Back at the other place @Cmaier and I had a discussion about the reportedly exorbitant cost of developing a new chip on leading edge nodes. Neither of us really believed the numbers, but here are some huge numbers from Marvell. I can’t remember if this is in that ball park but they are large. The caveat is important though as it is for development completely from scratch which no one does and thus vastly inflates these numbers compared to the actual costs. Still very expensive of course.
Cost to develop 2nm chips
- Thread starter dada_dave
- Start date
- Posts
- 5,350
- Reaction score
- 8,558
I saw that, and look at the “software” number. I was in charge of electronic design automation at AMD, and there’s just no way. I don’t have any idea what they could be including in that number.
- Posts
- 3,702
- Reaction score
- 9,010
- Main Camera
- iPhone
Wow! That’s breathtaking. I see those numbers and just have to smile and shake my head. Even if they were 10x too high.
Obviously the days of 4-5 guys getting together and successfully designing/producing/marketing/selling a handful of different specialized chips have been long gone. Glad I was able to experience that in the past, and compete with large “real” semiconductor companies we really had no business competing with. Clearly that’s a no way today.
Obviously the days of 4-5 guys getting together and successfully designing/producing/marketing/selling a handful of different specialized chips have been long gone. Glad I was able to experience that in the past, and compete with large “real” semiconductor companies we really had no business competing with. Clearly that’s a no way today.
That’s why you get bought out by a much larger company like say Qualcomm to fund your actual development.
- Posts
- 3,702
- Reaction score
- 9,010
- Main Camera
- iPhone
That’s why you get bought out by a much larger company like say Qualcomm to fund your actual development.
And that's what eventually happened, but by a different large (30-40k employees) US semiconductor company wanting to get into what we had developed. It was a large change in culture, with the original five leaving after 4-5 years.
Speaking of Qualcomm... I think that would have been a much better fit as we focused on communications-oriented high-speed (RF to baseband) full-custom CMOS digital signal processing tech that would have complimented some of their products.
