Recently Mr. Frisch published an article on Seeking Alpha regarding Moore's Law. I will not pretend to know as much about IC manufacturing as Mr. Frisch; he has forgotten more than I know. Hopefully, I change that when I go back for an MSEE, but I digress.
In his article he details how Moore's law has been used for 50 years to drive the semiconductor industry. In this article I would like to detail why I think Moore's law has had a good run, but at least from an investment standpoint it may be losing steam.
I completed my undergraduate degree at North Carolina State in Electrical Engineering. However I currently work in an unrelated field, but have a good grasp of the fundamentals and theory behind IC manufacturing. It has been a few years though, so I'm sure there are some industry experts that will have some valuable insight in the comments section, and the better part of this article may very well reside there and not in the body of the article.
Why Node Shrinks and Wafer Size Affect Cost
Nodes relate to the size of a transistor. The smaller the number, the smaller the transistor. IC fabrication starts with what is known as a wafer. A wafer is a circular slice of silicon that has been prepared so it can undergo a process called lithography in which it is etched with transistors, creating ICs (integrated circuits).
There is a formula that relates the number of ICs to the area of the wafer and the size of the IC. The way fabrication facilities drive cost down is either to make larger wafers, which allow more ICs to be etched on a wafer at one time, or to make the transistor smaller so the ICs take up less space. These cheaper, smaller transistors are why we can buy a smartphone today at a lower cost with more processing power than we would've spent on a desktop in the late 90s.
But at some point the industry will reach the turning point where it is no longer economically feasible to continue this increased spending. The increase comes from a few factors. At each node shrink there are new hurdles to jump over, and as we reach quantum levels these hurdles become even higher, meaning it takes more R&D spending to jump them. And when we do figure out how to make it over those hurdles, the cost of the tools to do so also increases. In addition, say Intel (INTC) decides to transition to a larger wafer. This requires the construction of new fabrication facilities, which is another expenditure.
Intel's Increased R&D Spending
Image taken from INTC.com
Intel has increased R&D spending on average 21% since 2009. Intel has spent on average $7.5B a year since 2009 in MG&A (marketing, general and administrative), but it has been inconsistent from year to year, so for my calculations I am using a straight average.
In 2013, Intel is expecting to spend roughly $19B in R&D and MG&A. Assuming Intel increases R&D spending 20% in 2013, yields a total R&D expenditure of $12B. Subtracting the yearly average of MG&A of $7.5B from Intel's proposed $18.9B budget yields an R&D expenditure of $11.4B, therefore in my estimates I will use an estimate of $11.5B.
Potential Impact on Intel's Financials
Taken from INTC.com
Image taken from itersnews.com
I do not foresee R&D expenditure tapering anytime in the near future. R&D spending will most likely remain at a healthy 15-20% growth YoY. As fabrication facilities continue to try and transition to smaller nodes, the effects at microscopic levels come more into play, which will give rise to the need for more extravagant tools and methods to etch these smaller transistors. Assuming Intel spends $11.5B this year, and a 15% a year increase for the foreseeable future, Intel's R&D expenditure will increase to roughly $23B by 2018. If I use its 4-year historical average of 20%, this cost is $28B in 2018.
Earlier I had mentioned that the price to build the facilities would also rise due to more expensive tools. According to Intel management, capital expenditure was $5B in 2010, and around $11B in both 2011 and 2012 (each, not total), with a projected CAPEX of $12B in 2013. So both R&D spending and CAPEX are rising as a result of transitioning to larger wafers and smaller transistors, among other things.
INTC uber-bulls like to remind me that these smaller transistors are what give Intel the lead over the competition, and allow it such high margins. If Intel needs to compete, it can just lower its margins.
How Would Lower Margins Affect Intel's Business?
For perspective, Intel generated $18.9B in cash last year. Let's assume that Intel did have to compete by lowering price, therefore lowering margin. If Intel had a 50% margin, its cash flow last year would have been closer to $12.6B. If Intel had a 40% margin, the margin Advanced Micro Devices (AMD) enjoys, Intel's cash flow last year would have been $7.1B. In 2012 Intel paid $4.4B in dividends and stock buybacks of $4.8B. Reduced margins would affect Intel's ability to return value to shareholders.
If you couple Intel being forced to reduce margins to compete, along with factoring in rising CAPEX and R&D spending to maintain its lead over the competition, you see how quickly it runs into trouble.
Taiwan Semiconductor Manufacturing (TSM) operates at around a 45% margin, and is one of the world's leading suppliers of smartphone and tablet SoCs. These SoCs are not very expensive and do not command the same price as Intel's more expensive Core line of CPUs.
Images taken from INTC.com
The first graphic shows the income of Intel's "Other" category, which includes mobile phones and tablets. You can see it is currently operating at a loss. The reason was stated as being:
Net revenue for the Other IA operating segments decreased by $627 million, or 13%, in 2012 compared to 2011. The decrease was primarily due to lower IMC (Intel Mobile Communications) average selling prices and lower netbook platform volume.
PC sales declined 3.2% in 2012, and are expected to decline another 7.8% this year according to the IDC. The decline in 2012 led to a drop in net revenue for Intel of $1.2B YoY from 2012 to 2013; a drop of 7.8% in a year could lead to a loss of around $3B in revenue from 2012 to 2013. The declining PC sales have been blamed on Windows 8 dissatisfaction, along with the increasing popularity of tablets and smartphones.
Tablets are priced much cheaper than laptops. Intel is trying to market its more expensive, higher margin Core line of CPUs for tablets, but these tablets will come at a much higher premium than what consumers are used to paying. I am expecting these expensive tablets to at most to offset declining PC sales. These consumers are the people who are willing to spend money on a nice laptop but prefer a tablet. I expect most consumers will want to remain with the status quo, and spend the price of an iPad 4 or less, which will most likely mean they're buying Bay Trail-powered tablets. The competition from ARM Holdings (ARMH) powered devices will be much stiffer than what Intel faces from AMD in the x86 space, and therefore Intel will have a hard time commanding the same price.
I consider myself to be cautiously bullish on Intel. There is no shortage of articles as to why everyone is extremely bullish of this stock. Intel has a process and an architectural advantage over the competition. But as Intel competes in the tablet and smartphone market where the average selling price of the SoCs are lower, and there is more competition, I do not see Intel being able to command the same margins, since I am expecting its most successful chip to be the Bay Trail-based Atom.
A few things that could offset my theory are a reversal of the declining PC market, a high adoption rate of Core powered tablets, or a reduction in CAPEX/R&D spending. The first two should be easy to spot. As for the reduced CAPEX/R&D spending, that will take longer to unfold. As Intel brings 14nm fabs online, look for CAPEX and R&D to increase - that will be a sign of how hard it is fighting to get to 10nm. Intel is supposed to start production of 14nm silicon Q4 this year for a 2014 launch. A delayed launch would strengthen my theory.
Lowering margins and increased CAPEX/R&D spending YoY strengthen my argument.
Additional disclosure: I may liquidate my position in INTC at any time.