The media already has Apple and Intel in talks to make the A11 SoC in 2018 as a result of the recent Intel/ARM IP licensing deal. This is probably one of the funnier media bumbles I have read in a while so let’s talk about it in a little more detail.
“According to Nikkei Asian Review, Intel is now perfectly poised to give TSMC a good run for its money in as little as two years because any Apple chips after the A10/A11 should be fabricated by Intel.”
First a little bit of history: As we wrote in “Mobile Unleashed” Chapter 7, Intel had the opportunity to make Apple’s SoCs from the very start but Intel again could not see the forest for the trees:
“According to ex-CEO Paul Otellini, Intel had been in discussions about a mobile chip for Apple before the original iPhone design. The cloak of secrecy on the iPhone gave Intel little to go on, and they were skeptical of Apple’s volume projections. “There was a chip [Apple was] interested in that they wanted to pay a certain price for and not a nickel more, and that price was below our forecasted cost. I couldn’t see it,” said Otellini. Intel passed.”
The rumors of Intel making SoCs for Apple have persisted and will continue to do so no matter how ridiculous it is. The good news is that I have won many lunch bets against Intel making Apple SoCs and will continue to do so, absolutely.
The latest rumor comes from the Intel Developer’s Forum (IDF). IDF is an Intel sponsored conference to created to promote Intel and products based on Intel technology (the first IDF was in 1997). The last IDF I attended was the launch of Intel’s 14nm product in 2014 where Intel CEO (Brian Krzanich) told us that 14nm was yielding on schedule only to recant at the next investor’s call. Apparently you can say things at IDF that you can’t say to Wall Street but I digress…
Also read: Intel Comes Clean on 14nm Yield!
This year’s IDF was again in San Francisco and now that the dust has settled there were two big revelations in regards to the manufacturing of semiconductors:
The ARM announcement is being overblown and the EUV announcement is being underblown so the Intel PR group is doing a great job.
Background:Intel has officially been in the foundry business since 2010 but has yet to get significant traction. At 22nm, Achronix and Netronome are lead customers with very low volumes. At 14nm, Altera and Spreadtrum are Intel foundry customers but we have yet to see silicon. Meanwhile the rest of the SoC and FPGA industry has been in 14nm/16nm HVM for more than a year.
Intel licensing the ARM IP for 10nm foundry business is wishful thinking at best. First, the IP that Intel licensed is the ARM standard cell and SRAM libraries, the building blocks of SoCs. Second, the big SoC vendors make their own standard cell and SRAM libraries and have already taped-out their 10nm designs and will be in HVM mid 2017, about the same time Intel 10nm will be ready to get STARTED with the new ARM IP.
LG is the announced first Intel 10nm customer which makes sense. LG does not have the volume to work with TSMC 10nm as an early adopter, LG competes directly with Samsung, and GlobalFoundries is skipping 10nm so what foundry choice did they really have? Besides, the majority of LG phones use SoCs from QCOM and MTK so this really is a low risk business proposition.
Bottom line: This announcement is a big fat nothing burger with cheese…
The “no EUV at 7nm” announcement was much more meaningful. At the semiconductor conferences earlier this year Intel stated very clearly that they would use EUV at 7nm while TSMC moved forward with a non EUV 7nm (TSMC will start 7nm production in 1H 2018 and hit serious HVM in 2H 2018 with the iPhone 8). So TSMC was RIGHT and Intel was WRONG about EUV. This is not a big technical change but Intel 7nm will be even more expensive since EUV is a significant cost reduction.
Yes, Intel will argue that their 10nm and 7nm are better than the foundries (TSMC and Samsung) but that will have to be proven at the chip level which is based on PPAC (power, performance, area, AND cost). The foundries have beaten Intel at every node based on SoC PPAC and I do not expect that to change at 10nm or 7nm. If you disagree I will cover all lunch bets.
Also read:If an Intel 10nm transistor fell in the ARM forest