Monday morning's announcement that Intel will outsource the production of some of its Atom processors to TSMC is roughly the semiconductor equivalent of the recent cancellation of the newspaper editors' national convention, or of the closing of major dailies like the Rocky Mountain News.
The late winter storm now pounding the northeast US isn't the only freak weather event currently in progress. Hell itself has frozen over, and the proof is that, for the first time in history, Intel is partially outsourcing the fabrication of one of its CPUs to a foundry.
Intel and TSMC have jointly announced the start of a long-term, "strategic" collaboration in which Atom IP will be ported to TSMC's manufacturing process, with the Taiwanese foundry fabricating Atoms for some customers.
"It is a long-term collaboration. It is not about capacity. It's a strategic relationship that allows us to expand the growth opportunities for both companies," said an Intel spokesperson on a call announcing the venture. "This relationship between us and TSMC is strategic in nature; it's the first time we've ported a processor externally outside of an Intel process."
Both parties are being deliberately vague on the timing and the process node at which this will start, stressing that the current announcement is just about the "memorandum of understanding" (MOU) that they've signed. (The press release is similarly vague.) News of specific plans and products will follow much later, once the details have been worked out.
Intel's Sean Maloney set out to give the company's official explanation for the move by touting Intel's recent announcements of investments in 32nm and stressing that Intel still sees itself as a manufacturing company. He then jumped to the non-sequitur that the TSMC agreement will allow Intel to "attack new markets" and offer "post-recession products."
In response to questions, Maloney stressed that the existing Atom roadmap is unchanged, and that the TSMC relationship was not about replacing Intel's existing fab capacity, but about new markets. "We're doing it because it enables us to get access to a market that we can't do alone," Maloney said. "This doesn't replace what we're doing already, or what they're doing already. It's additive to what we're currently doing."
Intel totally failed to explain why its massive new investments in 32nm, not to mention its current 45nm plants and substantial legacy fab capacity, aren't just as amenable to "attacking new markets" as outsourced fab capacity. But while Intel can't possibly admit this, the reason is as clear as day: Atom is just too cheap to support Intel's legacy business model, which is about funding expensive new fabs on the backs of high-margin sales of existing, high-performance processors.
Source: ars technica