Intel Corporation (INTC)

Estimates put peak densities (million transistors per mm2) at:

TSMC 16nm 28.2
TSMC 10nm 52.5
TSMC N7 is 91
TSMC N5 is ~171
TSMC N3 is ~290

Intel 14nm 37.5
Intel 10nm 101
Intel 7nm ~200-250

Note that these are peaks. Critical logic isn't made at peak densities.

From Ian Cutress of Anandtech.
Intel 10nm is denser than TSMC's so called 7nm. One of the biggest marketing gimmicks in semiconductor industry.
And that's why it is fair to assume Intel's house is in order once they are manufacturing 10nm.
The nanometers don't mean much, its a marketing gimmick.

Search for relevant tweet if interested in more details. Even if 7nm capacity at TSMC is freed up, Intel has not much to fear.

Grace server CPU will create only minuscule opportunity (if at all, in 2023) for nVIDIA where GPUs may be used with processors, x86 vendors have nothing to worry about it. In fact, it is other way round. With Intel's Ponte Vecchio to be released soon, nVIDIA may lose some GPU market to INTC. Intel Habana AI training chips are already being used in Amazon AWS and now in San Diego Supercomputer. And there are analyst on Wall St. who see this:

As per Barron's:

New Street Research analyst Pierre Ferragu took a different view. “We call [its] bluff,” he wrote in a Tuesday note. To Ferragu, the data-center chip can only capture a small piece of business tied to Nvidia GPUs (graphics processing units) in big data-center systems. The chip won’t expand the addressable market, and the potential opportunity is “an order of magnitude smaller” than the GPU business.

Let this dust settle, INTC will be rising again for many reasons, Intel GPU being one of them.

It's clear $intc attempts to quietly change its business to become a pure fab player like TSMC, diversifying its chip designs x86 that have no chance to compete with AMD, and future ARM made by AMD too and others. But it will be a long and risky journey with $10s of billions spending not guaranteed to make a profit.
New Intel's CEO continues previous CEO's move to stop depending on x86 chips only at a more extreme level. X86 could be sold in a few years, but not known if the contract fab business can replace the revenues lost.
Why risk such adventures? Sell Intel's shares now, buy AMD is the prudent choice to fo as AMD has successfully transformed its business to become the chips designer for the world with the Xilinx's merger.
Don't be fooled!

New Microsoft announced Surfaces laptops BASHING $intc !
"Microsoft wasn’t very clear on how consumers will know to pick between the AMD or Intel models. But it said the AMD models will offer better gaming performance and will be more appealing for people working with photo and video software given the boosted graphics. Part of that decision will be made for you depending on how much you want to spend. The entry-level model only comes with an AMD chip, for example."
Ha... Microsoft has just advised to select AMD as Intel is second best! WOW
https://www.cnbc.com/2021/04/13/microsoft-surface-laptop-4-price-release-date-intel-amd.html

A joke. .. so nVidia's has announced a PLAN to make servers ARM CPUs competing with $intc ... oh sure Intel's no ARM development. Competition is with AMD which has a lot of advanced IP in ARM plus the Samsung rDNA Exynos partnership on 5nm fab should be out soon. No, this is the typical BS that trying to get clueless retail investors to buy nVidia's shares. AMD has a long history of ARM based servers, developed their own the K12 and hold an ARM architecture license allowing them to develop their own ARM compatible CPUs just like AMD can do that with Intel's x86. I'm expecting AMD to launch ARM datacenters CPUs THAT ARE SOCKET COMPATIBLE WITH THEIR EPYC and use SAME IO CHIPLET as EPYC connected with an ARM 5nm chiplet from Samsung through infinity fabric! Do you understand how simpler it's to have the same server boards that could fit both Zen and ARM servers? Everything else is the same? With AND's engineering team plus Xilinx's there is no comparison with nVidia's just starting to develop an ARM processor product. Look at past nVidia's ARM APUs - horrible and stuck within Nintendo Switch. Plus some used for automotive applications. No nVidia's no talent to make CPUs especially compared with AMD. This should jump AMD's shares way higher, of course Intel's toast.
https://www.reuters.com/article/nvidia-arm/nvidia-announces-arm-based-grace-server-chip-in-direct-challenge-to-intel-idUSL1N2M51J6

Speaking looely, nVIDIA's new server processor's major is to ensure high bandwidth communication between CPU and GPU for training and specialized AI applications using NVME link. Needless to say, the incumbent x86 processors in servers won't support this over pcie for a competitor.

Few points to be noted. Without this feature, the writing on the wall for nvidia is loss of GPGPUs for AI in HPC/servers as both Intel/AMD have their own CPUs and server (Intel about to launch) GPUs. This has a real risk of putting nvidia in disadvantage. Ponte-Veccio, does 41 TOPS in FP64 TODAY in Silicon as opposed to nVIDIA's best Ampere's 11 TOPS of FP64. Indeed, as Pat Gelsinger pointed, it's a reaction by nVIDIA to stop INTC/AMD to eat into its server GPU market, NOT the other way round. The market, as usual, is irrational and prone to pump.

The first purported silicon will be available in 2023 and remember it will be competitive (BIG IF) only for certain AI workloads (Not all). A realistic competitor is due by 2025.

What's the chance the first Silicon will succeed? ZERO!!
Server Silicon is not a joke. Amazon worked on graviton for 5 years and only a few months back the processor has some semblance of competitiveness in BENCHMARKS, selling and making money from them is YEARS YEARS away. AMZN can put them into their own AWS, nvidia has to show profits by selling them ASAP.

How difficult it is to get a competitive silicon for CPU? Ask Samsung.
They had their Exynos SoC with custom ARM cores being designed for years into their OWN mobiles and other hardware. Yet, with all the multi-year effort in Austin Research Center, they could not beat Snapdragon for their flagship phones. And Samsung, let me tell you, is one heck of competitive company. None other than ex-AMD head of Opteron lead that development for a decade!

Plus, Intel is coming up with Ponte-Vecchio, and many more EMIB/Foveros packaged devices and they have till 2025 to deliver.

In summary, nvidia's so called processor is first of all targeting only a subset of workloads in servers, their chances of competitive silicon are low and selling them is not a given. Competitors are not going to sit idle.

If at all, I believe this is nVIDIA's shot at protecting its server GPU turf from INTC/AMD, not the other way round. The market reaction of clueless jokers on Wall St is of course foolish. Where is the money on table from the CPUs anytime soon?

Nvidia is up 3.09%, AMD is up 2.05%, Intel is down 0.28%

Is is clear to see who is the king of tech, the runner-up and the old dying dinosaur.

$NVDA $AMD $INTC