Intel Expects to Reach Process Parity With 7nm in 2021, Lead on 5nm
According to remarks Intel CFO George Davis made at a Morgan Stanley conference this week, the company still believes it has a ways to go before it matches the pace of its foundry competitors and retakes overall process leadership. Reports of his remarks at the conference suggest Intel won’t regain parity with TSMC and possibly Samsung until it launches 7nm parts in 2021, with the firm retaking leadership at the 5nm node. Intel has previously said it would launch a GPU on 7nm in 2021, and Intel CEO Bob Swann has stated that 7nm CPUs will ship in Q4 2021.
A report on the conference from Tom’s Hardware quotes Davis as saying:
So we bring a lot of capability to the table for our customers, in addition to the CPU, and we feel like we’re starting to see the acceleration on the process side that we have been talking about to get back to parity in the 7nm generation and regain leadership in the 5nm generation.
It’s not clear to me if Intel ever specifically stated it would regain leadership at 7nm or if it merely implied this would be the case. Remarks made by Murthy in late 2018 suggested Intel might be targeting 7nm for a return to its previous manufacturing cadence, but didn’t specifically state Intel would overtake its competitors at that node. Here’s the relevant quote: “One thing I will say is that as you look at 7-nanometer, for us this is really now a point in time where we will get EUV back into the manufacturing matrix, and therefore, I think, that will give us a degree of back to the traditional Moore’s Law cadence that we were really talking about.”
Intel hasn’t given a date for when it might introduce 5nm, but a two-year cadence would put the launch in 2023. TSMC expects to be building 3nm chips by 2022, but companies like AMD and Nvidia are node followers now, not the node leaders they used to be. If TSMC ships 3nm in 2022 for Apple, a 2023 launch date for AMD or Nvidia products would actually make good sense — though this assumes that Intel is actually aiming for a 5nm launch in 2023 in the first place.
Frank Words on 10nm
Davis’ comments are noteworthy for their overall assessment of Intel’s 10nm node and the company’s willingness to be honest about that situation.
As we said back at our analyst day in May of 19: Look, this isn’t just going to be the best node that Intel has ever had. It’s going to be less productive than 14nm, less productive than 22nm… the fact is that I wanted to be clear what was happening during the 10nm generation. The fact is, it isn’t going to be as strong a node as people would expect from 14nm or what they’ll see in 7nm.
There seem to be some parallels between the troubles Intel is having on 10nm and the knock-on effects on its products and some of the issues AMD had with fab production and node progression back over a decade ago, when it still owned its own fabs. Intel can’t just flip a switch to change a line from 14nm to 10nm or vice-versa, and managing demand for existing products while transitioning capacity over to building new hardware is a difficult balance to strike. Unlike AMD in the mid-to-late 2000s, Intel is making record quarterly profits — but new nodes often suffer from yield issues.
There’s a quote that I like to refer back to about the difficulty of microprocessor design. Supposedly by Robert Palmer, former CEO of Digital, it goes like this: “Designing microprocessors is like playing Russian roulette. You put a gun to your head, pull the trigger, and find out four years later if you blew your brains out.”
Intel’s difficulties at 10nm are a perfect illustration of the problem. Six years ago, Intel delayed finishing Fab 42. Had Fab 42 come online on its original schedule, Intel would be facing fewer capacity constraints right now. Intel also made the decision to move its smartphone modem manufacturing business in-house, which ate up some degree of additional capacity. In the past, Intel had decided to link its node transitions and new architecture development efforts in a sequence it called tick-tock. Tick-tock worked brilliantly for nearly a decade, before coming apart on the shoals of 10nm.
At some point, Intel made the decision to move on from 10nm more quickly than it would have if this had been a typical node. Frankly, that’s probably the right thing to do. 10nm simply hasn’t been a typical node, and it may be that EUV is capable of resolving some of the issues that have made it more difficult to put 10nm in production.
As for the question of where this will leave Intel versus AMD, that’s rather more complicated than it might seem. AMD’s Ryzen Mobile 4000 APUs will launch soon. We already know that Intel has a 10nm mobile refresh coming later this year and mobile has been the toughest place for AMD to gain market share against its rival. If Intel continues to improve CPU IPC and can win back some of the clock speed it gave up in the Coffee Lake – Ice Lake transition, we could see highly competitive scenarios even if TSMC has a process node advantage. We haven’t heard anything about 10nm desktop CPUs, but capabilities like AVX-512 support may also prove to be an effective bulwark for Intel in certain contexts.
Overall, Intel is warning investors to expect gross margin pressure from multiple fast product ramps and for the company to lose some market share to AMD as it ramps up its efforts in mobile and server. The guidance it gave at Morgan Stanley doesn’t fundamentally change that stance, but Intel did appear to be warning a bit more strongly than before that it might be facing headwinds.