Intel May Have 10nm Hardware In-Market Faster Than Expected

2018 hasn’t been a good year for Intel; the company has been rocked by security problems from Spectre and Meltdown and pinched by manufacturing issues related to the ramp of its long-delayed 10nm process. Earlier this year, the company told customers to expect 10nm shipments to be pushed back all the way to Q4 2019. It’s repeated that guidance several times since, but the timeline may have improved in recent days.

Last week, Intel acting CEO Bob Swan published a letter in which he wrote “We’re making progress with 10nm. Yields are improving and we continue to expect volume production in 2019.” This omitted the specific reference to “Holidays” 2019 that Intel had previously given, but it wasn’t clear if the company was changing its guidance or was simply choosing not to emphasize that it would be more than a year before it could ship 10nm parts in volume in a letter that was supposed to be positive and forward-looking.

A note from BlueFin Research Partners yesterday sent Intel’s stock price spiking, however, with claims that 10nm might be in-market sooner than expected. The note claims that Intel is making “significant strides,” and has held informal talks with suppliers about ramping products before June. If Intel could ramp before June, that could put hardware on shelves in time for the back-to-school season — and that’s a major buying cycle.

According to SemiAccurate, Intel has made significant changes to its 10nm to push the tech out the door, relaxing its design rules and changing the nature of the implementation. It’s not clear how significant these changes will be. S|A earlier argued that the new rules would leave Intel’s 10nm more equivalent to a 12nm process node, but its most recent update argues that “the new downgraded ’10nm’ process from Intel will not take as big a hit from the removal of this tech as SemiAccurate said earlier, but it will still take a hit.”

S|A should be taken with a grain of salt, but the idea that Intel would tweak 10nm to get it out the door more effectively isn’t surprising. Nor is the idea that the company might have had to back off its initially-aggressive plans for 10nm in order to make up for lost time. Intel’s original plan for 10nm involved an aggressive shrink it referred as “hyper-scaling,” complete with a new metric for measuring node benefits that would’ve played up the advantage of Intel’s manufacturing technology.

Right now all of this is speculation and it’s still possible that the Q4 2019 deadline remains the official goal. Intel could introduce 10nm in limited fashion in the first half of the year but wait to transition most parts to the node until yields improved. This would echo the original Broadwell ramp, when 14nm initially deployed in mobile but desktop chips waited to take the leap.

The impact on the competitive situation in the x86 market is similarly difficult to predict. If I personally had to bet on which issue represented the larger potential benefit to AMD: Intel being delayed on shipping 10nm chips (giving AMD the opportunity to ship 7nm first) or Intel’s demand constraints (possibly allowing AMD to ship CPUs to a larger percentage of the overall market) I’d pick the demand constraints. Obviously, in this case, the two issues are causally linked, but the benefits of node shrinks have become uncertain enough that I’m not sure how much upside AMD can really expect to see from a 7nm shrink alone. It’s not that there won’t be improvements — it’s that the improvements specifically attributable to process are an ever-smaller percentage of the pie, even as the pie is getting smaller.

As the graph above shows, AMD’s stock price has gotten hammered at the same time Intel’s has soared, but it’s not clear either shift is really warranted. Intel putting 10nm into production isn’t going to dramatically transform the CPU industry, and AMD competing against Intel’s 10nm a few quarters earlier than expected shouldn’t be a dramatic or particular problem for Ryzen 2, either.