Bunnie Huang goes shopping in Shenzhen.
He makes some very good points about how the end of Moore’s law has come for processor speed. The end of shrinking the gate size is coming shortly as well. Not necessarily because the physics don’t work but because the machines to make the transistors will have gotten prohibitively expensive. It seems that there was another exponential in Moore’s law.
It’s the cost to produce the machines that make the chips. As the scale and cost of each individual transistor head toward zero the size and complexity of the optics to get to the smaller size elements head toward infinity. It’s only a question of which curve crashes first. There’s a limit to how much a machine can cost before it become impossible to earn its cost back no matter how profitable the product is or how hard you run the machine. In the case of Moore’s law, it’s the cost of a number of machines that are getting truly expensive.
Here’s a video about Integrated Circuits in the 1960’s
And a more recent video.
As I point out here Silicon Valley has always relied on the Norwalk Valley.
From the 1970’s to now.
And even into the future.
There’s more here.
A Perkin Elmer photo gallery.
William Seigle goes over the history of lithography.
What’s interesting over the video interviews with the engineers is the same pattern happening over and over. Increased aversion as the company grows and a fatal misstep when technology is pushing. The innovation drive seems to get lost when the money starts rolling in and the money people forget how that money happened in the first place. Followed by misteps and mistakes that cause the enterprise to lose it’s edge and eventually go down the tubes to closure or purchase.
As the foundries get more expensive that cost has to be justified. As the element size goes down the investment goes up, way up.
I’ve seen it happen over, sometimes from the inside. That may be the secret of Shenzhen’s long term success. Nobody has a big chunk of the pie. That means that nobody is to deep into the risk. Outside the foundries and Foxconn(Apple) and other large manufacturing, most of what happens in Shenzhen is small outfits combining and sharing bits and pieces from other small outfits. That means that the time to product is very short, as we saw with the sudden emergence of hoverboards just before Christmas. Of course this does lead to problems as the battery issues for hoverboards demonstrated.
Once we admit to ourselves that the drive for ever smaller transistors has reached it’s end maybe that will be the signal to explore new innovations that have been pushed aside by the drive to push Moore’s law to it’s limits. Maybe somebody will explore 3D architectures, embedded small packaging, custom packages with plug and play chip elements, or something that nobody’s thought of yet. The end of Moore’s law doesn’t mean the end of electronics any more than the end of tubes did. That end just means the opportunities for exciting new beginnings.