March 15, 2010 – A year after calling a new deal with TSMC to make some of its Atom processors "an important step in a long-term strategic technology cooperation", that partnership apparently has floundered — and it’s a good example of the complex battles the top chipmaker is trying to fight, according to one analyst.
The deal announced in March 2009 ported the Atom processor CPU cores (including IP, libraries, and design flows) to TSMC, in an attempt to help broaden its appeal as a system-on-chip (SoC) for mobile devices — a move interpreted as a stab at the heart of ARM’s market. TSMC reportedly shifted its MPU lines from facilities in northern Taiwan to southern Taiwan and added backend tools, anticipating the work to contribute 1%-2% of 2010 revenues; some analysts have suggested upwards of $1B in annual business from 2011.
But a story from the New York Times reveals that the collaboration is now "on hiatus", though an Intel exec says it’s not officially ended.
CNet’s Tom Foremski more specifically points out ARM’s advantages in a war with Intel (large design library, low-poer and smaller design attractive for mobile apps, large installed base with smartphones and resultingly legions of developers familiar with its capabilities), though "Intel certainly has the talent and resources to make future Atom designs that are competitive with ARM in terms of power consumption and size."
However, Gary Smith from GarySmith EDA suggests that the Intel/TSMC/Atom deal wasn’t just put on hiatus because of slow customer demand — "it was a bad idea to begin with," and TSMC was the only party who would have gained. "By trying to mimic the ARM model, Intel would have only cannibalized their own revenue stream," he writes in a new research note. Trying to be a fabless vendor (for Atoms & mobile apps, competing against ARM) won’t work, he says; while Intel can compete with ARM technically, "they can’t compete commercially […] they have to compete with ARM’s customers" in various markets, and "trying to take on everyone would led to disaster." Meanwhile, its pedigree as an IDM (for its PC and server business, competing against the likes of IBM) puts them on top only in certain areas where resources can be focused and profits can be made — that means the biggest markets like PCs and servers, "so they definitely are less maneuverable than their competition," he notes.
Ultimately, Intel has to do what it has continued to struggle with — find ways to deeply understand emerging markets, then attack with the same skills it used in dominating PCs. "They are not going to win them all but they certainly need to win the majority of those they target," Smith writes. "ARM will still dominate a majority of the markets," he notes, but Intel needs only "to dominate enough of the largest markets so they can maintain their critical mass as an IDM."