I sit in on supply chain calls daily, none of the existing chip makers have capacity that is not already sold or allocated to other commodities. If they did there would not be shortage. It will take at least a year for this to get back to normal
Why not?I don't see things like semiconductor manufacturing ever being in house to car companies. Or, at least, until Amazon builds everything in our lives.
Because it is disturbingly expensive and difficult to keep pace in the semiconductor manufacturing business. Look at the history of AMD and IBM to understand why companies have worked so hard to ditch the manufacturing side of the business. With the incredible investment needed, you need some serious volume to make that money back. It's why we are down to a small handful of worldwide fabs capable of producing cutting edge semiconductors.
You start out making a chip by growing a big crystal of really pure silicon and separating it into thin slices. It costs a certain amount of money to make each slice. In general, the newer the process, the more "stuff" you can squeeze into each slice and the more money you make off each wafer. Given all that, as a manufacturer, would you rather make cheap, lower density chips, or expensive, higher density chips? Deprioritization basically means that the low-profit stuff gets moved to the end of the line.
Off topic, but what did you do? During my time in semiconductors I was in a support lab. Doing testing and data collection for development, failure analysis, etc. Fortunately I didn't need to spend a lot of time in the fab.
That definitely makes sense. And I can't counter your industry experience. However I have just two points.
That will only work until it doesn't. The carmakers are really caught up in old-fashioned "we're going to test this until it's perfect and never change it" design patterns that just don't really work for modern software/hardware solutions. (If they actually tested it until it worked I'd accept it, but they routinely release stuff that doesn't work well, and all the layers of bureaucracy simply make it unlikely that a customer will ever see a fixed version.) You can keep doing that as long as all of your competitors do the same thing, but I doubt that competitive pressures and new players in the market will let them continue to keep doing things the same old way because customers will eventually stop accepting this.
I'm still in the industry, I have worked mainly in product qualification and test development, and now I'm in a product engineering role. I work with a lot of test data, driving failure analysis and fab feedback. We find a lot of manufacturing related issues.
There are certainly arguments to be made for integrating chip manufacturing. If they can work a deal with a fab to buy old fabs or equipment for cheap, then they might be able to make it viable. The problem is that the cost of those old technologies starts to rise at some point when you need to maintain the facilities. And then at some point it will become obsolete, the chips they use today will not be the same chips they need in 10 years (due to safety requirements most likely), and so will their facilities be able to make the new chips they need? If not, what do you do now after you've invested all of this time and money into your chip manufacturing?
I certainly appreciate the insight.
You gotta have talent to do the work. A skilled Phd will cost you 200 grand at the minimum. I know 10-100 million in salaries might sound like chicken feed compared to a billion, but it all adds up.
All completely valid points, and certainly in the long run it seems at least partial vertical integration of the chip manufacturing would likely be a prudent decision. The problem is getting a CEO willing to stake their reputation on implementing a very expensive plan. Once you've made that investment, it's a hard ship to steer without risking a huge loss.
For some perspective on this, I once worked for a division that provided the base library (along with a set of industry leading, custom timing rules) for chip design. This library was used by a number of different design teams. But the actual division I worked for struggled to earn a profit. We generated a lot of "internal" revenue, but it didn't equate directly to $$ in the bank. There was never a time that we didn't have to justify our existence. Budgets were repeatedly slashed because we were a cost center. All expense, no revenue (in the eyes of the bean counters anyway). It made for a very stressful environment.
