Although the hoopla over silicon photonics (SP) has not been as pronounced at IBM as at Intel, including consistent involvement by technologists at the highest corporate layer of the latter, the former was heavily promoting the development of these chips for a long time as well. Each company had been spending R&D funds on SP since 2000. There is ample evidence that by the autumn of 2014, IBM decided to tone down any excessive rhetoric on the technology through changes in management and by shifting some of the public relations focus on leading-edge development away from silicon-based solutions.
fibeReality noted in October of last year that IBM’s announcement of $3 billion in spending on chip development placed a good amount of emphasis on non-silicon alternatives. However, it is now our impression that the money mentioned in the July, 2014 press release was not really a new investment, but had already been earmarked to various programs. Interestingly, the vast majority of that cash appears to be targeted toward the much more conservative types of development like 5nm or 3nm CMOS nodes for the standard, electronics-based transistors.
It is hardly surprising that the actual expenditures are for more practical purposes. While from a PR perspective it may be appealing to discuss the use of engineering talent in dealing with entirely new materials and processes (IBM mentioned carbon nanotubes, III-V semiconductor compounds, low-power transistors and grapheme as well as both neuromorphic and quantum computing) to potentially avoid the hurdles in producing silicon transistors lower than 7nm, the reality is that the most desirable outcome is to steer clear of the unknown pitfalls associated with entirely novel techniques — and first strive to milk CMOS as far into the future as possible.
Undoubtedly, IBM’s most recent public statement was fairly bullish on its claim that its SP demo will enable the production of 100G transceivers. Nevertheless, the overall tone of additional comments in news articles made by spokespeople, including the current manager of IBM Research’s Silicon Photonics Group, were more muted in nature compared to those in the past. For example, there did not seem to be any rhetoric about the firm offering a far superior solution than other companies.
Meanwhile, one has to wonder whether it was just convenient and natural for large conglomerates like IBM and Intel to find scapegoats for the lack of adequate progress on SP. These scientists were willing to work countless hours in the laboratory to try to overcome enormous technological challenges that have been discussed since 1969.
Of course, for Intel itself, the stakes were even higher in that it chose to prematurely commercialize its SP devices. In our opinion, the executives at the top were apparently unwilling to accept any responsibility — looking to put the blame elsewhere on overestimating headway and on being disingenuous regarding complications.
[written by Mark Lutkowitz]