In fact a lawsuit like this is very rare because such negotiations usually happens behind the stage. What Arm claims that Qualcomm does wrong is that they should not use another Arm licensee's IP without the confirmation from Arm, in other words, the license obtained from Arm cannot be transferred to another entity without Arm's consent. This makes a lot of sense, because it avoids the case that some entity does not directly obtained a license from Arm at all to use Arm's IP by getting the license/IP from a 3rd-party.
I suspect it is not the licenses themselves that deeply bother ARM as much as how much ARM charged for the Architectural license. If they were trying to cut a start up a large break they might license something at 50-60% of the cost to a large player who was far more stable (had lots of cash in the bank) and could pay full price. A too expensive Architectural license could 'kill' a start up that was pragmatically strapped for cash. If there is a 50/50 chance that the company is going to implode before they ever shipped a product in substantive numbers, then stressing them even further just raises the odds that they'll fail faster ( 70/30 that they'll fail).
If the company survives the first round if architecture license buy, then Arm will raise the prices when it won't distress the company quite as much on the next round. ( If Nuvia had only sold 30,000 of their first gen SoCs then can't get blood from a turnip. It is only a big payday if they survive to generation 2 or 3. )
What ARM likely does not want is super-duper deep pocketed companies buying up extremely discounted architecture licenses are bankrupt company fire sales. There is a pretty good chance that while Qualcomm held an older server architectural license they didn't have a new one. And that ARM was in no way going to sell Qualcomm the deep discount license to help them with corporate startup overhead and growing pains. They had a $1B of 'extra cash' to buy Nuvia, so far from broke.
Architectural licensees are a dual edge sword. It is lots of up front money and less long term money. If Arm doesn't carefully balance how ( to whom and for how much) they hand this out that will kill off the company . Or at least, any high range valuation of the company.
There is less of a danger for the Tech license agreements. If the buyer gets that and sells lots more, then Arm gets paid substantially more as the units are shipped. Getting approval for that is probably pretty easy to work out, but again any adjustment Arm made to not kill a start-up short on cash shouldn't be in play if the new "owner" has deep pockets.
But one important thing is that, Qualcomm also has an ALA with Arm directly just like Nuvia does, so Qualcomm could claim that their ALA covers the usage of Nuvia IP. We don't know what the ALA exactly says, so we don't know if that is the case or not.
That isn't important if they are not the same Architectural license. There seems to be some mythos that once you have a ALA that you can implement anything that Arm every produces for all time. That is highly likely not true. When Neoverse N2/V2 come along then need an update to the ALA. When ARMv9 comes along then need an update if only holding a v7 license.
Otherwise how would ARM fund long term R&D on the instruction set if give away the whole farm forever on everything new.
In their previous custom server efforts, Qualcomm never worked on a Neoverse era server architecture. They had not custom efforts going on in 2018. So it is a bit dubious that they ran out and paid ARM a very large check for Neoverse if they were not going to work on Neoverse at all. If Nuvia and Qualcoom had effectively "double paid at full price" for Neoverse I doubt ARM would be suing them right now. Qualcomm might have paid "I'm not going to do anything with this so here is some tip money to keep the server thing going that I might use in the future" and Nuvia probably played the "we're a broke startup who isn't going to sell 100's of thousands of chips so cut us a break" card.
Nuvia was also in such a rush to market I doubt that what they were doing was fully "from scratch". Pretty good chance they were going to take a large chunk of the uncore part of Neoverse and just reuse it. Which probably made them more attractive to get bought out also if could just for instance just couple their cores to the Arm standard core interconnect subsystem. Someone could buy them and plug just "fresh cores" into the existing infrastructure if it was based on slight/narrow mods to the Arm interconnect. The eventual Nuvia SoC was not going to consist solely of just Nuvia cores. It was a relatively small start up that was only going to be able to cover just so much ground. Lots of subsystems that they considered non-high-value-add were likely going to get bought off-the-shelf because it was both way more affordable to do and would take far less time and resources. (Apple did the same thing. They didn't try to do the
whole SoC all at once. And they had deep pockets. Nuvia didn't. )
I think they will likely to reach an agreement and ends by Qualcomm paying an IP transfer fee to Arm.
Qualcomm had every opportunity to do that all during 2021 and did not.
March 2021 completed acquisition.
https://www.qualcomm.com/news/releases/2021/03/qualcomm-completes-acquisition-nuvia
End of 2021 still wrangling with ARM over the issue. At over six months of negotiating ARM was done. ( the Nvidia acquisition of ARM was probably a limited distraction there. But by the end of 2021 it was widely obvious that deal was a giant turd in punchbowl that wasn't going anywhere. )
I suspect Qualcomm is just going to have to buy another Architectural license that is targeted to covering a marking of 5-15 Million units per year rather than some two or three order of magnitude smaller number. Qualcomm probably does not want to write quite as large a check until they know better how well it will work. They paid over a $1B for a company with no working product; not even a prototype of a working product. That is kind of nutty. It would be even more nutty to write another large number of digits check and find out that it still doesn't work. Qualcomm has money, but they can't just throw money away. Apple isn't going to keep stumbling on their cellular modem forever. And Samsung might figure out how to make their 3nm fab process work.
Sinking $20M in lawyers fees might be worth it to provide air cover during the delay. "we can't ship anything because ARM is getting in our way . blah blah blah" is a working 'dog ate my homework' story when this doesn't ship anywhere near the timelines Qualcomm was talking about at the end of 2021.
if it only works way better as a server SoC then Qualcomm my shift the whole thing that direction. Then claim ARM said the Nuvia license was just for servers and that's what we are doing. And pay a much smaller fee ( ARMs legal fees for the dust up.)
P.S. Qualcomm is also likely to stoke a fire under Apple, Microsoft, Nvidia , Intel , etc. about how ARM can blow up any acquisition deal if they want to when ARM IP is in the loop. It might be worth it to Qualcomm to pay to test the limits of Arms contract. ( similar to way folks are testing Apple's app store rules in courts. Often losing but getting some concessions pragmatically out of Apple indirectly. ). That also will put tons of "pepper spray" on anyone else trying to do what Nvidia was trying to do and buy ARM. Even if ARM won It will make the more "unpurchasable" because is it such an anticompetitive threat for a ALA holder to own the right of refusal over everyone else they compete with.
