Given you have said how CUDA proliferation is a problem for ROCm are you really unable to see the vast amounts of software written for x86 that has not been compiled for other architectures, really?
x86 gives them a different slice of the market and a very big slice, one they cannot compete in right now.
AMD / NV would cross licence, no way at all Nv would be allowed to buy Intel without agreeing to that.
Also, I was not being entirely serious, although your arguments against my not very serious idea have not exactly been that strong.
They are AND they are doing a high performance handheld chip too.
Once upon a time Jensen was open to Nvidia merging with AMD, so long as he got to stay on as CEO.
AMD didn't like that one bit, bought ATi instead and the rest is history.
I can see the Intel thing happening... So long as Pat GTFO.
I don't even know where to begin, so I'm just going to take everything in one broad stroke.
Nvidia's entire construction since the early GPU days has been to build exclusivity. This started to be particularly visible with how much they poured into software over hardware. You could just say it was "the right move", but really, it was also a way to expand on the control over their hardware, handling everything with limited hardware access that allowed them to vertically integrate and control the software stack away from pesky clients that could have done things Nvidia didn't want.
This was seen more strongly a few years later when CUDA started: not an open standard, no option to integrate your own stuff to Nvidia's stuff. NV offered the door to their hardware and made it very wide (CUDA works on all devices) and very approachable (CUDA is actually well built despite its age). You get Nvidia quality, but you don't get freedom.
This corporate culture allowed NV to strongly grow, because the deal with Jensen is a simple deal: they control everything, but they take responsibility to make it reliable/good enough that clients don't need to worry about insufficient support. This is an extreme contrast to AMD, who give a ton of development freedom and try every open standard they can find, but pour ridiculously low amounts of software and support in comparison with Nvidia. Nvidia is security under Jensen's thumb, AMD is freedom in an anarchic, underdeveloped ecosystem.
Past the early days of CUDA where it was mostly about adoption, between say 2012-2018 (after which AI, ofc), Nvidia did nothing but assert more control over their own software stack and clients. A little Nvidia golden cage that grew year after year. You can see the effects of that in things like NVENC also: NV did everything to control the entire system internally. It wouldn't have been hard to make a standard API for encoding/decoding that worked for Intel/AMD/NV, but NV is not interested in playing nice.
Along this, there's an important component about Nvidia's internal structure: they have very little of it. For a company of that size, the amount of corporate hierarchy and on-hands management is VERY low. This is because Jensen's philosophy on hiring is that if they're good enough to be hired at Nvidia, they're good enough to not need a manager chaperoning everything they do. If they're not good enough to be autonomous and bringing returns to the company, they don't belong. Period.
This sounds like some kind of management heaven, but it only works on the premise that you don't have "small hands". You don't have people doing the lame jobs and the dumb ones, you only have smart people doing smart things for smart roles. Problem is, almost every bigger company out there is going to have "small hands" and small jobs. This kind of "everyone is good, everyone is talented, everyone is smart, everyone gets well paid, everyone is happy" little world can only exist if you have high income, high added value for every job. Middling returns are simply not acceptable; you want smarts? You want talent? You pay for it, and that kind of money has to come from highly optimised returns.
To spell it out, Nvidia as a little powerhouse of talents can only exist BECAUSE they designed the entire company to make high net income products and services. They COULD have chosen to do more open source or relieve themselves of some difficult things; instead they took all the responsibilities upon themselves and took away the freedom of their clients. More hassle for them, but also more control and more money. Jensen didn't just "hire good people because good people are good". If it were this easy, HR in every company would've been thrown in the dumpster to choke and die on moldy orange peels, as God intended.
And of course, the ultimate expression of Nvidia the way Jensen built it is to end up being paid obscene amounts of money for access to overpriced hardware that is paid for because it's tied to a proprietary stack where they hold more control than AMD or Intel ever did. He designed Nvidia like a spear: all about piercing into markets with high margins and making away with as much percentage as possible.
In contrast to all this Jensenian business philosophy, Intel and AMD are both built the exact opposite. They're generalists with entries in a ton of businesses. Bought so many smaller companies. Focused on everything that entered a computer, Intel particularly (remember Optane? Intel wifi chips? Intel switches existed too I believe? the AI companies they bought? ARC? and so on). They don't care about splurging into a field if it feels like it will expand their market. This is a radical difference in every part of the business.
Where Intel and AMD will attempt to sell as much volume as possible, or historically have at the very least(AMD's margin obsession is fairly recent, they used to be far more balanced), Nvidia did everything to focus on the core business that brought in money. Intel tries Optane? AMD tried mobile at some point? Nvidia was about GPUs, where the money was. Intel gains a zillion monies with CPU, so they try to expand to everything? Nvidia gains a zillion monies with GPU, and they're very happy, please don't bother them.
This also goes in line with the fact that AFAIK, Nvidia never even bothered to have a Fab. Sure, lots of companies didn't either, but Fabs are a logical way to integrate vertically, aren't they? And yet they never bothered. Because their entire philosophy is about getting the golden sheep, and to ride it. They don't shear it to open new divisions or buy a ton of possible business expansions and side gigs. They just ride the one income driver forever. Opening a Fab would technically give them more control, but the margins/returns wouldn't be as high.
I said Nvidia was designed like a spear, maximise damage at the point of margin. Intel and AMD are designed like axes, cut large parts of the market in broad strokes that do not have the same effort/penetration ratio.
IF you gave Jensen Intel. That is GIVING him the company, one symbolic dollar for it, and he was the sole owner of the entire shares but HAD to run the company himself, he would destroy Intel within a few years.
He would try to find the moneymaker, and dump everything else. He would try to get rid of IFS. He would go around amputating the company of all the parts he deems gangrenous, which is all the parts that do not reach his lofty margin expectations. Which...is all the parts. He would kill Intel.
Jensen didn't build Nvidia as "a successful company", that is an incredibly Reddit tier take on them. He built it like his own little kingdom where everyone would be clever and make each other money together. He relented on unnecessary corporate control and just expected everyone to be smart. It's built sort of like a cool frat boy company, every cool guy goes to Nvidia and that makes them the coolest, and richest, company around. But cool guys don't touch uncool things.
Jensen doesn't want low margins. He doesn't want volume. He HAS the volume, sure, but in his mind it's because he built the cool thing. Always remember that Thermi was a pivotal moment in tech history where AMD had the better and cheaper product while Nvidia had the fatter, more expensive, and all around worse product. But he didn't care because it would allow him to create the Compute market and put his thumb over it for the next 20 years. Jensen went for the expensive, risky, and poorly designed thing that he could sell for high margins, over the standard, tried and tested, uncool "normal graphics card".
So for all the people that think that "Jensen will buy Intel because it makes sense", Intel is the last thing that Jensen wants to buy. It is the antithesis to what he's built his whole life. And the antithesis to how he views business. Cool frat boy company that controls the entire stack on the inside, dominant, nothing shared conqueror company on the outside.
And also, that bit about "he said he'd allow AMD to buy Nvidia, if he were the CEO of the new company"?
That's not a serious offer and he knew it. He was basically saying "you can pay to have me and my company but I'll be the boss of you because you're all terrible". Don't forget that Jensen did his classes at AMD younger. The man has a serious ego and loves to tell stories, I'm sure the story of him being a little guy at AMD is one he wanted to erase from history.
Now I'm not saying that Jensen will never not have Intel. I could see a situation where Intel, or more realistically the american govt, get desperate enough that they basically offer themselves up to Jensen to get saved. And nonetheless, I don't take a single word out of what I said: that would be Intel's death. He would gut the company down to nothing like what it was. Jensen's not a genius any more than Lisa or many other very smart people of the industry. He just found the way to make the most money out of the least diversity.