A lot of folks seem to think that Intel (INTC) should simply give up its mobile ambitions and become a foundry. While the economics of that don't really seem to work, there's another problem here. Being a foundry isn't as easy as simply handing over a chip design and telling a company to hit the "on" button to build the chips -- it's far more complicated and something that Taiwan Semiconductor (TSM) has far more experience at than Intel currently does.
Intel is trying to build experience
Remember how Intel and its foundry partner, Altera (ALTR) , have been making a big deal about their partnership? Well, that partnership is a big deal, but not in the way you might think. The actual revenue contribution from Altera to Intel from this deal is probably minimal. How do we know this? Well, look at Altera's cost of goods sold for the full year of 2013 -- it came in at $546 million.
Since we know that Intel is only going to be building the very bleeding-edge high-end chips for Altera (the mid-range and below will still be built at TSMC), let's say that 10% of Altera's 2016 COGS goes to Intel. This would translate into a whopping $50 million to $60 million in revenue for Intel, which does well north of $50 billion per year in sales annually.
It's clear that Intel isn't building chips for Altera to dramatically boost its top and bottom lines, but instead to gain experience building products outside its traditional comfort zone. Remember, foundries like TSMC are used to building a wide variety of products and working with a broad number of design teams with various skill levels and requirements.
Will Intel be a more aggressive foundry in the future?
Intel's best bet as a foundry is to keep it limited to customers that don't compete with Intel's core products. This means that building mobile apps processors that go into phones and tablets from the likes of Qualcomm isn't going to happen (especially as mobile devices encroach on Intel's PC cash cow). However, building, say, high-performance network chips for a company like Broadcom or large custom ASICs for a company like Cisco would make perfect sense.
Add up enough of these customers, and Intel could have a respectable, although probably not Earth-shattering, foundry business. It allows Intel to drive utilization of its factories even higher (which improves gross margins across the board), adds a non-trivial amount of revenue and profit, and on top of all of that probably provides valuable input to Intel's manufacturing and process teams that Intel's internal teams may not have the perspective to provide.
Intel's future isn't as a foundry -- TSMC and apparently Samsung seem to be on top of the high-volume foundry side of things -- but this could make a pretty decent side business that could ultimately drive utilization and provide valuable knowledge to Intel's teams. That said, if you want to be really pessimistic, this new knowledge ultimately serves as a plan B for Intel if its mobile efforts never take off and it needs to end up building chips for its mobile competitors.