Intel Corp (INTC)

[This Causes an Error]

Why can Apple and Samsung do what Intel can't, despite the fact that they are (economically) much bigger than Intel?

Samsung's logic wafer revenue is miniscule compared to Intel's, especially after having its large anchor customer taken away by TSMC. Seriously, I would estimate Samsung's logic biz at about $2 billion, not even in the same ballpark as Intel.

Anyway, you are missing something important. Samsung's core biz isn't logic manufacturing. Even within semis, its core business is DRAM and NAND. So Samsung can put a low yielding process into production, eat dirt for margins, and it isn't going to hurt the company's bottom line much, especially since most of its wafer volume these days is for chips that it wraps in expensive smartphones (bundled with its DRAM, NAND, etc.).

Intel's entire business, save for a few excursions in NAND flash/3DXP, is basically from X86 MPUs manufactured in its factories. If Intel's cost structure is poor (i.e. they take a low yielding process and shove it into the production fabs to build all of their products), then this will have a significant negative impact on the company's overall profitability.

This is one of the reasons that the foundries can "move faster." Samsung's core biz isn't logic, so low yields don't matter. TSMC's core business is being a general purpose foundry with a large portion of its chip shipments on non-leading edge manufacturing nodes. So even if it puts a relatively low yielding/high cost structure tech into production and sees relatively low margins on them (and you will hear it in TSMC's calls, new processes take a LONG time before they hit corporate avg margins), it still has plenty of revenue from mature, high yielding nodes to keep overall profitability up.

You are starting to see Intel transition to this model somewhat. For example, their 2018 desktop/high perf notebook lineup (Coffee Lake) is Cannon Lake but implemented in 14nm, while the thin and light notebooks go to 10nm. This way, Intel can go into production on its relatively immature 10nm process, taking it on the chin margin wise for those products, while keeping the bulk of its business thru the majority of 2018 (desktop PC processors, high perf notebooks, and servers) on the older, mature 14nm node.

I think we will see more of this over time, something that will allow Intel to transition to new nodes more quickly. We could very well see a 7nm low power Core CPU family in 2H 2019/1H 2020 (basically just the "Y" and maybe low end "U"), but the bulk of the company's volumes (servers, desktop, high performance notebooks) could be on 10nm+/10nm++.