bleab

robaba said:

Hope this doesn’t go through.  We need less consolidation within chip manufacturing, not more.

1. TSMC and Samsung have 74% of the chipmaking market. GlobalFoundries has 7%.
2. Chip manufacturing requires capital investments and expertise that few small companies are capable of. Only 3 foundries on the planet - TSMC, Samsung, Intel - are capable of even 14nm at high volumes. The fourth will be China's SMIC in early 2022. 
3. America has 3 real competitors in chipmaking: Taiwan, South Korea, China. The other 3 directly subsidize their chipmakers to a massive degree. Meanwhile America only voted to give its chipmakers their first subsidy in decades a few months ago ... and it was comparatively a very small one. So if you want "less consolidation in chipmaking" then you are asking for the federal government to get - for example - Texas Instruments from 130nm to 14nm and get GlobalFoundries' 14nm capacity from tiny to competitive, it would take more money than Intel's $78 billion in revenue last year.
4. The last one - Intels $78 billion in revenue  - is why no antitrust types are going to target them. Where Intel's $78 billion 2020 was a record year for them, Apple at times makes more than that in a single quarter.

Bottom line: you can either have a bunch of tiny chip manufacturers that can't compete with TSMC or maybe 4-6 that can. But you can't have both.

larryjw said:

A difference between IOS and Android are the customers. iOS customers are likely to be higher spending, higher income I would expect advertisers would get more bang from the buck on iOS. Will increasing advertising on Android give the advertisers their money back?

The answer to this can be illustrated with the decisions that TV networks made in the 1990s. Even though more people live in middle America than on the coasts, programming aimed at the coasts was more effective for advertisers because it allowed them to target their ads. The same was true of demographic/niche programming i.e. channels like BET, MTV and Country Music Television: ads for Ford trucks are far more effective on CMT than they are on NBC's Good Girls. 

So while there are more high income people with iPhones, thanks to ATT they are now equivalent to network TV viewing audiences' middle America. A lot of people but VERY different from each other. (Just as Tennessee very different from Utah. And even east Tennessee is very different from west Tennessee.) By contrast, app tracking allows advertisers take the larger set of Android users, section it into the ones who have - for example - Samsung Galaxy phones that cost from $700 like the Galaxy S to $2000 like the Galaxy Fold. Those are going to be your Los Angeles and NYC residents that advertisers are going to be able to target. 

Untargeted advertising is very wasteful. You may make $1 million but you have to spend $600,000 to get it. Meanwhile with targeted advertising you may only get $500,000 but spend $150,000 chasing it. Most advertisers would choose the latter, especially since you can take the $600,000 that you would spend in a single nationwide untargeted campaign and instead run $150,000 campaigns targeting San Francisco, Los Angeles, NYC and Boston. 

The biggest benefit to iOS for advertisers was that it was already a targeted market - San Francisco, Los Angeles, NYC and Boston - to begin with. But you don't want to try to sell beachfront property in NYC. Or snow tires in LA. Or even snow tires in NYC to single professionals who rely on the subway. That takes microtargeting and ATT precludes that on iOS. The overall Android market isn't targeted - it contains a ton of sub-$150 devices for the prepaid market for example ... that is the bulk of once proud Motorola's business right now - but it is still possible to use tracking to identify and microtarget the small percentage of people in NYC who drive family sedans, minivans and SUVs instead of catch the subway and push them snow tire ads.

darkvader said:

And then more ads push more people to iOS. 

1. Not nearly as many people care about this as you think. 
2. Pretty much everyone who does care about this uses iOS already.

tht said:

Nikkei Asia has a good track record for reporting on movements within Apple's supply chain. It very rarely makes predictions on Apple's future plans, but when it does, it generally gets them right.

Weird rumors here.

The iPhone 14 using a 4nm chip imply the 3nm chips are really for 1H 2023 devices. Ie, 3nm mass market production is expected to start in the Fall of 2022 while iPhone 14 needs it to start in the Spring 2022. So, Apple has to use a 4nm process. Nothing too weird here.

But, since when did Intel jump the line to get 3nm chips from TSMC? That essentially means they wrote a $5b check to TSMC about 2 to 3 months ago to get to the front of the 3nm line. This is where you should be skeptical. I could see Intel using TSMC 5nm processes, but 3nm, ahead of everyone else? Skeptical.

Intel won't use TSMC's 5nm process in 2023 because their own 7nm process will be operational by then and their 7nm process is equivalent to TSMC's 5nm process (similar to how TSMC's 7nm process is similar to Samsung's 5nm).

It isn't just process size but also transistor density. So in rising order from least dense to most:
Samsung: least dense. Their 5nm is equivalent to TSMC's 7nm. Which - in addition to yield problems - is why no one uses it unless TSMC doesn't have capacity.
TSMC: middle density. Their 7nm is equivalent to Samsung's 5nm but Intel's 10nm. They are #1 because of a combination of better density/higher yields than Samsung and smaller process than Intel.
Intel: most dense. Much more dense than Samsung - their 10nm and Samsung's 5nm are basically equivalent - but also more dense than TSMC. Meaning that when Intel actually reaches 5nm (and smaller) with their transistor density, that will be formidable.

Also, it has less to do with "Intel paying people off" (why do Apple fans always insist that everyone is corrupt but Apple?) and more to do with the FACT that TSMC won't be able to manufacture very many 3nm chips in 2023. Right now their current capacity is dedicated to 6nm, 5nm and 4nm. Qualcomm, MediaTek, Nvidia, AMD etc. will be using those for years. TSMC will have to build new facilities in Taiwan and New Mexico to accommodate 3nm and 2nm at scale. So, TSMC will only have the ability to fab tens of millions of 3nm SOCs in 2023. That will accommodate the iPad Pro and iPad Air devices that sell a year (the $329 entry level iPad doesn't use the latest SOC and neither does the iPad Mini) and some Xeon/Core i9/Core i7 that Intel will need to hold the line against the progress that AMD is making against them with servers with their Epyc line as well as in workstations and gaming laptops. That "may" be 100 million SOCs between Apple and Intel and could be as low as 50 million. However, about 225-275 million iPhones sell in a typical year and that isn't nearly enough.

A better question may be why Apple isn't using the 3nm SOCs for Macs. I was certain that I read that 4nm M1X chips with at least 12 CPU cores were coming this year for the 14' and 16' MacBook Pro as well as some Mac Mini and iMac models. If that is the case, 3 years would be quite awhile to be on 4nm. But if Intel "jumped in the line" ahead of anyone that would be AMD. 

coolfactor said:

Sweet, and I bet it's SUPER FAST!

I run the server version of the Ubuntu version of Linux on all of my VPSs, and I've been running Desktop Ubuntu on my ARM-based Raspberry Pi 4, but it's too slow to enjoy on a daily basis. So Linux has run on ARM chips for a while now, but now it sounds like they support even more.

Note, that this doesn't mean that everything that runs on Linux will run on ARM, yet. Individual "apps" (as we call them on macOS) still need to be compiled to support different ARM architectures. I've run into a few cases where stuff I use on Ubuntu on my x64-based servers do not run on my ARM-based Rasberry Pi.

Most stuff does though because ARM versions of the major distros have been out for more than a decade, and tons of ARM Linux servers are in the cloud. I installed a Linux app store on an ARM Chromebook a few months back just out of curiosity and all of the major apps were there and installed fine. Some stuff may need to be built from source but that is par for the course for Linux anyway.