Any one can explain why would Nvidia would buy ARM for 20 years of revenue's worth?($2B/year)?
In part, because over half of that isn't their money. It is pragmatically the stock holders money, so it is "monopoly playing money" to them. $21.5B ( or $23B if count the "make it rain" on some of the ARM employees ). Somewhat convenient if can just 'print in the basement" and use it as 'money'.
There is probably a huge implicit presumption that eventually new stock holders over time will do a partial "bail out" of this by driving the stock price up. [ Some presumption probably also that Softbank won't dump most of this in the intermediate term also. ]
Some hand waving finance folks will point to it is "just" $17B cash so perhaps only 8.5 years worth of revenue. Don't "have to pay back" the stock part (magic voodoo of the market will do that), just "have to" fill the cash reserves back up (and pay off any bonds used for this at super low interest rates. ). It is much less of a 'hurdle" is only have to pay back the cash used. ( but yeah if look at the profits ... even $17B is a head scratcher if look close. ARM's business model (high R&D overhead) is not a "print money at high margins" business model. )
Second part, is that Nvidia with use ARM resources to make more profitable Nvidia products. And that perhaps Nvidia products' higher margins partially pay for this. ( because perhaps taking the ARM instruction set R&D money and info collected to be skewed toward making better Nvidia products rather than helping ARM customers. ) A business that has much higher profit margins has a shorter return on the investment than ARM's relatively very thin margins can manage to eek out.
This may turn into a dual edge sword if too skewed because ARM customers would tend to start to exit. If Apple buys no more architectural upgrades that profit margin is zero. If most of the other Arch liensce holders also go to no new features ... again pragmatically probably have negative margin.
There is a big of a bet here by Nvidia that ARM is "too big" to fail. Big in as too broadly licensed, not that it generates lots of revenue. (or at least fail over the next 4-10 years. ). If ARM is going to collapse in 4-10 years then if Nvidia is controller the direction and scope of the collapse they can front run that process better than the others who don't have their hand on the controls.
Depends on how Nvidia's products go then perhaps they can operate ARM as a breakeven division. That it won't "pay back" the purchase price at all. The whole burden would be shifted onto the much higher margin datacenter products. That is probably the only way the "open license" model survives.
The problem with that is Nvidia already made a large Mellanox purchase that also needs to show an return on investment. That is why most of this on inflated priced stock. If the stock doesn't fall back to reality then they can get away with it.