At Intel's (NASDAQ:INTC) analyst day the company suggested that - if such a deal made sense - it would build mobile chips for competitors. While some investors may be excited by this prospect, such deals are unlikely to make much financial sense. Indeed, on a potential revenue-per-wafer basis (which is the key metric to look at here), building $20-$40 chips for (what would likely be a limited set of) competitors just isn't all that lucrative while, for those same wafers, building Atom processors is much more lucrative.
How much would an Apple deal be worth?
Apple (NASDAQ:AAPL) owns roughly 15% of the applications processor market. According to iSuppli, Apple pays Samsung (OTC:SSNLF.PK) about $19 per A7 chip (although in the industry, Apple would be paying for wafers and the cost would include whatever yield loss is present on the wafer). Given that the gross margins in the semiconductor foundry business tend to run at about 45%, Samsung is making about $8.55 in gross profit per chip. Now, the majority of Apple's products won't have the highest end SoCs - they'll use the older generation A6 and A5 processors that are, presumably, much cheaper to make. Assuming, then a weighted average cost per chip of $15, and assuming about 200 million units per year (iPod, iPad, Apple TV, iPhone), this is a deal worth $3B.
Of course, $3B on the top line is nothing to sneeze at, but at a gross margin profile of just 45%, this translates into $1.35B in incremental gross profit - a couple of percent increase from the company's roughly $32B in gross profit. This deal doesn't really look all that great, does it? These wafers can - and eventually will - be used to build much higher margin, Intel-designed products.
What does Intel need to do with Atom, then?
While Intel is extending into negative gross margin territory with its "contra-revenue" and additional NRE payments in a bid to aggressively gain market share in the tablet market, the company's mobile efforts won't be in the gross margin abyss forever. Indeed, once the company's products are - across the board - at the right design points and once the company's mobile efforts are well established with the major OEMs, a gross margin profile of at least 55% (1000bps higher than what the foundries charge their fabless customers) is well within the realm of possibility.
Further, the company's average revenue per chip - since it is selling the finished product and not wafers to a fabless semiconductor player - will probably be in the $20-$30 range for higher end parts, and in the $10-$15 range for the very low end parts. Assuming, of course, that Intel can do 55% gross margins on its highest end lineup, it would need to sell just half of the number of its own high end chips at $25 apiece to get to the same gross profit. Assuming Intel's die sizes are the same or slightly smaller, that's about twice the revenue per wafer.
More importantly, though, is that once Intel's apps processors are competitive and its modems are competitive (and, of course, once those modems are built at Intel's own factories - this happens in the late 2015/early 2016 timeframe), then it will not only be able to get paid well for its apps processors, but it will also get paid for the sale of cellular chips into the high end. Intel may have plenty of fab space, but as its mobile volumes ramp and as more of its cellular/wireless products get moved into Intel's own factories, utilization should not be an issue - and Intel will get paid both design and foundry margin for each and every chip.
An Apple deal or, frankly, a deal with any of its chip competitors just doesn't make fiscal sense. Given that Intel really does have only a finite amount of wafer capacity, and given that the mobile markets are exceptionally large (and the silicon content that Intel can capture here is quite high), it would be a surprise to see Intel's fabs - short of the company's design teams and its sales teams utterly failing to secure meaningful wins - be filled with any low cost mobile chips from competitors.
Large chips, like FPGAs, network processing units, and large custom ASICs, on the other hand, make plenty of sense for Intel's custom foundry. But that's another discussion for another time. In fact, my next Intel-focused article will be about what chip companies could very well use Intel's leading edge factories in a way that's mutually beneficial to both.
Disclosure: I am long INTC. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.