From their recent results call (Fiscal Q4 = calendar Q3) "For the Snapdragon 835 premium tier chipset, we are seeing strong demand exceeding supply. Our growth in operating margin relative to demand in the quarter is being limited by our ability to ramp volumes at 10 nanometer. We expect this to begin to normalize in the fiscal fourth quarter."
A lot of this is interesting but in many cases not relevent except for bragging rights.
Unless your volume is pretty big Intel and Samsung aren't interested, leaving TSMC and GF who are set up to deal with large numbers of tapeouts. If you want a lot of 3rd party IP TSMC will be first and Intel last. So almost everything except mobile AP -- a lot of volume but a small fraction of wordlwide chip tapeouts -- will end up at TSMC (or maybe GF) by default.
Mobile AP is an interesting one -- would you want your critical chip to be made by a potential competitor, even if they offered you a really good deal? Because that deal might be because they need to fill the fab, and when their own chip comes along who gets priority?
Technology benchmarks are great, but if the "best" technology doesn't have essential IP you need (when you need it) or the foundry isn't interested in your business you have to use the "non-best" one which meets these requirements -- which in very many cases means TSMC.
It's self-fulfilling, the more people who use them the more IP is developed and the more attractive the process becomes. And because pure foundry is their fundamental business model and they don't compete with their own customers, there's no risk of them pulling out of a market -- which is certainly more than you can say for Intel...
One thing is the theoretical density (TSMC focus for marketing), another one the density in shipping devices (Intel focus for marketing). All parameters can be changed on fly, for example Intel can increase 14nm density depending on customer if it ask for at a cost of lower yields at the beginning. Or it can decrease the density if more clock speed is necessary.
It's hard to say who has the higher density in these days, and there isn't a fixed border between nodes today but a grey scale instead with a stunning proliferation of subnodes.
I think your comparison with TSMC has some sense, anyway in this moment Samsung is producing like an IDM and we don't know about yields. They could be low but enough to do profitable Galaxy.
One thing is certain Samsung Qualcomm 10nm chips are not widely available to other customers and this is very suspect.
A little less intersting your comparison with Intel. Intel has an high power process to develop and even the mobile version is a medium power one. So it is more problematic for Intel to have a better process than 14nm in segments if its interest.
You can't have leading technology when you lack the talent necessary. Liarfoundries lacks the talent to develop any technology at this point. Most of the talent from IBM is either already gone or will soon be gone. They are hiring more H1B's as they fire long-tenured talent, claiming they are expanding. Expanding to China that is. How about we look at how many H1's Liarfoundries petitions for in exchange for local experienced jobs lost. Are they holding their promise? Of course not. They are liars.
What's left behind is an empty shell of managers and newcomers who have no idea what they're doing. AMD Ryzen is a bit late to the game, and Albany is a total tax-payer money wasting disaster "Research" facility. Their EUV program is so far behind it's a total joke.
Let Thomas "I will bend over to Arabs any day to keep my paycheck" Caulfield alikes enjoy their high salaries while the music lasts. As the talented engineers continue to suffer in this train-wreck of a company.
The foundry situation at 10nm might be a better example to describe who is leading. Samsung looks like it is leading with Galaxy S8 product coming out within a week. It is using triple patterning which was some concern but now it is ready. Intel just a couple weeks ago announced its 10nm which uses SAQP so its density is much higher than Samsung's, but their Cannonlake product is not yet available, maybe end of this year or start of next year. TSMC focused so much on developing 7+nm, I think it distracted them from 10nm a little. As a lower focus, 10nm will be later this year, for Apple. GlobalFoundries took this even further, will have no 10nm; readiness is N/A.