The Chip Insider®– The R&D Crisis redux

The problem of rising R&D costs has reemerged. We have, as an industry, solved this before. Some 20 years ago I was warning that the rising R&D costs would bankrupt the semiconductor industry unless we did something about it. In “The R&D Crisis,” I wrote: “Chip sales have been growing at around 6% a year since 1995. But R&D is growing at a 12% CAGR.” Then predicted that if nothing was done, “By 2020, RD&E spending rate would reach 40%. Clearly, this is unaffordable.” We were able to solve it through collaboration and consolidation by addressing it directly.

We could be at a similar moment again, as early process R&D for 7A starts. Today’s difference is that the opportunity for consolidation is minimal. TSMC is really the last stronghold of logic… Samsung is … Intel … IBM… China’s government is subsidizing its R&D cost, so … But China is mostly irrelevant, since >90% of advanced wafer production is bought by North American semiconductor design customers …

Nodal R&D costs needed to get to first product tape-out have been growing at close to 30% per node… If it continues to trend, the 7A process R&D cost will come just under the $10B high-bar… Even if we are able to slow the growth, I estimate it will still pass $8B. And this is just for logic.

R&D Crisis Solution Paths: Not solving it is unacceptable, as it means giving up the promised golden era of the AI age. There are three solution paths to the R&D Crisis: Consolidation, Collaboration, and Subsidization… We know what the problems are, as well as many of the solutions. All we have to do now is execute, which is often the hardest part.

"History never repeats itself but it often rhymes.” — as Mark Twain is reputed to have said