Hacker News Reader

Comment by nn3 | Hacker News Reader

I'm sure Intel would be happy learn from you how they could avoid investing all that money.

To misquote George Burns, too bad that all the people who know how to build fabs are busy hanging out on HN instead...

The point is that if you have a new chip design as a fabless semiconductor start up (like mine), you can get a chip fabbed on a fairly new process node (28nm) for less than $250k for a small quantity run, and go into production for under $5 million through a pure play fab like Global Foundries, TSMC, Samsung, UMC, SMIC, etc.

Right. Who need power plants when the power is coming out of the wall plug.

See the problem? Someone has to build the fab in the first place so you can use it.

Oh of course there needs to be fabs, but plenty of them exist. All I was saying is that if you want to make a chip, you don't need to take on Intel's capex.

Then you end up like AMD, with designs that have to be cut back and compromised when a third-party fab doesn't manage to do what they expect on a new process.

There's something to be said for tight integration between fab and design when you're trying to go in to uncharted territory.

If you don't need to be at the edge of process technology, you're completely right of course.

Why do you have to be like AMD - why can't you be like Apple, Qualcomm, NVIDIA, or, well, about any successful/large chip maker other than Intel and Samsung?

Regarding the edge of technology - Intel doesn't beat others to a new node by making both chips and fabs, it would beat others even if it was only building fabs. And it's not like TSMC is that much behind Intel. Grandparent who mentioned 28nm could mention 16nm except that there the masks will costs much more but it's still millions, not billions.

Our next shuttle run is on 16nm... Depending on a bunch of different options, it is "only" ~2-4x the cost of a 28nm run (and should be declining in the next year). Doing a full mask set is still extremely expensive though (~$5-10m range)

It's easy to build a cheap fab. You just target 45nm and use the old tried and true tech.

Reply on news.ycombinator.com