source: http://www.digitimes.com/news/a20170823PR202.html Semiconductor market data firm raises 2017 revenues estimates: -- global market will be +17% this year; previous guidance +11.5%, now higher due to rising chip prices -- memory sector revenues to be +50% this year; previous guidance +30.4% 2018 guidance +4.3% revenues
This is a trade I got completely wrong were I placed technicals over the broader trend developing into AI and AR. $107.89 needs to hold into year-end to see carried momentum into next year, but more importantly, $103.90 is a major level of support that if broken before the end of the year would put an end to the momentum trade and lead to further consolidation.
I hesitate to go against CNN's opinion piece but they seem to thrive on ignorance. GlobalFoundaries is not exactly working on discrete components with their 14 nano node. They can build world class stuff with good power consumption profiles. 14nm means they can't compete at the highest density level with 16 core CISC systems, etc. They can do 8 core fully modern CPU dies with near-leading edge power consumption and they have. That's almost every chip made. Intel doesn't have 7 nano figured out, yet. They keep announcing delays. How are they listed as a top fab? Meanwhile at Samsung, I understand their second gen 7 nano process is better but it's not fully online, as far as I know. Even here, I don't know how many features can tape out at 7nm, as the first generation 7nm node wasn't fully 7nm. I won't speak to TSMC, as I don't know how good their 7 nanometer node is but, even if it's the world beater it seems to be, it's not like you can't build the world's most powerful computer using 14nm chips and lots of fabs can do 14 nano. I don't want to dismiss the importance of the 7nm node but it doesn't seem intelligent to dismiss the 14nm node as "not advanced". Semi fabs are famous for announcing a new lithography node while only being able to tape out a small portion of the features at that size level. If they can tape out *anything* on that node, they say they have that process capability. Often times, the next node is barely better than the last one, even though it sounds like it's twice as good. It takes time to improve the efficiency of it and broaden the application. Suck my moist fart, CNN.
