issue. I think it's more on how to actually implement the ISA into silicon that still is the big challenge for everyone. Even in ARM, you can see how Apple vs Qualcomm competes in performance. They both design their chip using ARMv8 ISA, and yet Apple M1 vs Snapdragon 8cx performance gap is so massive.
SiFive also discontinued their HiFive Unmatched (and I guess the Freedom U470 SOC too) and now teaming up with Intel Foundry to design a new "Horse Creek" chip based on Intel 4. I think that shows that it's not the ISA, it's more on everyone still trying to gain some expertise in how to actually laying them into silicon. I bet everyone are tinkering with RISC-V ISA in their lab and has some sort of prototype there, be it the Chinese, India, Russia, Japanese or even western players like AMD/TI/Infineon. And I guess they just haven't found the best recipe yet to make the chip performant enough against their own x86/ARM design.
@@paulrebellion9548 I have a guess, maybe others know better and can correct me: Architectures usually aren't as 'simple' as they say they are on the label. For example, x86 CPUs haven't been x86 CPUs since the AMD K5 and one of the pentiums in the sense that they decode instructions from x86 machine code and convert them in to instructions that run in an internal RISC like processor. In intel speak this 'real instruction set' is 'micro ops'. Strange people will tell you this is why x86 uses more energy but really that's a lie as even in the pentium 4 it was only 2-3% of die for that extra logic. Also, we can look at the architecture of the CPU and say, OK, 32 bit x86 has the EAX register, the EBX register the ECX register, buuuut really those are called 'architectural registers'. The real CPUs can have an array of registers much larger than the architectural registers that allows the CPU itself to do stuff like re-write your crappy code on the fly via register renaming and other tricks. ARM surely has much of this stuff going on too, all of it behind some very impressive patents and IP protections. Meanwhile, RISC-V is an open source ISA meaning you can take what's there, but what's there is in some respects less advanced than 1990's AMD or Intel designs, making RISC-V perfect for 'high end microcontroller' work, but not a lot more. If you want to implement a super advanced design that will respect the RISC-V ISA but under the hood do all the stuff that is needed to reach current performance standards then it's allowable to build on the ISA, but I'd expect lawsuits can be expected if you infringe on patents from the established players. TL:DR; I expect RISC-V to replace ARM, MIPS, and x86 CPUs in embedded devices that aren't performance critical, but not a lot more.
@@paulrebellion9548 they have, it goes slow. Espressif has and is investing in riscV and their chips can be found all over IoT market which is going to dominate everything in the years to come. All the stupid knick-knacks that shouldn't be on the internet but some reason people want on the internet :^)
If Arm prevails and they still don't get to an agreement with Qualcomm, then I wonder how quickly all the Arm ALA holders that are startups will pivot to another architecture. Sure Apple would stay, and the big ones who are using their ALAs substantially, but the smaller players looking at IPO or Buy-out are going to be watching with lots of interest.
That’s exactly my point that ARM could be 💩Ing in their own mess kit as the saying goes. The value add ARM can provide in supporting customers should have been the approach.
@@PlanetFrosty IP is now, IP lookout!!
Apple has licensed their instruction sets, and implemented their own architecture
It's not ARM, it's it's own thing, that supports ARM
To what you said about doing mostly hardware reviews, i for one find your analysis and Op-Eds on another level and the content that lets me think of STH as the #1 source for industry news for many years now, Patrick! No knock on your great reviews, but your analysis is what huge companys pay millions of dollars for, and we get it for free! There can never be too much STH analysis content!
Thank you for the kind words Steven
@@ServeTheHomeVideo Only the truth! Well earned! :)
The analysis is what I follow for. It's not something you get in other media.
We already had thousands of reasons to use RISC-V, but sure let's have another one.
Maybe out of spite Nvidia will make a RISC-V “super chip”
RISC-V won't get anywhere unless they start releasing open source peripheral IP and drivers, and even then, competing silicon will probably still be cheaper
If everyone modifies the CPU, it is much less likely that any of the variants will slide far enough down the cost curve far enough to be a real challenger to volume merchant parts.
Besides, we are already seeing the first working silicon for disaggregating the multicomputer so that what we now call a datacenter will actually be just one gigantic machine with many millions of processors. Transistors are cheap enough to put 50 BILLION of them into a desktop and that’s without counting storage. So how hard are we willing to work to save or share them? Non-volatile memory is coming that reads a little slower than DRAM, writes about 30% slower, doesn’t need refresh or sustaining power, and doesn’t wear out. How is that going to change processor and system architecture?
We have bigger things to think about than everyone creating new instructions. I can see it now:
GITFAB - where a build actually carves-up a wafer…
If you were to make a list of how not to make your suppliers happy, pretty high on the list would be to lobby to block their merger agreement, and then to suggest you want to control your supplier by proposing you and other large customers should take significant equity stakes in the supplier. Qualcomm is being pretty aggressive towards Arm and the last thing Arm wants is to be a small, independent company forced to get almost all its revenue from a few giant customers, especially if those customers carve it up for their own purposes.
Amen
It wasn't a merger, it was a "Nvidia buys ARM". Also the only ones displeased by the above are the ARM owners, which is SoftBank or some other japanese investment/pension fund
@@marcogenovesi8570 That's a merger. It's irrelevant if Arm was or wasn't "the only ones displeased", because it's Arm that's looking out for Arm and the Arm ecosystem. Not Qualcomm, not Google, not Microsoft. Those companies are all looking out for themselves. And they most certainly don't want strong suppliers, they want weak and compliant suppliers. How do you know if small Arm customers were pleased or not by the merger? When did you hear from them? What about the potential Arm customers that haven't formed yet?
But you mentioned a very strong opposition to the dead merger. It's off topic, but I think it's misplaced. As far as popular opinion of tech enthusiasts, so many always want everything for free, unless they are supplying rather than receiving the service, of course. They thought Nvidia buying Arm would make Arm more expensive and now it seems so many of them are chanting risc-v when Arm defends itself.
Let me ask you this. Samsung has 35% of the Android OS market share like Qualcomm has 30% of the market share of Arm mobile SOCs. Would Android OS be better off as the "Switzerland of OSes" instead of under Google? Wouldn't you be afraid of Samsung twisting, warping, or splitting the Android ecosystem? Because Samsung does not have the Android ecosystem's interests in mind as they have no stake in it. They only care about using it. But Google is strong enough to keep Android together. And note that Google competes with Samsung and makes Pixel phones. Why doesn't that fall apart? Which companies would be happy or not for Android to be a small independent company instead of under Google? Then note that Android doesn't need the same ongoing investment that Arm does. Qualcomm would love nothing more than Arm to be an IP shell company that doesn't have the money to build effective cores which allow small companies to compete with Qualcomm's Nuvia cores. So when you say what others outside Arm wanted, ask yourself who they were and why they wanted it.
@@mattmexor2882 That's not a merger, it's an acquisition. A change of ownership where the buyer retains its brand and company.
The current owner is SoftBank and those are the only ones that are relevant in the argument.
I don't understand why would ARM care about who the end owner is, it's a publicly traded company, they answer to shareholders anyway regardless of who the shareholders are.
Also, NVIDIA is a (big) client of ARM, they use ARM cores in their embedded and automotive designs. So your statements that ARM should be specially butthurt about Qualcomm "trying to control them" by opposing a deal where a client is buying their supplier outright? Bich please.
ARM sued Qualcomm because of other far more valid reasons, probably they got fed up of Qualcomm's shenanigans with their licensed stuff which is a valid reason, not because they somehow are pissy because Qualcomm offered to make a industry-wide consortium to have a shared control over ARM instead of letting them get bought by a single company.
@@marcogenovesi8570 Bich please? Are you 3?
So what if Nvidia is now a customer of Arm? How does that refute my point? If Nvidia owns Arm they then have an interest in Arm for Arm's sake. A customer does not have that interest. Google is not going to disrupt Android OS to benefit their Pixel phones. But Samsung would. A small supplier needs to be protected from their customers. Look at the data center. Who else besides the hyperscalers can make much money from their business? Big companies like Intel, AMD, Nvidia, and Broadcom. What is the profit margin Apple's small suppliers are able to maintain? Now consider what happens to Arm's TLA business if Qualcomm can inherent an ALA chip and expand it across their entire lineup and cancel then the TLA purchases. Do you think Arm will have the money for more TLA R&D? Do you think they will be able to prevent complete consolidation of their customers into the giant haves and the little have-nots? Without access to capable TLA's startups won't be able to emergeto compete with the giants.
Regarding "merger", Softbank was to take a significant stake in the combined company. It's a merger. The distinction you're trying to make is irrelevant, though. You're trying to use it pejoratively to say that Arm was somehow being destroyed or corrupted in some way. But it was just fear-mongering. I have a feeling among popular commentators those who worked to spread that fear are the same ones who now call for risc-v when a soon-to-be independent Arm defends its IP.
I find it odd that you seem to argue that Arm management has no stake in Arm's future.
I never said that Arm sued Qualcomm because Qualcomm worked to block the merger. I never remotely implied that. I turned around the quote from the servethehome video on how a supplier usually wants to avoid treating a customer by highlighting how the relationship was already strained to help explain Arm's decision. Qualcomm's aggression towards Arm goes beyond the effort to block the merger. That effort is simply one example of the aggression. Other examples are Qualcomm's "solution" to a cash-strapped independent Arm and Qualcomm's pursuit of the Nuvia merger even after being warned by Arm. Arm's decision to sue Qualcomm did not depend on Qualcomm's attitude or disposition and I never suggested it did. In fact I explicitly said that Arm's decision was made in order to secure Arm's own future by allowing it to continue to service and grow its ecosystem. Qualcomm, on the other hand, has interest in the opposite.
I own Qualcomm stock, by the way, and have no financial interest in Arm, lest you think I am somehow being biased here even though my reasoning isquite clearly laid out. Arm's lawsuit hascost me money. But for Arm it is something that they must do and it should be clear what Qualcomm's end game is here. Qualcomm has always been very aggressive in this way and they've gotten away with it because of their geoolitical importance regarding China. That's one good reason to own Qualcomm stock.
Happy Birthday dude!
I hope that you have/had a great day, make sure to take the opportunities to spend time with family & friends because those moments are precious and you never know when you may never get the opportunity to do it again!
I love the work that you do and I want you to know that your content is invaluable to us Technology Fanatic's who don't have the time or opportunity to work with the kind of hardware that you do, I love your videos!
Take care dude and all the best to you :-)
Thank you much. I hope you have a super day.
On whether Arm would know if Qualcomm was developing new cores: as I understand it, all custom Arm cores must still be submitted & verified by Arm first to ensure they satisfy the ISA standards (so that all ARM-ISA CPUs work right with ARM-ISA compilers & software). And so, if Qualcomm had submitted no designs, then Arm would've known that QC's ALA was being unused for new core development.
This licensing problem seems weird: Apple, surely one to be litigious and fearful of their designs, had no problem sending its vaunted IP to Arm, but Qualcomm chickened out?
I would be unsurprised if apple has a unique agreement with arm given their very long history.
Apple was part of the consortium that formed ARM. At one time it was its major shareholder. This is the reason ARM is called Advanced RISC Machines instead of the original working group name of Acorn Replacement Microprocessor. Apple sold its share of ARM when it nearly went bankrupt and needed working capital to buy enough time to regroup.
It can take years to develop CPU cores...doubting anything needs submitted until it's at some maturity levels.
0:18 Never steal or kill another person´s dog. You never know if it is John Wick, who will kill your gang and ruin your business.
Very good explanation. I like the break down of the different parts of the complaint.
Very important, as why not go RISCv ?
Oh that is a great one
@@ServeTheHomeVideo You know an even better one? Broadcom creates an open core RISCv cpu that has no accelerators. In five/ten years ARM is dead.
RiscV is the way to go, no licence, no nonsense!
Thanks for discussing this. I look forward to watching this play out.
Me too!
Is it messed up that I hope ARM wins because that outcome will turbocharge investment and development of RISC-V?
Either way this case goes, Arm claiming ownership over Nuvia's IP and suing is a major red flag. Anyone considering Arm's ALA should seriously consider switching to RISC-V after this, especially startups. Then next Nuvia that gets acquired won't go for as high since the acquiring company will have to pay Arm a cut too
I think it's a good video, but you could have put a context in the beginning (as you did in description). Qualcomm bought Nuvia and is trying to use Nuvia's IP, based on Arm IP together with it's own (Qualcomm's) IP that is based on Arm IP to create new cores and Arm has lots of clauses that prevents that.
Because, since I for some reason didn't hear about Nuvia being bought I was wondering what the hell was going on for first 9-10 minutes. But it's a good video and as a person that visits lawtube way too often, I would say that you did a good job of explaining things (which is not easy. IP, Trademarks, Patent Law's etc. those are among the most complicated). Just for future reference, you might start with the context. If of course you plan to do more videos like this - I think you could at least follow this case.
That was quite the roller coaster…nice job simplifying the situation!
Thank you John!
More power for RISC V! :)
That once in a lifetime moment when you actually get to put your college degree to use! 😂
On the other hand, Qualcomm is getting a dose of its own medicine that it’s imposed on customers at times. It still is a bad move for the industry.
Who could have thought that arm would fight tooth and nail for future earnings? Especially after the nvidia deal broke down
Honestly if I'm starting a cpu startup I'm going RISC V - no license bs at all (which some insiders dissatisfied with the nuvia acquisition did)
RISC V could be worse if you are a startup, RISC is base, math, graphics, network, and usb/storage controllers aren't there, do you want a separate chipset that handles a a bunch of stuff like the old north/south bridge controllers, how do you talk to them? none of that is included. There was talk of using ISA since the patents have run out on that, but the industry tools are copyrighted not patented, so you are still going to need to licence those instead.
its all a mess if you haven't got IP in the game already.
@@Duncan_Campbell except that there are many existing designs that use RV with all sorts of controllers via pcie and bus protocols (mostly AMBA which is open source from ARM)
@@eladdvash3052 Yes, 100% this. The RISC-V ISA has absolutely nothing to do with the surrounding IP/peripherals like graphics accelerators, network MACs, storage controllers etc. Most hardware designs have all of those peripherals sitting on a generic, often-times open standard bus. It's entirely possible to just swap an ARM CPU core out for a RISC-V one without changing any other part of your design and have it "just work", so long as both of your CPU core designs support the bus standard you're using. For example, if I had an entire SoC built around an ARM core, utilising an AMBA bus, I could very easily just replace the ARM core with any other core that interfaces with an AMBA bus.
Now sure, the memory and cache controllers are usually core specific, and likely couldn't be reused, but those components are par for the course when you're designing SoC's, and aren't a major blocker when it comes to developing RISC-V based hardware.
I don't care.......I just buy Apple products...
This sounds like basic companies arguing over contracts.
The court will ask one simple question
"Nuvia, would you have been able to create your IP without Arm's IP"
They would answer "No"
and lose the case.
What if Nuvia say we would have gone with RISC-V if we'd known Arm would claim ownership over our own IP?
Except, the ARM IP that Nuvia used to create their own IP Qualcomm already have a license to, remember, Qualcomm already have a (likely larger in scope) ALA and TLA license agreement with ARM, so the assumption being made by Qualcomm here is that Nuvia's IP falls into a subset of licenses they already had from ARM.
@@joshdau
well, if thats the case then theirs not much else to it. but usually contracts are independent of each other. even if they say the same thing. qualcomm would have to renegiate a new contract.
Another component that may go into licensing fees beyond the ALA and TLA is the total addressable market that each player was going for and expected sales revenue.
So when Nuvia was negotiating their licensing they might have said like ok we are going after the data center, we expect say $3b in revenue for this product over the next 5 years and ARM does the license fees accordingly.
Qualcomm can't use that design with that projected revenue across their product stack for a different revenue target without owing ARM money.
I would’ve never pegged Patric for a lawyer 😵
Ha. But it's criminal defense! Which I might add is something that is tragically lacking from the highest court in USA, save for a single exceptional new Justice.
THANK YOU for giving an overview of this whole cluster-fudge.... when I heard about this the other day I had NO CLUE what any of it meant or what was going on... PLUS I had no idea about ANY of these companies or that ARM even licensed their designs the way they did... You have educated me SO MUCH just with this video (and in MANY of your other videos too) - Thank you!!!!
Excellent synopsis of the ARM & Qualcomm dispute. It is a great shame that ARM and Qualcomm couldn't resolve this through negotiation. I was looking forward to seeing Nuvia designs appearing not only in data centres but also Windows / Linux laptops that could compete with Apple Silicon, but once the courts get involved this has the potential to add significant delay and further strain relations between ARM and Qualcomm. I suspect the only immediate beneficiary from this is Apple.
Fantastic breakdown, Patrick! Here's a request for more great tech jurisprudence content!!
Happy Birthday 🎂 Thanks for the useful info 👍
Thank you Arun
13:50 It's actually opposite (I have information from around 2015) it was cheaper to have finished design, but cost increase with more freedom to change core, most expensive was architecture license.
I am not the biggest fan of ARMs outreaching licencing model , but Qualcom is not somebody to feel sorry for. This company is known for shadier bussines practices than intel, way shadier. So i do not wonder ARM sues them
How could ARM make a profit without very strict licensing? Plus the more complicated the customisation, the harder it may be for them to verify compatibility. So without knowing the specifics, I don't think we can judge if their licensing is bad or not. It certainly seems to have worked in allowing the huge explosion in smartphones over the years.
I agree, Qualcomm can be shady and they’ve used/licensed some IP I’ve been a part of in the past. I level near some of the founding board members and friends, but their legal team is shady. The board member I dealt with was a very early founder.
Very good article. It would be very interesting to hear your take on qualcomms filing in response to ARM suit. There are quite a lot interesting bits in there too and your take on it will be very good
Put a bit on Twitter this morning on it. The "broadly overlapping" I was very confused why they would leave that in unless Nuvia licensed something that is not covered by Qualcomm's license
Interesting and detailed explanation, nice to see someone shed some light on these more complex legal matters in the industry.
Patrick I love you and your energy in these videos!
Happy Birthday Dude!
Thank you. I hope you have a great day.
Totally jamming on the gaps in IP law and hoping there's a follow up if this goes to trial.
Arm, in their filing, seem to imply that royalties are set differently for server CPUs vs PC / tablet / phone CPUs. So that NUVIA perhaps got a great bargain rate because it was only for servers. But now that the target market has expanded, it's violated the contract (not to mention change of control).
Notably, Arm claims Qualcomm admitted its license was terminated and then sent a brand-new design for Arm Ltd.'s verification.
Yes. Tried to make this simple for folks to understand.
However, it seems the assumption being made on Qualcomm's side is that they already had ALA and TLA's signed for the broader scope of products they typically ship (ie. mobile SoCs, server processors, DPUs etc). So by Qualcomm's reasoning, it seems like they believe ARM has already given them permission to develop their own custom cores in those target markets, and so the Nuvia's IP it acquired fell into a subset of the ALA/TLA licenses it already held.
Personally, I feel like ARM is grasping at a technicality in some contract somewhere, and they've seen this as an opportunity to either protect their interests (if Qualcomm move to using Nuvia-based cores, then Qualcomm will no longer need to pay the TLA license fee on every Snapdragon SoC they sell), or to extract additional money from a bigger player.
Considering your back-ground and the impact of futur ARM CPU' 'incarnation'', it was a great idea to produce this feed. Thanks ! You made a great documentory... ;)
Well, there's a new pitch:
Use RISC-V, not ARM. You won't get sued.
I love STH videos.. though sometimes I think the intro is there to remind me I need coffee :)
I usually record before coffee to slow myself down
@@ServeTheHomeVideo hah - Love the enthusiasm!
Thanks for explaining all of this Patrick. I followed you for mini-PC reviews which i enjoyed (I went with the Beelink GTR5 and it rocks ;). i've stayed because i also enjoy learning about the networking devices and i found this very interesting. I don't think this news would have made as much sense to me if i hadn't watched this video of yours explaining both the technical and legal aspects so well.
Thank you Anya
Qualcomm played their hand poorly. They should have considered the valuation of Nuvia with regard to ARM's IP. The acquisition negotiations should have included consent with ARM from the beginning. Qualcomm knowingly chose to take the backdoor to ARM's IP without consent first. ARM chose to slap Qualcomm's PEEPEE very hard because of this.
woah, wonder how this will play out. Awesome breakdown broski
It's a bit sad that we might never see the nuvia design in a product because of this legal battle. On the other hand, while Nuvia was very promising for the future of high performace arm products, by the wording of the license, it doesn't look like they were going to make desktop products anytime soon, so their impact as an "Apple Silicon Competitor" might have been greatly overstated.
Qualcomm won't lose. The worst case for Qualcomm will be to pay ARM a good amount. But Qualcomm has probably the best legal team in silicon valley. And the fact they let the situation get to this point means that they're pretty confident that they can win.
I think ARM is in the right, without their IP, there is no Nuvia IP, Nuvia chose to license ARM as a base to work off of and under that agreement ARM was going to make a profit of sales of the combined product IP, that’s how ARM makes its money, is it not?
Thanks for the breakdown, and happy birthday 🥳
Not sure what to make of this so far. My gut feeling is Qualcomm is in the wrong. They don’t exactly have a clean history, and I don’t think Arm would take legal action on a customer lightly 😅
Thank you Aaronage
Great piece Patrick
That was an interesting video. Even with your fast talking, I was able to completely grasp all of the concepts of the suit. Good job!
The first time absorbing a competitor does not translate you extending your reach? Also where are gov't agencies involved in anti-trust and competition laws on this? Is buying Nuvia killing competition in this segment?
What I don't understand is what is it that ARM has that everyone else chooses to license from them vs. build their own from scratch? That was the question I was hoping you'd answer the entire time I was watching the video.
I am not an IP lawyer, so I don't know anything about change-in-control clauses and their limits, but from my limited understandig I don't see a basis for this lawsuit as both Qualcomm and Nuvia had the needed architecture license. I don't see why it should make a difference if a core was designed by an acquired company or in-house. I assume ARM's payments are structured in a way that they might get a fixed sum up front but more heavily rely on a cut from every sold chip. They could simply charge the rate intended for Qualcomm instead of Nuvia which should be a plus for ARM as Qualcomm's market reach is much bigger. If not, that was simply a negotiation gone wrong for them with long-term consequences alienating their customers.
A licensor restricting transfers of a licensee is extremely common practice.
@@ServeTheHomeVideo I am hooked to learn what Qualcomm's legal team has to say about their claims. :)
I don't think that ALA licensees are allowed to sell their cores to others without explicit permission from ARM. Qualcomm had to deal with this when they licensed the Centriq core to the JV in China. I cannot disclose how this was resolved with ARM.
Yea, the ARM case seems like a homerun based on their licensing, but a very bad sign in every other regard. Looking forward to hearing QC's reply.
@@spencer5051 There is a price at which ARM would relent, but it may be too high for QCOM!
The details of each agreement matter. I wouldn't normally sign an agreement that restricts where I can apply my developed product and no partner should expect such with significant value add at little cost. It seems like ARM is trying to grab money and claim ownership of the acquired companies development and double charge QUALCOMM but I can't be certain of that. This should absolutely make people reconsider basing development investment and return using ARM as a platform.
This is interesting, but it also sounds like a 2nd year standard contract law case study for law school students where you would have a mock trial amongst the students to be able to present your arguments, in class, in a competition for who can present and argue their case better.
As far as contract law goes, the thesis of the case is actually relatively academic, but the theory is applied in practice.
As to who's "right" -- I can definitely unless there's a suddenly fluke judge, I could see Arm winning this tort/contract law case. Qualcomm needs to be able to enforce both it's IP and their IP license agreements. If they can't uphold their IP and license agreements, then their value as a company would basically plummet/evaporate.
To me, the case is fairly cut and dry.
I came up with an idea. You can license the idea from me and/or you can license the ability to use the *principle* that the idea is built off of and then go off and do your own thing.
But you can't make a product with that principle and then sell yourself as a company to someone else so that they can get away with NOT having to pay the licensing fee for licensing the *principle* of the idea.
Makes sense to me.
Qualcomm should've known better.
It's an anti-competitive, anti-capitalist behaviour, but that IS your capitalism at work for ya.
Simple, easy, straightforward.
Except Qualcomm have already payed for a license to ARM IP, they had one before they bought Nuvia. ARM are essentially trying to charge them twice for something they already bought just because they bought another company which also bought the same thing. The only way I can see this suit making sense is if Qualcomm's existing license only covered servers and not desktop/laptops or something like that. I could understand if the law was written this way for some reason but in that case the law needs to change, not Qualcomm's behavior. Not that I really believe in our current laws or capitalist system to begin with of course.
I don't necessarily like Qualcomm as a company but this area I think they are in the right; I also want this project to succeed for my own reasons.
@@harryhall4001
Different license though.
That would be the difference, for example, between having the license for the right to manufacture vs. making a knock-off from the original design. (cf. Amazon making a knockoff of the "nice design" bag)
You can have the license to use Arm's intellectual property to implement Arm's designs which is separate from the license where you can use the Arm SPECIFICATION and then come up with your OWN design.
Suppose that I designed a recipe for a cookie that meets a very specific taste profile that you can ONLY get with MY specific recipe -- I can sell you a license that you can use MY recipe so that you can make your own cookies and they might be in different shapes, etc., but fundamentally, it uses my recipe.
But if you were to then use how I DEVELOPED that recipe to make your OWN recipe for your OWN cookie that has either identical or nearly identical taste profile (such that the customers won't be able to qualitatively tell the difference between my original recipe vs. your new recipe) -- but if you didn't license the right to make your own recipe (as stipulated in our contract when you wanted to license my recipe to get you going/get you started), whilst yes, you did pay for *A* license, but you didn't pay for *THE* license (to make your own).
And that's the difference here.
Nuvia licensed the right to make their own recipe from Arm.
Qualcomm, as far as I can tell, only licensed the right to use Arm's recipe.
And in both contracts/license agreements, I am sure that Arm would have a clause in there about the transferrability of the licenses in the event of an ownership change (i.e. the change-in-control or change-of-control provision which as Patrick noted, IS quite standard for these types of contracts).
You can't have another company pay for the license to make their own recipe and then you buy them up and call that design yours when you didn't come up with that recipe yourself because if the other company was say a much smaller startup and therefore; and let's make up some numbers where let's say if Qualcomm were to license the right to make their own recipe and it costs $100M because Qualcomm is a bigger company with deeper pockets. Whereas with Nuvia, which is more like a startup, doesn't have that kind of money, so Arm agrees to license the right for Nuvia to make their own recipe for say $10M instead.
Well, now, Qualcomm looks at Nuvia and goes "I like what you've got there, so I'm going to buy you up, which includes the design that you came up with for $50M (yes, I'm making up numbers)". Then Qualcomm, instead of paying $100M for the license to make their own recipe, just saved $40M by buying up Nuvia for $50M where Nuvia spent $10M for the license which allows them to make their own recipe.
If Qualcomm were to let this happen, then ANY company that wants a new Arm design would spin off their design division into a "startup company", claim that they don't have the kind of money that Qualcomm has, and argue for a lower licensing cost, develop the recipe, and they get acquired by the original, parent company, arguing that they should've never spun off that division in the first place.
If they EXPLICITLY said that they wanted to do that, I am pretty sure that that's actually illegal (fraud).
But if they filed it under "normal business operations" or "an oopsie", and this wasn't some grand master plan all along, even though it could very well have been, then this move could be legal and Arm would lose out on revenue as a result of allowing this to happen just ONE time.
(It's just like how with some rental agreements, you can't sublet without the explicit written approval/consent/authorisation from the party who owns the rental unit that you're renting from. So if you rent from Some Housing Corp. and you want to sublet your apartment, and your rental agreement says that you can't do that unless Some Housing Corp. grants you the permission for you to do so (why would they? They can lose out on the income that *YOU* would be making), and you decide to sublet anyways, then you just violated the terms of your rental agreement and Some Housing Corp. can kick you and the person to whom you subletted, out from their property/apartment. This is the same jist.)
This is why I think that as a tort/contract law case, it's quite simple/easy/straightforward prima face.
@@ewenchan1239 I didn't even have to read all of that to tell you have gotten something very wrong.
Qualcomm didn't just have the license to manufacture existing ARM CPU designs. They also have the Architecture license that lets them make new designs. Similar to the license Nuvia had to make their own core. It tells you they have this because Qualcomm also tried to design their own core not based on ARM's design for the server market but failed wheras Nuvia apparently succeeded.
You apparently didn't watch the video properly before commenting any of that wall of text.
@@harryhall4001
"I didn't even have to read all of that to tell you have gotten something very wrong."
"If I don't write it, that's my fault. If you don't read it, it's yours."
"They also have the Architecture license that lets them make new designs."
You clearly didn't read Qualcomm filed complaint.
You might want to read it before commenting.
"26. Even though Qualcomm has an Arm ALA, its prior attempts to design custom processors have failed. Qualcomm invested in the development of a custom Arm-based processor for data center servers until 2018, when it cancelled the project and laid off hundreds of employees.
27. Qualcomm's commercial products thus have relied on process designs prepared by Arm's engineers and licensed to Qualcomm under Arm TLAs. Discovery is likely to show that as of early 2021, Qualcomm had no custom processors in its development pipeline for the foreseeable future. To fill this gap, Qualcomm sought to improperly purchase and use Nuvia's custom designs without obtaining Arm's consent." (Source: images.law.com/contrib/content/uploads/documents/394/11978/Arm-v-Qualcomm.Arm-complaint-1.pdf, p.7)
"You apparently didn't watch the video properly before commenting any of that wall of text."
You KNOW that Patrick EXPLICITLY talked about this at timestamp 7'11" (cf. here: th-cam.com/video/_3uJPWBtxkw/w-d-xo.html).
You MIGHT want to review that part of the video and the complaint PDF itself before writing such an idiotic remark which shows your poor grasp of the nature of the subject matter.
Patrick also EXPLICITLY mentions the failure of Qualcomm's Centriq 2400 custom Arm processor that Qualcomm DID design under its own ALA, and that's what Qualcomm's ALA was used for, and NOT for the ACQUISITION of Nuvia's design, nor Nuvia's ALA.
(Remember: the license agreements ARE NOT transferrable under a change-in-control or change-of-control of the company that licensed the license agreement occurs. When a change in/of control occurs (which happens with mergers and acquisitions (M&As), said merger/acquisition will trigger the change-in-control (or change-of-control) clause as stipulated in the license agreement/contract whereby the custom design that Nuvia produced via it's ALA with Arm -- that license agreement cannot be transferred.
Further, Patrick also EXPLICITLY tells you in the video (if you had bothered to watch it at all), that the acquisition itself was subject to approval/authorisation by Arm, also as stipulated in the contracts/agreements. (cf. here: th-cam.com/video/_3uJPWBtxkw/w-d-xo.html)
(Which you will also find on paragraph 37, on pages 10-11 of the complaint PDF.)
You might want to watch the video again more carefully next time before making such idiotic comments. And whilst you're at it, also read the ACTUAL complaint PDF yourself as well (given that it's only a mere 48 pages which you should be able to polish in half an hour or less).
@@ewenchan1239 I read your other comment just to make sure but that's not at all what you argued.
You are trying to claim ARM needed Nuvia's license to make a custom ARM core. Then in the second reply you correctly state they already have a license from when they were designing their own core. You're contradicting literally contradicting yourself.
My understanding is this: ARM have an architecture and IP blocks. They licensed both to Qualcomm, the IP blocks for phones and tablets while the architecture was licensed to make their own server CPU design, these were probably separate deals that happend at different times. Nuvia also got the architecture license to make their own core, albeit at a discounted rate because they were smaller. Nuvia succeeded in doing that while Qualcomm failed their own design. Qualcomm buys Nuvia and it's core design. ARM rightly removes Nuvia's license as they are entitled to do so but Qualcomm already has their own license they tried and failed to use earlier. Sincs Qualcomm already paid for their license at the full price before they even purchased Nuvia so how can they be breaking any rules?
Whats the problem for big company to start they own design from zero? Time?
Happy birthday!
What is arm seeking? More fees? Cease development? Are they planning on bringing this ip to market with another partner? As u suggest in the intro, hard to see a productive partnership after this
Arm is seeking Nuvia's designs destroyed as the main remedy. More fees would have been during pre-suit negotiations.
@@ServeTheHomeVideo oof. This will get ugly. Thanks!
Belated happy birthday!! 🎉👏🎂😁 Hope you had a great day!
Great explanation and summary of this case and I hope you do follow-ups. I probably would not mind hearing about similar cases or news.
Thank you Gjermund. I managed to get a pretzel so I was very happy.
Loved the breakdown! To us mere mortals it does seem like a tiny distinction, but I guess if you're an IP-only company, that tiny distinction means big $$$. 👍
Thank you so much for this video, this was fascinating to watch. I’ve always been curious about the way that ARM licenses it’s IP but would never grok the articles I was happening upon.
You put it in a clear and concise way and in a way that actually helps the layperson in tech figure out which product is ALA or TLA (aside from your mention of RPi and Apple)
what a mess. But I guess ARM can define (limited to thier contract) thier license. Is ARM architecture really that superior? Why not pour RnD into RiscV.
Software wise RISC-V is still not where Arm is. So it is not a trivial gap to close.
Sounds like ARM gave Nuvia a discounted ALA license and they would charge much more for Qualcomm to use the ALA and the license contract was with Nuvia as a startup in the data center market. Not with Qualcomm. Therefore Qualcomm must re-license the ALA with ARM and the terms and cost of the ALA license would be significantly different. Licensing fees and terms of use likely vary from licensee to licensee. This is all ironic when you look at Qualcomm's own predatory licensing model to it's customers.
Qualcomm must have been told that they would have a loophole when they bought Nuvia. They think they can threat ARM financially so they will got better term than play straight.
Now it's depend on ARM. Will they go the furthest, destroy Qualcomm and take colateral damage?
For Qualcomm, it can be only one way:
- you've got greed and a bad lawyer
- or you've got greed and a good loophole lawyer.
My prediction? Wouldn't it be more interesting if we see dimensity in flagships instead of snap?
I hope they work it out because I thought Nuvia was going to provide stiff competition to intel/amd x86_64 in the pc space. Kinda odd that arm would do this since Nuvia stood to make arm look good.
It will likely take some time. Getting to this point likely means a long process of working it out did not reach a successful conclusion.
@@ServeTheHomeVideo yeah but don’t cut off your nose to spite your face. What if Qualcomm and other licensors abandon ship to risc-v. Then less revenue for arm. This whole thing is odd. Qualcomm is probably the largest or one of the largest licensors of arm… snapdragon is everywhere in android devices.
Happy birthday Patrick!
I don’t think this is the first time this (ARM ISA licensee getting bought out) has happened, didn’t Cavium already have a custom ARM core when Marvell took it over?
And these core design startups (Nuvia and Rivos) are probably formed with the goal of getting acquired by a major silicon vendor, ala P. A. Semi. I’d be surprised if Nuvia didn’t try to get favorable license transfer terms from ARM.
p.s. Thanks for the detailed Hot Chips coverage on your site!
Thank you.
We do not know what Cavium's terms were and if Marvell had Arm agree if they had a similar CiC provision.
Happy birthday Patrick!
Just to be clear, do you fully expect this lawsuit to go through? Reading through the complaint, I had formed an opinion that this is ARM last phase to force QC to fold back into negotiation.
Well the claim has now been filed in court. That is what the complaint is.
The court will usually attempt to do mediation first, which will involve a court appointed mediator to assist in further negotiations. If those negotiations go nowhere, then the case is likely to be heard before a judge.
If the comments in paragraph 22 translate to specific clauses about nontransferable licenses in the Nuvia ALA, then Qualcomm's due diligence team was not very diligent (and hence Qualcomm is going to have a very bad time when they get to court).
It sounds like Arm alerted QC before they acquired Nuvia and they proceed regardless. Anything can happen, but it seems QC will either be paying Arm a new license fee to utilize Nuvia's IP or just dead in the water. Attempting to leverage and aggregate their license with the Nuvia one also screams bad faith which is probably why this has gotten to the point it has.
On TLA vs ALA cost structure my research suggests that TLA is much cheaper generally up front to get rights to use the designs, but has high royalty rate once going out. The ALA is other way around and is quite expensive up front, but then royalties are lower or maybe can even be negotiated to not exist in some market segments since the upfront cost and development costs are so high it need amortized into products for them to be viable sometimes...but seemingly all are unique per customer.
Generally I think ARMs real issue here is I don't get the impression they have the right limitations clear on Qualcomms and probably others who might buy out smaller firms in the larger firm ALAs to prevent them from rolling unlicensed by termination 3rd party designs into the still kosher after the deal ALAs and Qualcomm likely has very favorable royalty stuff for their Smartphone and other ALAs they have used in past until they got behind and now ARM wants some higher royalty rates to make up for the TLA stuff Qualcomm will be dropping. So I am sure it creates a bit of animosity generally when you pay however much up front to get different ALA deals with clear cost structures and such and no limitations on whether the designs need to be 100% original or somehow acquired if you go and spend $1.8B also and then are told no, no your ALA and all the money and agreements you made with it are off the table and only this little firms CiC provisions rule all. Those CiCs only invalidate the smaller firms licenses in my mind and so it all then revolves around what Qualcomms licenses say and if there are no provisions preventing qualcomm from buying 3rd party ARM designs and rolling them into their own ALA, that's all tough nuggy for ARM for not writing a better license with qualcomm to prevent them from shopping startup designs to use with their ALA.
Soooo, should qualcomm sue the old Nuvia rights holders for false representation and evaluation maybe get the SEC involved?
If the one thing your company does, no longer functions/is not the property of that business any more after an acquisition, is that company worth anything more than the physical properties/building leases.
This should have been seen during diligence. The action you describe would be only if Nuvia actively hid the ALA. Even if they did, then it would not change Qualcomm's obligations to Arm or Nuvia's obligations to Arm.
it's a really great video, others of this kind will be appreciated for sure!
It seems Arm would make much more of it's money from the TLA licences via "Royalties" ($$ / unit sold), and the bulk of it's money from ALAs via "consulting" on the custom design work of the licencee. If that's the case, I don't see why the "designs" can't be transferrable, even if the licences are not. Now, it all depends on the wording of the licence agreements, and that's likely to be the focus of the case... Or, if it is explicitly worded that the designs are also not transferrable, Qualcomm would argue "fair use", as the bulk of the work was done by Nuvia employees while paying Arm consulting fees.
If my assumptions are correct, and it's a matter of who has the rights to Nuvia's designs, I usually look at who paid for them, and who got paid for them. Nuvia paid it's own employees AND Arm for "consulting" on those designs, so they should have the rights to the designs, and those designs should be transferrable via purchase of the company. Now, those designs are worthless without a licence, but Qualcomm already had TLA & ALA licences of their own, so they should be covered there. If anything, the "tailoring" of licences is the only thing to stand on for Arm, and was obviously the sticking point.
Really, this should be an arbitration case, and not a lawsuit... But lawsuits are more valuable, so there you go!
Happy Birthday (for which ever day that was)!
Thank you Sean!
If I was Qualcomm I'd fire the mergers and acquisitions dept. unless Qualcomm wanted the litigation.
If Qualcomm decided to ditch ARM for RISC V imagine how fast the architecture would mature and evolve. It'll probably never happen but still
If ARM didn't invalidate the License either prior to or at time of purchase, Qualcomm should hold them liable for the price they paid for Nuvia, since ONLY ARM is benefitting from Nuvia's work and expansion into the Datacenter market (as ARM stated, with preferential treatment due to EXPANDING ARM itself).
The ARM licenses are based on the potential market for the resultant core. Nuvia's market was for servers, Qualcomm's is for everything under the sun (much much larger). So if Qualcomm wants to think it can buy a small company to get an ARM license on the cheap, I think they are sadly mistaken. Basically, Qualcomm is trying to screw ARM out of their slice of the pie.... and get the ARM license in effect for much less than it would have cost if it had been sold to Qualcomm directly in the first place.
Makes me curious what Amazon and Apple think about all this... Both companies are big enough to acquire the entire Softbank group if they had too, but Apple in particular might have quite a lot at stake here.
I think if Apple tried, it would raise concerns similar to NVIDIA's attempt. Also - thank you for joining!
I think part of the issue is that Qualcomm is famous for playing real hardball in patent and other IP matters. It has been asserted that Qualcomm negotiated very attractive terms when they first went with Arm and I suspect Arm got tired of feeling abused. I was told that the ALAs are for a specified marker segment and specifically address no presumed transferability between market segments by the same licensee, and certainly not in a change of control. I suspect Q thought they could muscle through the issue like they have been able to do in other disputes. Arm’s negotiating
position is now very different and Arm saw no reason to acquiesce. I suspect there are a number of folks privately cheering for Arm.
I was actually surprised when Softbank bought Arm instead of Apple. However, that “pass” may have been a shrewd decision on Apple’s part to keep Arm in neutral hands. Nvidia’s attempt was worrying to some as it could be too easy for Nvidia to be a proxy for another party.
A wise insight.
I don't think it's certain Qualcomm wasn't working with a pretty broad ALA with Scorpion, Krait and one gen of Kyro. That takes you all the way back to 2007 or even 2006 they would have had to get an ALA or some ALA and TLA mix to do those. But generally wouldn't be surprised if in that early era of ARM catching on as a mobile thing with Apple and everyone else taking phone level computing mainstream and there being extreme competition driving massive ARM growth that didn't give out some generally broad and do what you want ALAs. Now sure as they started to target markets outside of their early growth that was predominantly in more consumer electronics style stuff with more very application specific OSes and such they maybe started to see opportunity in the ALA where they wanted it to be less generic to have better royalties from an ALA where there was margin to do so, but really somehow doubting they were thinking that far ahead when they were still running on mostly consumer electronics style devices and very niche narrow market mobiles and still nowhere near the sorta designs needed for servers or any other higher margin stuff at the time.
To me all these realities mean that Qualcomm probably has a VERY favorable ALA to do with as they please so long as they pay whatever menial royalties they had promised. I'm sure they have to do some new stuff with like ARMv8 and v9 and such, but even at v8 back in 2011 I doubt they had to start making special per segment stuff like a lot of the more recent efforts to get out of mobile and CE space to higher margin spaces have caused. But, to me that means ARM has very little leverage on ancient ALA holders like Apple/Qualcomm until someone wants an ARMv9 license. It's even possible ARM was seeing the better margin potential by the time someone like Apple purchased PA Semi and wanted to use the team for ARM and got more lucrative ALA royalty agreements than they have with a bit older ALA holder like Qualcomm. And I am sure the TLA heavy royalty nature of qualcomm since they gave up on Kyro custom efforts has been a boon for cashflow. Doesn't mean they can just try to demand more from the standing ALA if Qualcomm uses a 3rd party to drop designs in an ALA without any terms that disallow that.
Very disappointing to see this, I was neutral on Arm vs RISC-V, but now I'd definitely prefer if in the long term everyone eventually switch to RISC-V
You could hazard a guess that Qualcomm is trying to buy itself out of the jam of not being able to create a true competitor to Apple by buying Nuvia. A short cut that is understandable given that for the past few years, it has been having problems putting out quality chips. Heck, Mediatek is coming out with better designs now.
While Qualcomm seems to be all over the place. Mid range chips that are actually slower than the last year model. Overheating flagships.
Do you want RISC-V server startups? Because this is how you get RISC-V server startups.
Happy birthday Patrick!
I think you are right. And thank you Max!
I think it's called SiFive
I'm inclined to agree with ARM since they are not using this as a method of extortion due to Nuvia being sold to another party, but because of the other party looking to take advantage of the merger.
"because the other side wants to take advantage of the merge" what? Why, then, does the "merge" exist as a commercial procedure between companies? For the sake of a joke?
@@arround1 Hans, the Nuvia license only covers designs for Nuvia products sold for datacenter use and Qualcomm is trying to use it to produce their own cell phone products and sidestep their agreement with Arm. It's quite a clear case. Qualcomm violated the terms of the Nuvia contract with Arm.
Loved the video, super interesting for me :D
Excellent explanation of IP legal issues in company acquisitions!
Thanks David
Qualcomm does not own all of the ALA IPs to be able to make use of the Nuvia IPs, so should they opt to not to purchase the correct licensing the Nuvia IPs are likely tainted as how can one separate out any of the work derived using the IP from ARM, even if it all there was were simple interfaces between ARM's IP to Nuvia's IPs, who's to say that engineers from Nuvia looking at ARMs IP was not inspired by ARM's work and elements unintentionally made it into Nuvia's IP.
I dont see how Qualcomm can get out of this without paying for those licenses.
Also, Happy Birthday!
Thank you Carlos!
When are we going to get a new updated HP Microserver Gen10+ by the way?
Great question.
Happy Birthday! 🎉
Thank you!
It makes sense. Qualcomm would be skipping out on loads of licensing fees, especially when the owners of ARM are upside down in their investment.
Since the acquisition is already completed, I don't see how it gets undone.
Qualcomm tried to do an end-run around ARM's IP licensing model by purchasing Nuvia and then announcing that it intended to go beyond Nuvia's ALA agreement on datacenter servers to other areas WITHOUT ARM's agreement. What ARM has now done is basically make Qualcomm's purchase of Nuvia worth nothing by cancelling the licensing agreements AND cancelled Qualcomm's TLAs as well.
That IS a big deal. Qualcomm called ARM's bluff that it wouldn't sue its largest customer and it could get messy from here. Qualcomm's dependence on ARM's technology IP means if that it cannot release any more new processors on ARM's tech IP, which could allow competitors to do so undermining Qualcomm's market share in key areas.
How it could get undone is simple: Qualcomm agrees to pay a hefty fee in return for a) Nuvia being allowed to continue developing for datacenter hosts contingent on b) Qualcomm disallowing itself from using those processors in any other device or market without ARM's consent and allowing c) Qualcomm to get its TLA license back.
Anything else will simply fill attorney's pockets with money while Qualcomm loses market share waiting for a judicial ruling and appeals process to take place.
Full steam ahead with RISK-V then!!
Could be the season of risc-v especially for baseband chips from Qualcomm
I can't imagine how would a huge company like Qualcomm made an acquisition without sorting out the IP and licenses.
I saw arguments that Qualcomm might already had an ALA license from Arm. First of all, Arm is now claiming Qualcomm does not have one allowing them to do what Nuvia is doing. Secondly, in my opinion, if Qualcomm does have one, they would have used their resources to design their own CPU cores, instead having market share eaten by MediaTek and the poor performance compared the Apple's Arm CPUs.
As far as I understand, the problem is that ALA licences aren't general licenses but specific to particular product development. Qualcom's ALA license was specific to the development of Centriq 2400, so it can't not be used for Nuvia's IP. Qualcom should have negotiated a new ALA specifically for the development of new cores based on Nuvia's IP.
The availability of a non-x86 chip that can compete with M1 was very exciting. ARM is being so dumb in my humble opinion.
Wow this is crazy! Time for open source CPUs
Open source CPUs don’t protect you from patent lawsuits, Qualcomm for example is always suing other companies for unlicensed patent use.
this is getting very risc-y for arm. i wouldnt be amazed if apple is behind this wrench being thrown into themerger. apple is known for doing shady stuff like this.
So apple is risking here for doing ARM cores design for desktop and laptop?
Or their ALA and TLA were laid for that use in the first place but only used for mobile devices in the begining?
I'm not sure if Apple being the company it is would be eager to pay for ARM's blessing for desktop and laptop since the beginning or their relationship with ARM when they only made mobile SOCs.
I think this move from ARM will make some companies to review their TLAs and ALAs to not being caught between a rock and a hard place.
Happy (belated) Birthday!
Thank you!
To say don't sue your customers does not and should not apply to a company like Arm. I believe Qualcomm needs Arm, more that Arm needs Qualcomm.
This is the core business of Arm and no one is allowed to 'Bend It Like Beckham'.
14:00 Why you selected exactly these colors? disgusting.
ARM is going to find themselves of being accused of “Working for Apple and Tencent.” This could be a publicity nightmare for ARM. It creates incentives to create non-Arm processors. ARM is killing success.
ARM wants a piece of that 1.5 billion and they don't produce their own chips, so of course they were not going to allow transferring (a non transferable) license without a cut. It also seems like they would have to defend their licensing otherwise it could be used against them in other cases..? It'll be interesting to see QCs legal reply. I also wonder if Arm gave Nuvia a sweetheart deal as a startup.
Sounds like good news for SiFive. Bad news for everybody else.
Well Patrick I can observe here on what I'm thinking including on the legal and industry political wrangling. mb
I think ARM is right here and happy B-day m8
Thank you! Much appreciated
IMHO ARM is cutting a branch it's sitting on. I suspect a fair amount of startups in this space will consider going with license-free ISA like Risc-V instead of ARM. Since most startups are created with a specific aim of selling them to "big boys", this licensing nonsense can make those startups un-sellable.
Overall, none of the rhetoric used by ARM in the complaint matters. The only thing that matters is what was actually in the licensing contract that Nuvia had. Until we see that, no informed opinions are really possible.
"you are not allowed to use any of the technology created by Nuvia" What? How can you defend that? ARM can't require people forget what they've learned. ARM can't compel a company to delete their own data. Nuvia could created a patch file that contains only their changes to the ARM source materials, call it 100% original work. Then apply the patch to Qualcomm's source materials. ARM is in the wrong here. They keep hanging on the "licenses do not transfer" aspect of the contract. Qualcomm doesn't need to transfer the license because they have their own. Qualcomm purchased knowledge and employee contracts. Qualcomm used their knowledge and their workers to create a chip design with their license. I've heard nothing to suggest any different.
Imagine if Qualcomm had to halt all ARM processor production.....
So many companies (even huge ones) just use others 'stuff', and have no idea how to do much of anything.
Happy Birthday!
Thank you Jan!