Yesterday we did a quick post on the crazy/not crazy idea that Intel could spin off its fabs into a separate company. A reader asked us how much would that entity be worth, and since we are clearly obsessed with this science fiction topic, we thought we would take a rough stab at putting a number on it.
Looking at it from a strictly mathematical, financial perspective, this looks attractive. (Please note that this is all very rough math using figures from Yahoo Finance.) Intel generated $78 billion in revenue over the past twelve months, with gross profit of $44 billion. That means their cost of goods sold comes in at $34 billion. So a spun-out Intel Foundry would count that $34 billion as revenue, again roughly speaking.
Intel is currently trading at 3x Enterprise Value to Revenue, by contrast Taiwan Semiconductor (TSMC) is trading at 9x EV/Rev. So in theory, Intel Foundry would be worth $309 billion, which is more than Intel’s current enterprise value of $211 billion. This, of course, is too good to be true.
First, not all of Intel’s cost of goods sold goes to their internal fabs. Second, TSMC is a clear leader in the foundry business with all the best customers. By contrast, Intel Foundry would only have one customer to start – the Design side of Intel. Factor in a whole range of other factors like restructuring costs, dis-integration and taxes, we think Intel Foundry would likely end up trading at a significant discount to TSMC. For the sake of simplicity, let’s say Intel Foundry trades at half of TSMC’s multiple, so $154 billion.
Now things get even more complicated. If Intel is worth $211 billion today (in EV), and the Foundry business is worth $154 billion, that means that the rest of Intel is worth $56 billion or 11x EV/EBIT. This is a significant discount to other fabless companies, such as Qualcomm which trades at close to 40x times EV/EBIT. So on its face, this split looks like it make a lot of financial sense.
That being said, there are some considerable problems that would need to be resolved. Factors like who ends up owning the company’s Intellectual Property (IP), how do they divide up the debt and cash on the balance sheet, how much additional corporate overhead would two separate companies require. These are difficult issues to sort out in small companies, let alone the biggest chip company in the US.
In addition, there are significant operational issues which we glossed over when we split out the company’s gross margins. One of the poorly understood financial factors of semiconductor companies is the way they use their gross margins. Calculating gross margins is complex in semis because much of the long-term costs get amortized into gross margins. This is meaningful in fabless companies, but huge in companies like Intel which own their own fabs. Moreover, companies like Intel have operationalized this into their product strategy. The trick is that fab utilization plays a big role in these calcuations and Intel has a large bag of tricks at their disposal.
A simple-ish example can demonstrate this. Many of Intel’s chips are big by semis standards. The production statistics of semis means that larger chips are more likely to have a bug when they come out of the fab. When Intel produces a wafer of its CPUs, this means that some of the chips will have errors in them. But they have designed that into their model. The small number of bug-free chips are binned into premium products because they have more functioning cores. The parts with bugs in them are designed to still function even with those bugs, but they are slightly less capable, and are thus priced lower as a different product. It is the same chip and the same design, but some parts work a little better and thus command a premium. By integrating the design and foundry side of their businesses, Intel can fully capture this, and have an advantage in working this phenomenon into their designs.
There are many more ways that this plays out. Put simply, they give Intel a cost advantage when their fabs are running at high utilization. All of this has been diminished in recent years because their leading-edge fabs are operating well below capacity given their ongoing problems making the transition to 10nm and 7nm. But Intel still has tremendous leverage in their model, and this will become more apparent assuming they can turn the corner on their manufacturing processes.
All of this is a roundabout way of saying that there are considerable dis-synergies involved in splitting the company in two. Simply getting the company in a position to split in two would take several years and billions of dollars of consultants and lawyers. The end result would likely trade at a premium to the current combined company, but the pain it would take to get there would likely mean years of diminished valuation for the company.
How did AMD get around the issue of bifurcating products by number of defects when they spun out their fabs?
The fabless companies can do this too but they have to pay the foundry real dollars as they get the process to work. The internal Intel teams have much closer relationships with their intern fabs which makes the whole process easier.
How did AMD get around the issue of bifurcating product lines by number of defects when they spun out their fabs?