Our 6/15 report of more 10nm Intel delays confirmed by leaked info…
The delay appears to have interrupted Intel’s Tick Tock cadence…
Kabylake replaces Skylake – Cannonlake pushed out over horizon?
The news we broke is now confirmed…
On 6/15 we put out a report that broke the news of further delays at Intel 10nm. This appears to now be supported by a report of leaked Intel documents on 6/23 by the website benchlife.info which shows specifications for a new line of processors for 14nm called Kaby Lake to replace Skylake (14nm) and precede Cannon lake at 10nm.
The information shows another 14nm processor family called Kaby Lake which is to follow Skylake which follows Broadwell. Skyline was originally supposed to be followed by Cannon Lake at 10nm process. It would now appear that Cannon Lake has been pushed far enough into the future to be beyond the visible horizon.
Tick Tock skips a beat.. “The case of the missing Tick”….
Intel has been doing what it has called “Tick Tock” changes in technology, with a “Tick” being a node shrink and a “Tock” being an architectural improvement at the same technology node.
Broadwell was a “Tick” down in technology nodes to 14nm with Skylake a “Tock” at 14nm but with a new architecture. Cannon Lake was supposed to be a “Tick” down to 10nm but instead we are getting a second “Tock” of Kaby Lake at 14nm.
Broadwell = Tick
Skylake = Tock
Kaby lake = Tock
Cannonlake = Tick
Maybe Intel can fix this by calling Kaby Lake a Toke. Warmed over Kaby Lake marks time while waiting for real Tick to 10nm Cannonlake. It seems rather obvious that Intel has come up with a bit of a stop gap excuse while waiting on 10nm. Not only does it upset the cadence they have set in the market but it clearly reduces manufacturers desire to use Kaby Lake as its obviously less of an improvement over previous models of product.
Imagine a car manufacturer having problems with their 2016 model car for some reason and they decide to put out a 2015 and a half model year as a stop gap. Probably not going to get a lot of buyers. People will stop buying and wait for the real thing at 10nm.
10nm delay must be very significant…
We can’t imagine that Intel would go to all this trouble if 10nm were right around the corner or didn’t have a significant delay. This is clear evidence of a longer than expected delay.
Not only is it a replay of 14nm but could be worse as 20nm did not have an extra Tock thrown in while waiting for 14nm. You are not going to do Kaby Lake for a 3-6 month delay…its more like a year.
Energizing TSMC & Samsung..
In the continuous LeMans race of technology, TSMC and Samsung have perennially eaten the exhaust of the long time leader Intel, usually from a good distance. Now, not only are they closing the gap and up on Intel’s bumper but smoke may be coming out from under the lead car.
It could be that TSMC & Samsung experience the same issues as Intel at 14nm and 10nm but so far 14nm does not appear as much as a problem for them as it was for trail blazer Intel.
If we were in their shoes we would likely be stepping on the gas as hard as possible. As we mentioned in our prior note this does not bode well for Intel’s mobile aspirations nor its foundry hopes as the only advantage Intel had was its technology lead.
As expectations for Intel’s Capex have continued to fall this news is not likely much of a surprise. However for those expecting an uptick in 2016 spending by Intel we think the odds of that are low given that not only has 10nm been pushed deeper into 2016 but it has probably started to roll into 2017 at this point.
We don’t expect 10nm to get out of Intel’s R&D fab until the first half of 2016 which probably translates into equipment deliveries starting mid 2016 at best and rolling forward.
15 ASML EUV tools likely pushed out….
It was already clear that Intel wasn’t going to use EUV at 10nm and they also said they had a clear path to 7nm without EUV as well so we found it odd to hear about a 15 tool order as we thought it was clearly not written in stone.
Well the 15 tool order may turn out to be written in the sand if Intel pushes its schedule out for 10nm. This is a significant negative turn of events for ASML given that the order was announced not that long ago.
On the other hand this likely increases the time available for ASML to get EUV to work as recent progress seems to have slowed a bit.
Obviously the delay helps multi patterning companies with Lam being at the top of that list followed by AMAT. Further delays at Intel do not help metrology and yield management companies as Intel tends to be heavier in this spending area than most others. It may be made up for by Samsung and TSMC stepping on the gas but its not yet clear.
Coupled with DRAM issues we think this makes 2016 look relatively flat overall as Intel and DRAM weakness is offset by 3D and NAND spend.
The challenges are unclear yet clear…
We have not heard much about the problems Intel is having as it is deeply under wraps. It could be a combination of technical and financial issues at work…we just don’t know.
We would like nothing more than to see Cannon Lake come out on time and wake up to find Kaby Lake and 10nm delays to be just a passing nightmare….but we don’t think thats the case.
The issue of 10nm is more problematic than the embarrassment of the 14nm delay as Intel is in a more critical situation with both its competitors and its markets than before at the 14nm node. Pulling this off will test the mettle of management.
“Bad companies are destroyed by crisis, Good companies survive them, Great companies are improved by them.” Andy Grove
Semiconductor Advisors LLC
Also Read: Further Delays for Intel 10nm?Share this post via: