According to Digitimes (via JPMorgan), noted FPGA vendor Altera (NASDAQ: ALTR), which had been loudly championing its shift to Intel's (INTC 7.61%) foundry for its high-end FPGAs, is crawling back to Taiwan Semiconductor (TSM 2.15%). Altera touted that Intel's 14-nanometer process gave it power, performance, and density advantages over products built on TSMC's 16-nanometer process (which is really a 20-nanometer with FinFETs). The real question is whether Intel really did lose Altera, and if so, what the implications are.
Intel says it didn't lose Altera
Upon reading the Digitimes piece, I shot off an email to Intel's Chuck Mulloy asking if Intel did indeed lose the Altera business. His response was a clear "no." It seems likely that, at this stage in the development cycle, Altera would have to spend significant time and effort to make a switch to TSMC's process. Further, with Digitimes claiming in that same piece that TSMC's 20-nanometer yield is at about 50%, this would suggest that TSMC is far from true "volume" production of 20-nanometer parts. To put it into perspective, yield of 50% suggests that half of the chips that get built end up getting tossed. TSMC's 16-nanometer yields must be even worse.
Taking a longer term view
While Intel probably didn't "lose" (in the strictest sense) its Altera business, there is a chance that Altera may be designing future FPGAs based on TSMC's 16-nanometer process. If this is true -- and this is a big "if" -- the implications wouldn't be pretty for Intel. While such a defection would not necessarily reflect on the quality or advantages of Intel's manufacturing technology (Intel is known for process leadership), it would serve to solidify TSMC's claims that there is more to the story than just technology leadership.
Foolish bottom line
Intel has been a laggard of a stock for several years now, wildly under-performing semiconductor peers such as TSMC, and it seems that it just can't get a break on the news-flow/rumors front. While it is very unlikely that Intel has lost its current Altera business, and while TSMC is likely to continue to lag behind Intel in semiconductor technology development, the real question is whether Intel is well-suited to run a foundry business long-term. This will remain to be seen, and what Altera does at the 10-nanometer generation will be a pretty big "tell."