According to a report from Digitimes, a leading Taiwanese news outfit, Intel (INTC 1.77%) is beginning to place orders for its next-generation 28-nanometer cellular baseband design. This chip, known as XMM 7260, is the company's second-generation LTE modem supporting LTE-Advanced, category 6 transfer speeds (i.e. 300Mbits/sec), and both TD-SCDMA and TD-LTE modes, giving it much broader appeal than the firm's first LTE attempt, the XMM 7160.
That's not the interesting part
While the speeds, feeds, and features of the company's next-generation cellular baseband parts are certainly interesting, what's more interesting is where Intel is actually having these chips built. Intel isn't building these chips within its own bleeding edge factories, but rather outsourcing production to Global Foundries.
Intel's next-generation modems will be built on Global Foundries' PolySiON (i.e. lower performing than high-K metal gate) 28-nanometer process (which should be rather cheap by now). This is the case because Intel has yet to port over its modem IP (which was likely in development at Infineon, the company that Intel acquired for modem development, well before Intel acquired it) to its own, more sophisticated manufacturing process.
The implications
Interestingly, at Intel's investor meeting, the company made it clear that it would have an LTE-Advanced modem with category 6 speed support shipping to customers during the first half of 2014. Qualcomm (QCOM -0.20%), on the other hand, will be sampling its first category 6 modems during the first half of 2014 for broader availability at some point during the second half of the year.
While Qualcomm is no doubt a modem juggernaut, it's interesting to note that there is an ironic twist here in favor of Intel. Qualcomm's next-generation modem is built on Taiwan Semiconductor's expensive, bleeding edge 20-nanometer manufacturing technology. This brings performance and power improvements, but at the same time, these newer nodes are typically more expensive than mature, fully depreciated nodes.
This means that Intel may have a significant time-to-market and cost advantage by virtue of the fact that it built its modem with a similar feature set to Qualcomm's latest on a cheaper, well-known process node. It's in Qualcomm's long-term strategic interest to move modems to new nodes as quickly as possible and to integrate those modems into apps processors, but this could be a short-term advantage for Intel.
Foolish bottom line
Form an even longer-term perspective, all of these foundries will lose Intel's baseband business as these products are moved in-house (to, of course, get performance, power, cost, and integration advantages), but in the shorter term, Intel needs to build these parts externally. Given that most of Intel's mobile volume through 2015 will likely be built externally (either at Global Foundries or, in the case of SoFIA, at Taiwan Semiconductors), this is an interesting and deep irony for the company with the world's most advanced manufacturing technology. However, time-to-market reality means that Intel needed to bite the bullet and fork over the foundry margin, at least for the next few years.
