3 ms·

Monstrously capital intensive industries are conservative for obvious reasons. Computers were never as capital intensive as a large power plant, and transistors

by api 11y ago

Monstrously capital intensive industries are conservative for obvious reasons. Computers were never as capital intensive as a large power plant, and transistors and ICs made them exponentially less. When a project that costs twenty billion dollars fails companies go bankrupt and politicians lose their careers. Lots of pressure to stick with what you know will work. Also everyone blames the hippies for nuclear stagnation but nobody looks at the fossil fuel industry. The latter has far more money and political influence and is more than happy for nukes to get stuck in the 50s.

jacquesm 11y ago
There isn't much else that is more capital intensive than a modern chip factory, the costs are - roughly - comparable to a nuclear power plant.
- chopin 11y ago
  But this haven't been true in the 60's or early 70's.
- api 11y ago
  They're just as conservative too. They'll go to .22nm from .32nm but they're not going to invest billions up front in a fab to produce a completely untested chip technology based on graphene and optics or something. Not unless someone (like the government) coughs up the money risk-free. Same with nukes. Gradual evolution of the overgrown 50s submarine reactor is the AP1000. To really make fission work and scale would require something like the liquid thorium cycle or accelerator driven reactors or pebble beds... something not an evolution of 50s tech.
  - jacquesm 11y ago
    Indeed. So, now with Moore's law coming to an end there will be an incentive to try 'strange' and off-the-beaten-path technology, it will in a way be the most interesting time for semiconductors in many decades. There is just a chance that we'll hit on something that would not even be considered only 10 years ago that will actually make a huge difference. Before that all you could pretty much set the clock by was a predictable speed increase, now we're going to either get stuck or there will be something revolutionary in computing to move us to the next level. Some of the options: - 3D (heat issues, but biology got that one solved to some extent by using massive parallel low power, extremely slow and clockless architectures) - Quantum Computing - biological computing - exotic semiconductors (for instance: GaAs) - optical computing - Totally new software architectures to take advantage of existing hardware in an un-foreseen way - non Von-Neumann machinery (for instance, computing memory) And a whole raft of even lesser known and researched tech. I'm really very curious what the next decade will bring.
    - api 11y ago
      Why? If end user devices are just thin clients for the cloud and we can always just make data centers bigger, where is the trillion dollar incentive necessary to make the leap? The constraint on tech is usually more about what is economically justifiable rather than what is physically possible. It's 2016 and passenger flight is still subsonic for example.
      - jacquesm 11y ago
        That depends on the amount of computing capacity required for those data centers and if the last years have been any indication the amount of data center capacity is not even close to where we will end up. Even the 'thin clients' will acquire features and more capacity as time goes on, that's just the eternal terminal->personal computer pendulum on one end of it's swing at the moment.
- 7952 11y ago
  Cutting edge chip production is still highly dependant on possessing the required experience within your institution. You may be able to build a FAB from parts, but that doesn't mean you can immediately start mass producing the latest CPU's or memory. And there are other comparably complex project's with similar issues (like designing rocket engines). I think the same applies to reactor building where only a few companies know how to do a particular thing (like build a reactor vessel). Cutting edge development is unavoidably going to be limited to a small group of people who have limited capacity. Merely maintaining that capacity is tricky.
vidarh 11y ago
We're stuck with massively large nuclear plants in part because the only way of making a new plant cost effective given the regulatory regime is to make them big.

by api 11y ago

jacquesm 11y ago

There isn't much else that is more capital intensive than a modern chip factory, the costs are - roughly - comparable to a nuclear power plant.

chopin 11y ago
But this haven't been true in the 60's or early 70's.
api 11y ago
They're just as conservative too. They'll go to .22nm from .32nm but they're not going to invest billions up front in a fab to produce a completely untested chip technology based on graphene and optics or something. Not unless someone (like the government) coughs up the money risk-free. Same with nukes. Gradual evolution of the overgrown 50s submarine reactor is the AP1000. To really make fission work and scale would require something like the liquid thorium cycle or accelerator driven reactors or pebble beds... something not an evolution of 50s tech.
- jacquesm 11y ago
  Indeed. So, now with Moore's law coming to an end there will be an incentive to try 'strange' and off-the-beaten-path technology, it will in a way be the most interesting time for semiconductors in many decades. There is just a chance that we'll hit on something that would not even be considered only 10 years ago that will actually make a huge difference. Before that all you could pretty much set the clock by was a predictable speed increase, now we're going to either get stuck or there will be something revolutionary in computing to move us to the next level. Some of the options: - 3D (heat issues, but biology got that one solved to some extent by using massive parallel low power, extremely slow and clockless architectures) - Quantum Computing - biological computing - exotic semiconductors (for instance: GaAs) - optical computing - Totally new software architectures to take advantage of existing hardware in an un-foreseen way - non Von-Neumann machinery (for instance, computing memory) And a whole raft of even lesser known and researched tech. I'm really very curious what the next decade will bring.
  - api 11y ago
    Why? If end user devices are just thin clients for the cloud and we can always just make data centers bigger, where is the trillion dollar incentive necessary to make the leap? The constraint on tech is usually more about what is economically justifiable rather than what is physically possible. It's 2016 and passenger flight is still subsonic for example.
    - jacquesm 11y ago
      That depends on the amount of computing capacity required for those data centers and if the last years have been any indication the amount of data center capacity is not even close to where we will end up. Even the 'thin clients' will acquire features and more capacity as time goes on, that's just the eternal terminal->personal computer pendulum on one end of it's swing at the moment.
7952 11y ago
Cutting edge chip production is still highly dependant on possessing the required experience within your institution. You may be able to build a FAB from parts, but that doesn't mean you can immediately start mass producing the latest CPU's or memory. And there are other comparably complex project's with similar issues (like designing rocket engines). I think the same applies to reactor building where only a few companies know how to do a particular thing (like build a reactor vessel). Cutting edge development is unavoidably going to be limited to a small group of people who have limited capacity. Merely maintaining that capacity is tricky.

vidarh 11y ago

We're stuck with massively large nuclear plants in part because the only way of making a new plant cost effective given the regulatory regime is to make them big.