With slowing revenues from the PC sector, slow progress in the mobile market, and excess capacity on its industry-leading production lines, Intel is set to move into the foundry business in a big way – possibly by stealing Apple’s mobile chip business away from Samsung.
Over the last 12 months, Intel has slowly become a foundry – a company that lets other chip makers use its semiconductor fabrication plants (fabs). Intel began its foundry dabbling last year, when it started producing chips for FPGA makers Achronix and Tabula, which are both privately held and rather small.
Now, Intel has announced that it will produce chips for Altera: A big, publicly-listed company, and one of the largest FPGA and ASIC makers in the world. It’s safe to say that Intel is now a bona fide foundry. By moving into the foundry business, Intel will compete against TSMC, which primarily produces chips for Qualcomm, AMD, Nvidia, and Samsung, which primarily produces Apple’s “A” series of iDevice ARM SoCs.
Which brings us neatly onto Apple. For a long, long time there has been a persistent rumour that Apple and Intel are “in talks.” The first set of talks, as we now know, turned out to be Apple’s switch from the PowerPC architecture to x86 in 2005. Now, of course, the focus has switched to the iPhone and iPad, which use ARM SoCs manufactured by Samsung.
The question is, would Intel simply take on the foundry work and produce ARM SoCs for Apple – or would it try to convince Apple to switch from ARM to x86? The Atom SoCs (Medfield, Clover Trail) aren’t quite there yet, but next-gen parts (Merrifield, Bay Trail) due at the end of 2013 will be serious contenders for the mobile processing crown.
Moving from the fanciful realm of speculation back to the land of hard facts, Intel’s custom foundry vice president told Reuters last week that his group was “ready to take on a potential large, unidentified mobile customer.” Intel also announced that its capital expenditure (capex) budget for 2013 would be $13 billion (£8.6 billion) – a significant bump from previous years. Such a huge figure – $2 billion (£1.3 billion) more than last year – coupled with the slowing PC sector (Intel’s main revenue stream) has caused some trepidation on Wall Street, which was expecting a capex of around $10 billion (£6.6 billion). With Intel already idling some of its fabs due to excess capacity, what is it planning to do with that $13 billion?
According to Reuters, Intel’s foundry shift will coincide with the arrival of a new CEO, who will replace the retiring Paul Otellini. With new CEOs often come big decisions, such as taking on considerably more foundry work.
As to why Intel would move into the foundry business, the simple reason is money. Unless Intel can break into the mobile market in a big way, the only real avenue for continued revenue growth is increasing its foundry business. The problem with becoming a foundry, though, is that it gives Intel’s competitors (Qualcomm, AMD, Samsung) access to the very same advanced CMOS processes that have allowed Intel to gain such a dominating lead in the PC market. The profit margins for foundry work are lower, too: Whereas Intel’s gross margins are around 60 per cent, TSMC’s gross margins are around 48 per cent.
According to one analyst, if Intel did manage to bring on Apple as a foundry client, the production of iPad and iPhone SoCs would net Intel an additional $4.2 billion (£2.8 billion) in revenue. $13 billion (£8.6 billion) of capital expenditure doesn’t seem quite so crazy when you look at it like that. At least as far as the industry analysts are concerned, moving into the foundry business is the only real route for Intel. With the value of its stock dropping 19 per cent over the past year, Wall Street would seem to agree.
I think we should probably wait and see how Merrifield and Bay Trail do, though, before jumping to any conclusions about Intel’s future as a foundry.