This week, Intel and ARM announced that they would be entering a partnership whereby for the first time in years, ARM designs will be manufactured by Intel.
The chips will be for a low-risk client, South Korea’s LG, and it signals a historic rejoining between ARM and Intel since they worked together many, many moons ago. Much fanfare was made following the announcement, but there are a few reasons to believe it’s a shrewd move by Intel.
In 2013, Intel’s marketing department drove home how its Custom Foundry would find the chip giant opening its fabrication plants – where semiconductor chips are produced – to other businesses. But the truth is Intel has been producing custom chips for some companies for years: one such is Hewlett Packard (before the turn of the millennium) and even its predecessor, Compaq. Now it makes chips for others like Altera, but it does so to very specific guidelines, and as more of a contract manufacturing model than as a pure-play foundry like rivals Samsung and TSMC, which have both traditionally been the de facto choices for the production of low power and mobile chips.
“Intel is very picky about which customers it will have, and that’s partly because it’s very picky about its process and its design methodologies,” says Malcolm Penn, principal analyst and founder at Future Horizons. ”If you want to use Intel as a foundry you’ll design it their way, and that way is non-industry standard.”
Traditionally, Intel has blown hot and cold about whether or not it plans to shift to a larger-scale foundry model. However, there are some convincing arguments as to why this could be beneficial, not least undermining Samsung and TSMC – both of which have claimed to have cutting-edge technology, but according to Penn, is nothing of the sort.
“Even though the competitors all say they have a 10nm or 14nm process, they’re not identical, and the Intel process is definitely more advanced by a generation than Samsung or TSMC,” says Penn. “That’s probably allowing large companies like LG to want to get into bed with Intel.
“There’s also no doubt about it, although Samsung and TSMC have closed the gap with Intel over the years, they are still not process development leaders – and to be a leader in the same way Intel is, they’d have to spend huge amounts of money on R&D.
“You can see how if Intel did want to offer a foundry business on a larger scale, to only a select top tier of customers it would be a formidable force. There’s nothing that TSMC or Samsung could do that would come up with a chip of better performance than an Intel-designed chip.”
What’s in it for Intel?
Fabrication plants are highly complex and expensive assets to maintain, and they shouldn’t be running under capacity. Although it’s very unlikely Intel’s are sitting empty, the current level of PC and server demand in the market suggests they might not be as full as Intel would like them to be.
When Intel pushes its upcoming 3D memory products out of the development stage this will change dramatically – but for now, it’s safe bet that there’s capacity going spare.
With this deal, Intel is filling its fabs with a contract from a top-tier client – LG – and gaining practical experience in turning out ultra-low power chips at a very high volume. Not only is that going to provide Intel with a heap of information about low-power chip production, but also provide it with a better understanding for realising its vision of taking the lion’s share of the burgeoning internet of things market – a critical objective for Intel.
Ultimately the deal serves as a warning shot to the pure-play foundries that it has the technology to lead, fills its fabs with production while they’re running low, and provides Intel with the crucial experience needed to tackle the ultra-low power market.
“It’s a good learning curve,” says Penn. “They’ll be building some of this stuff and also gaining huge amounts of data on real-life products that they’re selling.”