Moore's Law hasn't slowed innovation. The cost of fab toolsets and the cost of chip design are the culprits. Many cutting-edge tools (etchers, deposition, othe

Moore's Law hasn't slowed innovation. The cost of fab toolsets and the cost of chip design are the culprits. Many cutting-edge tools (etchers, deposition, others) now come with price tags of >$5M. A new 200mm fab can easily run north of $4B. The relentless trend of feature size shrinkage is still happening - some designs are underway using 16nm FinFet technology - but at the cost of increased power consumption and increased cost-per-transistor. I believe it was Nvidia that put a shot across the bow of chip manufacturers by saying they saw no economic advantage in chasing the lastest process node. A clean-sheet-of-paper microprocessor design can easily hit $50M in expenses before the first chip is seen. Much of that expense is sheer manpower & verification processing cycles - a billion transistors' worth of real estate is worthless unless you can run a test suite against it. If you're spending >$50M on a new design, you'd better have a locked-in customer. Projecting that amount of investment for a squishy market like a new consumer electronics widge is the surest way to get your project killed. Running a microprocessor at ever-greater frequencies doesn't help either. First, you're cooking your chip. Frequency = heat. Second, every processor vendor in the world is limited by the processor-memory chokepoint. You can't read data & code from memory into the processor(s) fast enough to keep the processor pipeline full. And the gap is widening.