In a research report published by Morgan Stanley's Joseph Moore, the argument is made that at a $3.7 billion operating expense run-rate for Intel's (INTC 4.88%) Mobile and Communications group (a value published first here on The Fool), Intel would need to generate anywhere from $8 billion to $10 billion in revenues, assuming gross margins between 30%-50% in order to break even. As a result, Moore believes that Intel should simply stop developing cellular modems, as he believes that this is the bulk of the expense.
Intel would be condemning itself if it were to stop cellular/connectivity
Let's put aside the fact that the idea that it takes $2 billion/year to keep a pipeline of cellular basebands and connectivity, as alleged by the report, is absurd, and explore the idea that Intel should simply "stop" developing cellular basebands. Without the cellular/connectivity solutions, Intel would immediately suffer the following consequences:
- Current Mobile & Communications Group revenue would be within spitting distance of $0 (most of Intel's business today in this division is modems, not apps processors).
- Intel would no longer have discrete modems to pair with its high-end smartphone/tablet solutions, and would have precisely zero chance of selling into the mass market/low end without an integrated solution (which, again, requires that baseband IP).
- Intel develops connectivity (Wi-Fi, Bluetooth, NFC, etc.) for use in its PC platforms, and a logical next step to capture more of the PC bill of materials is to integrate this capability onto the same piece of silicon for low-end PCs.
Without both cellular and connectivity, Intel's long-term prospects in both tablets and smartphones -- particularly as these are cost sensitive and will almost all require integration -- would be absolutely grim. The suggestion from Morgan Stanley is tantamount to Intel quite literally throwing in the towel on Mobile altogether.
Connectivity and cellular do not cost $2 billion to develop
If we assume operating expenses of about $3.7 billion, then the entire R&D for the division is perhaps between $2.5 billion to $3 billion. Now, take a look at a modern mobile system-on-chip, and how many individual "blocks" are required for just one product:
Source: Qualcomm
Are we really supposed to believe that the design, validation, and test of a single IP block in what is a multifaceted system-on-chip, each with its own very sophisticated IPs, is what costs $2 billion/year?
Intel probably has in its pipeline, at any given moment for mobile, at least three (possibly four) "base" generations of system-on-chip designs. Each of these requires the following:
- CPU
- GPU
- Modem
- Multimedia
- Memory interface
- Camera/Image signal Processing
- Display
- Connectivity
Further, Intel -- or any other company -- is likely to want to design and validate multiple variants of these chips for different market segments. So, not only do each of these individual blocks need to be designed and validated, but each full system-on-chip needs to be designed and validated. Not only is all of that true, but the company also needs to write drivers and the other various support software needed to actually get these products launched. And I suspect that this is an area that Intel is having difficulty with, given that its 22-nanometer yields are excellent, but its 22-nanometer Android products have yet to roll out.
Source: Broadcom
So, no, cellular and connectivity probably don't cost $2 billion
As a quick head check, Qualcomm's (QCOM 1.05%) entire R&D budget (remember, Qualcomm is the No. 1 vendor of modems) is about $5 billion. However, if we take a look at Qualcomm's financials, we can see that the company doesn't even spend all of that on its chips:
Source: Qualcomm
Indeed, QCT -- the chip division -- did $16.7 billion in sales, and generated $3.189 billion in operating income. Now, given that the cost of the equipment and services revenue was $9.82 billion, this implies that Qualcomm's operating expenses here were $3.69 billion. It's tough to imagine that $2 billion of that spending -- in light of all of Qualcomm's CPU, GPU, ISP, etc. IP -- is going to the modem, as important as it is.
Foolish bottom line
It needs to be faced -- leadership in the mobile market is expensive and costs just south of $4 billion in operating expenses to really be in that upper echelon across the board. Only Intel and Qualcomm can afford this level of spending long term outside of the rich-device companies like Samsung and Apple. Given, however, that Samsung's entire semiconductor R&D spend last year was south of $3 billion (including process technology R&D), I would say only Intel and Qualcomm are spending at the levels needed for long-term, sustainable leadership, and that any company that spends less than this will never be first tier.
Intel needs to be first tier in mobile and, as a result, needs to spend at this level. Any "cost cutting" and, in particular, halting development on the critical cellular/connectivity components, and Intel may as well fold up shop and quit mobile altogether. Intel's management team is too smart to do that.
