Samsung may have scored a one-time coup over TSMC for Apple’s business at 14nm, but that increasingly appears to have been a lucky strike. TSMC has been Apple’s sole supplier since the A10 in 2016, including this year’s A12 Bionic with the Apple iPhone XS, XS Max, and Xr. Combined with recent news on TSMC’s 7nm EUV and 5nm risk production, however, the implication is that we won’t see much in the way of a node improvement for Apple to work with in 2019.
TSMC has just announced that it’s taped out its first 7nm EUV design, with EUV being used for those parts of the chip that don’t require a pellicle. TSMC also stated it would begin risk production on 5nm in Q2 2019, but risk production typically takes a year to ramp into the kind of volumes that Apple would require for an iPhone launch — and the EUV pellicle solution required for 5nm isn’t ready yet, as far as anyone knows. That means the gains for a second-generation 7nm could be small indeed, given that EUV isn’t really expected to deliver any in and of itself.
It’s expected that we’ll first see a new A12X before the A13 debuts, with the A12X dropping into updated versions of devices like Apple’s iPad Pro. TSMC is looking into using a more advanced form of packaging to improve 7nm performance and capabilities, with its integrated fan-out (InFo) technology making the jump to 7nm (it’s previously been deployed on older nodes). Using fan-out technology allows for heterogeneous components to be packed more densely, which can reduce power consumption and improve performance by packing bits more closely together.
We talk about the foundry business as being a multi-way race, but at this point, it’s really down to two companies — TSMC and Intel. Samsung is said to have only Qualcomm and its own internal foundry work as a customer at 7nm. GlobalFoundries is out of the race. The number of customers who can afford to make transitions to new nodes gets smaller every time a new node debuts, and TSMC commands 56 percent of the merchant-foundry market. That’s a staggering market share for a single company in a field ostensibly crowded with competitors. The 2017 foundry revenue rankings show just how lopsided the market is:
As things stand today, it’s increasingly shaping up like a two-way fight between TSMC and Intel at the leading edge, even as the question of how much being on the leading edge matters is increasingly uncertain. Second-generation 7nm EUV isn’t expected to drive major performance improvements or area reduction, as that won’t happen until 5nm. Expect a lot of discussion about how much it matters for AMD or Nvidia or other major companies to be sitting on TSMC’s 7nm next year while Intel is deploying its own 10nm, but assuming both process nodes work properly, the gaps may not be that large. As the benefits of each node shrink, the difference between being stuck on the last one and using something at the cutting edge will be commensurately smaller.
Now Read: TSMC Announces First EUV 7nm Risk Production, 5nm Tapeouts in Q2 2019, AMD Moves All 7nm CPU, GPU Production to TSMC, and Report: Intel Will Outsource Chipset Production to TSMC