From the book:

Fabless: The Transformation of the Semiconductor Industry-2013

Aart de Geus
Chairman and Co-CEO of Synopsys, Inc.
There are fabulous technology advances underway at just the time
when the explosion of applications is gearing up to bring in BIG money
into the semiconductor techonomic ecosystem. What we’re seeing is the
latent pull of the applications market, riding on the back of the past 50
years of semi technology push.

And the waves of applications are just beginning to hit the market
shore. No matter what field you look at, you notice a substantial uptick in
hunger for silicon: “If you could just make the chip design a little faster/
lower power/smaller, we could do it!” In other words, semi innovation
will be most visible through utilization. It will be a connected world of
“smart everything,” where everything, and possibly every one, will have
chip-driven computational capabilities and an IP address.

Paul McLellan
Semiwiki.com bloggers and author
It is clear that Moore’s Law will continue, in the sense that we know
pretty much how to build chips all the way down to 7 nm. And by the
way, 7 nm doesn’t have anything on it as small as 7 nm, it’s just a name.
But what is a lot less clear is whether the cost per transistor will continue
to drop or whether 28 nm will be the cheapest process not just to date
but forever. As for memories, at some point the capacitor that holds the
data in a DRAM will also stop scaling. 3D chips might turn out to be
another game-changer but currently they are also too expensive.
We are entering an era where the economics that have driven
semiconductor for five decades may be coming to an end. Even if each
process is still a little cheaper per transistor, which Intel reckons is the case
for them, we are not going to have the sort of 10,000X reduction in price
that resulted in our smart-phones having more graphics capability than
million dollar flight simulators of 30 years ago. They will never become
like credit-card calculators, so cheap they can be used as giveaways.
This is something that the general public and the press have not yet
noticed. Everyone just considers that Moore’s Law will continue since
the iPhone 5 is more powerful than the iPhone 4. But that is a very
limited market that isn’t affected by a few dollars extra for an SoC. The
iPhone 5 is not cheaper than the iPhone 4. The PC you used to buy for
thousands of dollars is now just a few hundred. But that trend is not
happening with iPhones.

On the other hand, something may change. Some technical breakthrough
in semiconductor technology or EDA may break the logjam. Carbon
nanotubes? Directed self-assembly? Optical interconnect? Quantum
computing? EUV may even work. Or maybe this time it really is different.

Daniel Nenni
Semiwiki.com blogger and author
As I see it, the biggest challenge for the semiconductor industry moving
forward is economic. I have no doubt that the fabless semiconductor
ecosystem will clear the technical hurdles required to continue down the
process node path. I do however think we are underestimating the effect
of the increasing financial burden of modern semiconductor design and
manufacturing.
The advent of the fabless business model opened the doors to all
comers who had an idea and a modicum of capital. The shear competitive
nature of this paradigm shift brought us innovative products only Gene
Roddenberry could have imagined at palatable costs. Unfortunately, as
our industry matures, consolidation comes knocking and that is when
the doors of innovation start closing again.
A reduction in the work force comes with consolidation and that is a
very difficult trend to reverse. Today less than 5% of the work force is
engineering and science based. How do we attract new talent when college
graduates see the semiconductor industry as old and boring compared
to the likes of Google, Facebook, or Twitter?
Attracting new capital is also a problem. According to the GSA
January 2014 Market Watch Report, in 2013 semiconductor funding
deals decreased 63% from what was raised in 2012. The total number of
fabless semiconductor deals closed during the same period decreased by
30%. And the average value per deal decreased 13%.

If you really want to
know what will finally kill Moore’s law the answer is economics, absolutely.
We as an industry do not do a great job of communicating our amazing
story to the outside world. That is the real motivation behind SemiWiki.
com and this book; to remind everybody that the fabless semiconductor
ecosystem made semiconductors what they are today—a critical part of
modern life.