I’m skeptical that will work, but Son was dumb enough to pay $31B with no strategic value.
At the time I thought Son had clever telco synergies in mind, but I gave him far too much credit
The problem is that Son needs cash, so he's flogging off everything he can to get it.
https://www.businessinsider.com/running-list-softbank-invest...
Grossly so. They paid like 45 percent above what the stock was trading at the time.
They'd be paying a premium for a path to an all-nvidia datacenter & supercomputer.
Consider HPC applications like Oak Ridge's Frontier supercomputer. They went with an all AMD approach in part due to AMD's CPUs & GPUs being able to talk directly over the high-speed Infinity Fabric bus. Nvidia's HPC GPUs can't really compete with that, since neither Intel nor AMD are exactly in a hurry to help integrate Nvidia GPUs into their CPUs.
This makes ARM potentially uniquely valuable to Nvidia - they can then do custom server CPUs to get that tight CPU & GPU integration for HPC applications.
There is [0] https://en.wikipedia.org/wiki/NVLink
which is supported by [1] https://en.wikipedia.org/wiki/POWER9
those two combined give you [2] https://en.wikipedia.org/wiki/Summit_(supercomputer)
currently the worlds number 2 supercomputer(only very recently dethroned) according to the article.
Installed at Oak Ridge.
So they are already there, just needing some premium POWER?
Amazon paid 350MM for Annapurna, ~ 1/100th of 32B.
For embedded devices, Nvidia already ship Jetson boards with Arm CPUs and Nvidia GPUs.
https://hothardware.com/news/nvidias-64bit-tegra-k1-the-ghos...
> .. this chip is fascinating. NVIDIA has taken the parts of Transmeta's initial approach that made sense and adopted them for the modern market and the ARM ecosystem -- while pairing them with the excellent GPU performance of Tegra K1's Kepler-based solution.
https://www.extremetech.com/computing/174023-tegra-k1-64-bit...
> there’s an interesting theory ... that Denver is actually a reincarnation of Nvidia’s plans to build an x86 CPU, which was ongoing in the mid-2000s but never made it to market. To get around x86 licensing issues, Nvidia’s chip would essentially use a software abstraction layer to catch incoming x86 machine code (from the operating system and your apps) and convert/morph it into instructions that can be understood by the underlying hardware.
Which other Arm licensee has been talking about x86/Arm instruction morphing in 2020?
If the goal of acqui-billion-hiring the Arm reference design team is to prevent other companies from using those designs, that would endanger smaller vendors in the Arm supply chain, along with many of the devices that run modern society. Regulators may not like that.
Is Arm not profitable as a standalone business? They recently raised some license fees by 4X.