ARMing Intel

The industry has kept a close eye on Intel’s foundry business. The question is whether Intel will merely dabble in the business or become a major player.


For some time, the industry has kept a close eye on Intel’s fledging foundry business. The question is whether Intel will merely dabble in the foundry business or become a major player. The answer? It’s still too early to tell.

Not long ago, Intel entered the foundry business and announced a smattering of small and niche-oriented customers, such as Achronix, Netronome and Tabula.  Microsemi is also a customer. For these foundry customers, Intel will offer its 22nm finFET technology. All told, Intel’s move into the foundry business is part of an ongoing diversification strategy.

Then, the industry really took notice when Intel recently announced a foundry deal with FPGA giant Altera. Until then, Altera’s sole foundry was Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC). But apparently, Altera wanted finFET technology sooner than TSMC could deliver, and so, the FPGA house turned to Intel. Under the terms, Altera will gain access to Intel’s upcoming 14nm and 10nm finFET technology, giving the FPGA vendor a lead in process technology over rival Xilinx.

So what’s next for Intel and its foundry business? Unlike before, Intel is now willing to talk to foundry customers that use ARM-based chips, said Nathan Brookwood a research fellow at Insight 64. ARM, of course, competes against Intel in several processor segments. “I think (Intel has) changed their position,” said Brookwood. “They will do ARM-based products.”

The first sign of the new strategy occurred in recent weeks, when Altera said its ARM-based Stratix 10 FPGA would be made on Intel’s 14nm finFET process.  But long term, Intel would also like to secure an ARM-based customer with larger volumes, such as Apple or Qualcomm.

In fact, for some time, Intel has been wooing Apple as a foundry customer. Until recently, Samsung was Apple’s sole foundry partner. But amid a bitter legal battle with Samsung, Apple has recently switched foundry vendors and moved to TSMC. It appears that TSMC will make Apple’s 20nm, ARM-based application processors.

Now, according to industry sources, Apple’s finFET foundry business is up for grabs. TSMC appears to be in the driver’s seat. GlobalFoundries, Intel, Samsung and UMC are also vying for the business. On the surface, Intel appears to have an edge over its rivals. In finFET production, Intel has a two-year or more lead over the rest of the industry. On the other hand, Intel still has an unproven track record in the foundry business.

And in fact, Intel is just beginning to understand the subtle nuances in the arena. It’s more than just making and selling leading-edge wafers. Leading-edge technology is important, but a foundry is really involved in the service business. Price, delivery and flexibility are just a few of the key components in the foundry business. Dealing with multiple customers with different needs can be a headache. For example, the holiday season is typically slow, but foundries must quickly react if customers require sudden and so-called “hot orders” during the period. Cancelled orders are also common.

IP protection and access to capacity are also important. In some cases, the IDM foundries tend to offer plenty of capacity during the slow periods, but they will give their own products priority in the fab during the boom cycles.

Even if Intel begins to understand, and implement, many of these components required for the foundry business, the company will never overcome one obstacle. To one degree or another, the company competes with many large potential foundry customers, such as AMD, Apple, Broadcom, Nvidia, Qualcomm and others.

On that one point alone, Intel may have trouble attracting these customers. Most would rather deal with the pure-play foundries, which do not compete with their customers. But to keep customers happy, the pure-play foundries must continue to deliver on a timely basis. This is especially true with their upcoming finFET process launches.

But if the pure-play foundries are unable to deliver and are substantially late to the market with their finFETs, then the tide could turn. Foundry customers may have little or no choice but to consider Intel–even if the chip giant doesn’t fully understand the intricacies of the foundry business. And then, look for a brave new foundry world.

Leave a Reply

(Note: This name will be displayed publicly)