- Intel's data-center group benefits heavily from investment in the PC Client Group.
- It is my belief that the data-center group would see an operating margin drop from 47% to 30-35% if it were a standalone entity.
- This leverage illustrates the importance of PC Client Group, and why a valuation based solely on DCG is ill-advised.
It is my belief that Intel's (NASDAQ:INTC) server story benefits heavily on the operating margin side of things from investments made in the PC Client Group. Indeed, it is my belief that Intel's data-center group, which runs at roughly a 47% operating margin rate today, would offer something more along the lines of 30-35% operating margin if it were a standalone entity today. This illustrates why valuing Intel solely on the basis of DCG and excluding the PCCG business is ill-advised.
Meet Intel's Datacenter Group
At Intel's 2013 Investor Meeting, the company showed the following slide to illustrate just how robust a business the company's data-center group is:
While the 2013 actuals came in slightly lower than these numbers ($11.24B in sales, $5.16B in operating income), the point remains that this is an extremely healthy and high margin business. My thesis is that without the PC Client Group serving to amortize a good portion of the R&D and, to a lesser extent, without PC Client Group paying for the leading edge factories (which significantly lowers depreciation per Xeon unit, since Intel's server products typically do not run on leading edge nodes), this operating margin percentage would be much lower.
Running The Numbers
At the 2012 Investor Meeting, Intel showed the following slide to give investors an idea of just how Intel's R&D dollars are typically allocated:
Notice that "technology development" (which is process technology) is "over $2 billion" (but Intel won't tip its hand as to just how much it actually spends). Note also that the "shared processor and graphics" investment is also "over $2 billion". The latter includes all of the Atom/Core CPU cores as well as the Intel's in-house GPU. I argue that Intel's spending on CPUs is higher than that of its graphics processors, although I don't have enough information to give granularity beyond that.
But you see the point, right? There's a lot of sharing/reuse here that helps to improve margins for every division as the costs are amortized among multiple product lines across multiple businesses. So, today the DCG operating margin is ~47% (really more like 46% if you look at the 2013 actuals). What happens to that number when you make Intel a "server-only" company?
Well, quantitatively, I want to do the following to arrive at the right numbers:
- Let's assume that for a server-only company, Intel no longer needs to invest in graphics. This cuts, in the best case, "shared processor and graphics" to ">$1 billion". I believe this is probably grossly understates the investment given that GPGPU may be an important part of future server applications, and even many Intel Xeons (which go into workstations - the E3 variety) have integrated graphics.
- Let's also assume that the "PCCG + DCG" investment of $2 billion is now cut in half - it's just $1 billion.
- Next, assume technology development is roughly flat as Intel would not go fabless here or license outside process technology.
- Further, assume that the software investment is cut to $250 million/year.
- Finally, assume that SG&A (which is currently $8 billion for all of Intel) goes to a mere $1 billion/year.
Running the numbers, and assuming that DCG has gross margin of 80%, this is what I get:
- Revenue: $11.23 billion
- R&D: $2 billion technology development + $1 billion processor development + $1 billion DCG-specific R&D = $4 billion
- SG&A: $1 billion
- Gross Margin: $8.9 billion
- Operating margin: $3.9 billion or 35% of revenue
Now, I believe that the processor spending + graphics could potentially be higher, and I believe that my SG&A estimate, while reasonable, may be a slight over-estimate. However, it's important to note that if we take SG&A to $0 (these chips will not literally "sell themselves"), we still end up with meaningfully lower operating margin dollars than we currently have ($5.16B v.s. $4.9 billion).
In short, I believe that this is a fair justification of the following claim made in my previous piece, "Intel: It's Not That Easy",
In short, without the PC group essentially doing a lot of the heavy lifting for DCG, DCG wouldn't be anywhere near as profitable. More to the point, the success of PC Client Group and its ability to keep utilization high and to fund much of the IP development is the secret sauce to Intel's DCG profitability. If those PC volumes go away, DCG suddenly becomes a lot less attractive of a business.
Discussion and/or fine-tuning of this model is welcome.