I almost can't believe this comment. TSMC literally has an entire paragraph in parents comment that addresses it. The comment is about the much larger semiconductor business, not bleeding edge. I fail to see from any perspective how you can read that comment and post this.
Parents Comment >> People talk a lot about TSMC and the bleeding edge business(computers, phones), but that's just one part of it, but there's a whole industry of embarked software that will likely be much bigger.
Your comment >> TSMC literally has an entire paragraph in parents comment that addresses it.
That looks like a single sentence, not an entire paragraph... And it's a paragraph without supporting evidence. No shocker c7DJTLrn wasn't convinced about parent commenters weak claim.
There is no global semiconductor market. Saying it exists is the same thing as saying that global vehicle market exists and Boeing competes with Ford in making the best selling vehicle.
There is, however, "consumer chips foundry market", which is dominated by TSMC with 55%+ market share. And TSMC's market share is limited only because of their production capacity.
A vehicle is a finished good. A semiconductor is a semi-manufacture. Comparing it to a finished good with an end-user purpose is nonsensical. You could compare the global semiconductor market to, for example, the global steel market.
Semiconductor market is PARTIALLY finished good, if we consider creators market (gigs economy), where sell microcontrollers (RPI2040), and some other semiconductors.
It is market of small series custom devices, prototypes (like Oculus Rift), aftermarket upgrades, like cars tuning.
Yes, creators market is not too large, total number of sells just around million, but it exists.
BTW it is interest question, how large part of steel market also belong to creators market, if it is larger than semiconductor creators market.
>People talk a lot about TSMC and the bleeding edge business(computers, phones), but that's just one part of it, but there's a whole industry of embarked software that will likely be much bigger.
I have no idea what "embarked software" means. So-called "bleeding edge" makes up a significant portion of semiconductor profits, so how can you disregard that and say semiconductor is going bust?
Those semiconductor companies can make lots of revenue, but operate on thin margins, because a competitor can also build a fab. Think graphic cards, Radeon(now AMD) has consistently built worse graphic cards, with less R&D, but sold at a cheaper price point and survived throughout the years.
TSMC has some extra expertise that others doesn't, and has good profit margins because of that, but together with NVidia, they are the champions, and others are constantly catching up. As with many things in a capitalist system, prices face a race to the bottom.
Related to bleeding edge fab. 14nm has suffered some shortages, but even that has changed.
One example is Apple, even though they've managed to increase their revenues (and products sold) a lot because of Covid, that demand is down. Apple is projected to have less revenues this year than the last years.
The same has happened with PC sales. What has been really driving revenues and profits up in this space is AI.
I don't believe this will continue. More competition will drive profits down, meaning it'll be a less attractive field to invest for VCs or investors.
Semiconductor business is more akin to package food products in my point of view. Some products can enjoy a nice margin because of their brand, taste or being hyped at a specific moment, but the majority of it compete on thin margins.
It isn't like Meta, a closed garden, that has plenty of data and have a product that got everybody addicted and dependant on it, either social networks or communications, that because of network effects is extremely sticky. Where only regulations and government intervention could pose a risk to it.
But wall street and people's expectations(and valuations) on semiconductors is suddenly as if they are building the next monopoly. Monopolies only exist because there's only one, what I see is a reaaaally fragmented market, with high CAPEX, thin margins and low moat.
> but operate on thin margins, because a competitor can also build a fab
Building new fabs, especially for the advanced nodes, costs billions, takes years, and requires a lot of know how and special equipment. There are only a handful of companies operating and building new fabs.
> TSMC has some extra expertise that others doesn't [...] but together with NVidia, they are the champions, and others are constantly catching up
Not sure what you mean. Nvidia uses TSMC for the manufacturing for the 40 series (and Samsung for the 3000 series) and doesn't have their own fabs.
> Think graphic cards, Radeon(now AMD) has consistently built worse graphic cards, with less R&D, but sold at a cheaper price point and survived throughout the years.
You mean ATI (now AMD)? They got rid of their fabs back in 2009 and put them into GlobalFoundries. They now also mostly use TSMC for their CPUs and GPUs (and profit from the more advances nodes compared to GF) and GF for a few things in larger nodes.
Apple also (just as everyone else you have mentioned) uses TSMC for (most of) their chips and is one of TSMCs largest customers. Intel and Samsung have their own fabs, but also seem to lag behind TSMC.
I'm not comparing TSMC with Nvidia, I said they are champions in that business, given that there are companies that design chips and also has fabs, my comment makes sense?
I'm also sure here people know how fabs are expensive.
I believe you understood my comment related to Radeon being part of AMD. And also you need to work on being less pedantic.
