The Truth Behind Intel's Manufacturing Lead

Summary

• Real nodes are apparently different from the advertised nodes.
• But even after normalizing nodes, we see that Intel is set to lose its process leadership post-2018.
• This may boost the appeal factor of GlobalFoundries, TSMC and Samsung and force Intel to slash its prices to stay competitive.

The longevity of Intel's (NASDAQ: NASDAQ:INTC) manufacturing lead is being questioned rigorously of late with its competitors coming up with smaller nodes on aggressive timelines. Bullish investors and tech enthusiasts argue that TSMC (TM), Samsung (OTCPK:SSNLF) and GlobalFoundries have process nodes that are far inferior than their advertised figures but it doesn't really help Intel's case by a great deal. I did some sleuthing of my own and found that even if we create normalized-node (not advertised node) roadmaps of all four companies, Intel's process lead would vanish in the next two years. Let's take a closer look.

(Image Source)

Normalizing Nodes

Let me start by saying that process nodes advertised by Intel, Taiwan Semiconductor, GlobalFoundries and Samsung are drastically different from each other in reality. For starters, TSMC's 16nm process reportedly uses a 20nm backbone. Chips fabricated by each of the fabs tend to have parts with dimensions vastly varying from each other. These advertised nodes have been basically reduced to just marketing material for the aforementioned semiconductor firms and the advertised figures aren't an accurate representation of physical dimensions used by any means. In fact, a chip manufactured on the 14nm node by Samsung or GlobalFoundries may not have any part that exactly scales 14nm.

The current state of semiconductor manufacturing is such that nobody follows a universal node naming process. In a bid to simplify this confusing scenario, ASML, which is the world's largest supplier of photolithography equipment for semiconductor manufacturers, has reportedly come up with a formula to normalize nodes based on their effective feature sizes. Granted that the formula is far from perfect and doesn't factor in other features such as SRAM cell sizes, but it's currently the best resource available to us to separate reality from marketing FUD. The calculation goes as follows:

Standard Node = 0.14 x ((CPHP x MMHP)^0.67)

Here CPHP = Contact Poly Half Pitch, and
MMHP = Minimum Metal Half Pitch

A research report published by IC Knowledge last year calculates these values based on its CPHP and MMHP projections for the nodes to come over the next 4-5 years. The indicative results are as follows:

	32nm/28nm	22nm/20nm	16nm/14nm	10nm	7nm	5nm
Global Foundries	25 nm	NA	17 nm	NA	9.2 nm	7.1 nm
Intel	31 nm	21 nm	13 nm	9.5 nm	5.9 nm	NA
Samsung	27 nm	20 nm	17 nm	12 nm	9.2 nm	7.1 nm
TSMC	27 nm	18 nm	18 nm	12 nm	9.2 nm	7.1 nm

Click to enlarge

Making Sense of the Data

Key thing to note here is that Intel's nodes are the closest to its advertised nodes, at least as per the aforementioned formula. Samsung, GlobalFoundries and TSMC may market their fabs to be groundbreaking or industry leading, but all of them seem to be approximately one node behind Intel as of now. But the above table doesn't provide much insights if we look at it in isolation. So I plotted the above values on a time-scale to have a better understanding of the evolving trends.

(Compiled by author)

It is evident from the chart above that Intel is enjoying a process lead as of now. It's 10nm chips won't be bested by offerings from Samsung, GlobalFoundries or Taiwan Semiconductor for another year. In fact, during 2018, Intel's competitors are projected to marginally surpass it in terms of fabrication process. The difference between Intel's 9.5 nm and 9.2 nm chips from its competition is quite small to make any significant dent in the industry dynamics.

However, 2018 marks the end of Intel's manufacturing lead. The chipzilla has a relatively advanced manufacturing process but its lead would vanish on an advertised as well as normalized basis next year when Samsung and TSMC start mass producing chips based on their 7nm advertised node (or 9.2nm calculated node).

Things deteriorate for Intel from thereon. TSMC, GlobalFoundries and Samsung would take the driver's seat and leave Intel far behind in terms of chip fabrication technologies. Intel won't be releasing it's 7nm chips (or 5.9nm calculated) until 2021-2022 and this provides its competitors enough room to surpass the chipzilla for good.

A similar roadmap was released by EETimes last month. Although it has a different set of normalized node values, the conclusion drawn out of it is the same. Intel is projected to not only lose its manufacturing lead beyond 2018, but it could also start lagging its competitors in the fabrication process department thereafter.

(Source: EETimes)

Impact on Intel

The impacts of Intel's loss of manufacturing lead are manifold. For starters, its future iterations of x86 CPUs may no longer have a performance or efficiency advantage compared to AMD's (NYSE: AMD) offerings on a common normalized node. Let me remind readers that AMD's Ryzen flagship already has been seen matching Intel's high-end Broadwell-E i7 6900k chip, in spite of the former being on an "inferior" node. So it's a very real possibility that AMD's future chips get even more competitive in both retail and server segments once the chipmaker gets on the same node as Intel, post-2018.

Also, ARM-based server chips from the likes of Qualcomm (QCOM), AppliedMicro, Cavium etc., haven't been able to make a dent in the market yet due to myriad reasons. But by advancing to a common node with Intel, they might pack in a good fight in terms of performance, efficiency and costing metrics, thereby encouraging enterprise clients throughout the industry to reconsider ARM as a potential x86 alternative. If ARM manages to corner even 5% of the server market, it would be a big blow to Intel as it happens to be the most profitable business segment for the chipmaker.

I also believe that Intel's pricing power would take a hit due to this emerging trend. The chipzilla won't be able to charge a premium for its products when cheaper x86 and ARM alternatives from AMD, Qualcomm and others would be available at the same node, fabricated using similar processes, with the only differentiator being the chip architecture. So unless Intel's competitors botch up their future releases big time and come up with duds, there's a high chance that Intel would be forced to discount its products to remain competitive.

It is due to these reasons that I'm expecting Intel's profit margins and ASPs to deteriorate over the next five years.

Investor takeaway

I'd like to point to readers that the above calculations may not be absolutely correct as they are based on estimates. In fact, the normalized node values provided by EEtimes, IC Knowledge and Semiwiki are all different from each other. So we have to factor in some room for variance as well.

But the projected nodes provided by all three media outlets does indicate one common trend - Intel is set to lose its manufacturing lead post-2018. Therefore, I recommend that investors rethink their investment thesis. Intel may not be able to sustain its industry-wide dominance in the chip industry for very long without its manufacturing lead.

Editor's Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.