Globalfoundries has also done what may seem illogical and said it intends to invent the fab-lite foundry. That is the implication of dividing Globalfoundries' business between a foundry part and an ASIC development business that is free to go other foundries at 7nm and beyond. Is it the case that the company formed by AMD's desire to exit capital intensive chip manufacturing now itself wishes to get out of capital-intensive chip manufacturing? What we do not yet know is how significant a business ASIC development will be as a proportion of the whole. It seems unusual and may receive double-takes from potential customers before its receives their business.
if we also step back and look at this radical rethink we must also question whether it the beginning of end of Globalfoundries? Or the end of the beginning for a company that could yet prosper?
Undoubtedly what we are seeing is a familiar process of re-alignment imposed on a chip manufacturing company because of the exponentially increasing cost of R&D at the leading edge – and certainly below 10nm – and the exceptional success of Taiwan Semiconductor Manufacturing Co. (Hsinchu, Taiwan) at meeting demand for miniaturization in a timely manner.
It is also ironic that an inability to keep AMD as a client at 7nm could have made Globalfoundries' roadmap to 7nm and beyond unviable. The word in the first half of this year was that AMD would split its 7nm business between TSMC and Globalfoundries. But that talk of a split may have just been a fig leaf to spare Globalfoundries' blushes, with the lion's share of the orders going to TSMC. In any case it now seems that AMD will now get all its 7nm FinFET silicon made by TSMC.
The only other leading-edge FinFET manufacturers left in the game are Intel – arguably also failing to compete with TSMC – and Samsung, which is cross-subsidized by its mobile phone, television and broader engineering businesses.
Globalfoundries has said it will put its 7nm R&D on hold indefinitely and re-assign leading technologists to work on finding differentiating options at 14nm/12nm and on FDSOI. It is effectively throwing in the leading-edge towel. It can never resurrect that 7nm R&D. And on this occasion, it is not even asking if it can license a process from Samsung to stay in the game, something that it did with the 14nm FinFET process.
Next: Going different