- TSMC will give up the impossible chase of Intel or be smothered by Intel capital spending.
- Other than for Intel, any process below 28nm is uneconomic.
- Intel will be the last company standing in the 50-year run of Moore's Law.
I've spent the weekend looking at numbers, lots of numbers. The numbers have to do with the sustainability of capital expenditures in the semiconductor industry. For this article I'm looking to illuminate some gross disconnects between Intel (NASDAQ:INTC) and TSMC (NYSE:TSM) on capital spending.
Everyone who reads my articles will know that I am an unreconstructed Intel bull, albeit with bad timing. Also, I have nothing against TSMC, as a matter of fact I have great respect for the company and the company's chairman, Morris Chang.
TSMC has to be the focus of my analysis because it supports so many Intel competitors, and the simple fact that there is no comparable company with access to public financial records to compare Intel with.
So, here we go:
The numbers I want to generate are sales dollars per cap ex dollar and gross profit dollars per cap ex dollar. Of necessity, these numbers will be rough approximations, because it has to be that way. I need to first get into some definitions.
A company's annual depreciation and R&D can be considered the capital expense to support that particular year's revenue. In a steady-state environment, annual depreciation should approach new capital expenditures and R&D is what it is, to support the business.
In the semiconductor industry, by far the heaviest cap ex is for the leading edge technology, so I will try to identify the piece of each company that the bulk of the cap ex will be spent on.
In the case of Intel, the company is about 90% semiconductor, all of which is at the bleeding edge of technology, so the spending will apply to about 90% of the company.
In the case of TSMC, at least for 2013, the bleeding edge (28nm) is about 30% of the total sales, per the most recent earnings conference call.
For Intel, the adjusted numbers are about $44.3 billion revenue in server and client CPU products; The adjusted gross profit (correcting for some of the losing non-semiconductor divisions) is about $34 billion for a 77% gross profit margin on the semiconductor segments. That is an incredible number. The combined number for 2013 depreciation and R&D is about $17.4 billion.
For TSMC, the 30% of bleeding edge technology amounts to about $6 billion. The gross margin for TSMC has traditionally been 50% give or take, so I will use $3 billion as the gross margin on the 2013 28nm segment of the TSMC business. The combined number for TSMC 2013 depreciation and R&D isn't available yet, so, for the purposes of this article, I have to use the 2012 numbers. The 2013 numbers will be available in the first few days of April and, of course, will be much higher than the 2012 numbers. All that being said, the way I'm doing this will put TSMC in a much better light than if I had the 2013 numbers in hand.
When the 2013 numbers become available, I will update by an author comment.
The TSMC 2012 combined number for depreciation and R&D is about $5.9 billion.
|Sales B$||Gross Profit B$||Dep+R&D B$||Sales $/D+R&D$||GP$/D+R&D$|
So for every dollar of, let's call it operating capital spending, Intel spends, it generates $2.55 in revenue and $1.95 in gross profit. On the other hand, TSMC returns a $1.02 in revenue for every $1 it spends on bleeding edge technology, and only $.51 of gross profit.
Obviously, this is an unsustainable situation for TSMC; the company HAS to be losing big time on every 28nm wafer it ships. The other 70% of the business is subsidizing the new process efforts.
There are two reasons why Intel can make money in this environment, while TSMC bleeds heavily. One is the much-maligned PC business. While the PC business is no longer a growth business, it is a cash generator of galactic proportions and the justification for continued spending by Intel on advanced processes. $44 billion of dependable, repeatable sales generating $34 billion in gross profit is something that TSMC can never attain, no matter how much it spends.
The other element is the 77% gross margin on client and server chips. To calculate final price, you divide cost by the reciprocal of the stated gross margin. Thus, $1 of TSMC cost, at 50% gross margin, sells for $2. $1 of Intel cost, at 77% gross margin, sells for $4.35. That is another number that TSMC will never achieve, largely because the fabless business model can't afford it.
The unavoidable conclusion here is that TSMC simply has to stop spending money trying to catch Intel; it is at an impossible disadvantage.
Another unavoidable conclusion is that the TSMC customers who need bleeding edge processes in order to remain competitive will, I guess, not be competitive with Intel on any business that Intel wishes to pursue. So, the fabless model begins to collapse at the bleeding edge of semiconductor technology. These things may not happen overnight, but happen they will.
There is more and more writing on the idea that Moore's Law ended at 28nm. Again, quickly reviewed, Moore's Law states that the number of transistors on a chip will double every couple of years, THUS REDUCING THE COST OF TRANSISTORS BY SOME NUMBER APPROACHING 50%. That cost thing is the part that many people miss.
I have repeated many times (and absorbed a lot of heat for it) that TSMC (or anyone else) will never ship a commercially attractive 20nm planar chip in volume.
If TSMC is to follow Intel into the 10nm TriGate realm, it will have to spend a minimum of $10 billion dollars, and it will always remain 1-2 nodes behind Intel and lose massive amounts of money doing it.
In bleeding edge semiconductor technology, Intel will be the last company standing, supported by huge margins and a stable or slowly-growing PC business.
The next shoe to drop will be, if 28nm IS the end of the line, the three companies capable of making 28nm; TSMC, Global Foundries, and Samsung will commencing a price war of such intensity as to threaten the very existence (in N-1 or N-2 logic node) of all three companies.
During the next year, Intel will make major inroads into the tablet computer segment, and a year later, it will make major inroads into the smartphone business. From a capacity standpoint, Intel is trimmed up to do $80-$100 billion in revenue.
Apple is said to have given its A8 business to TSMC, maybe so, but I will believe it when I read it in a reliable teardown report. I think Intel will have the Apple fab business in the end and most people will actually be surprised by something so obvious.
Intel is now shipping Broadwell processors on a 14nm TriGate process, while all the next best-in-class are stuck at 28nm.
About a week before Intel announces earnings on April 15, the third -party marketing companies will give us their guess on first-quarter PC sales numbers. Based on some recent information, we could see some improvement in the PC business. I'm thinking maybe a 3%-5% increase YoY. That sort of information could produce a couple dollar bump in the Intel stock price, and a follow-up earnings beat could get another couple of bucks.
You can buy the stock by about April 10 to include possible good news on PC sales and the earnings report, or there are some Intel April 27 call options trading for $.08 for those that like to roll the dice. Eight cents!