I have this theory, supported by anecdote alone.

In the US there is a "scam premium". The scammier a business, the higher the multiple.

Fabbing is decidedly unscammy. So according to the theory of scam premiums, everything associated with fabbing should trade at a discount.

Well, maybe another way to put it is that it's a business where the costs and value are absolutely clear. And it's a pure physics/technology catchup business. So margins get competed very thin.

Versus what you call "scammy" industries, where people attach emotional value, social value, network effects, irrational likes/dislikes -- there's a lot of margin that can be milked out of stuff like that.

The expensive part isn't getting the labor, it's building and operating the facilities - and then the profit margins aren't even that great.

The existing chip manufacturers are sufficient to supply global demand, creating more competition there is pointless from an economic perspective. It only makes sense from a national security perspective.

And from the perspective of the labor, what good are your highly specialized skills on a million dollar machine if you don't have a million dollar machine? Your only choice is to work for the few companies that own one and take whatever pay they offer you to try and make up for the decade+ of lost income from grad school.

Very specialised work with a low volume of competitors seems a great way to stagnate wages.

It’s probably been undervalued for a long time. Not just fabs , hardware comp used to on par or higher than SW many years ago. Now it’s fallen behind on average. But eventually shock (eg trade embargo) will cause a correction

If you look at the number of chip companies in business starting 30-40 years ago to today, you'll notice that the trend has been one of consolidation.

Honestly that would mean there's ample room for competition considering what's happening especially with Intel, Nvidia, and AMD. Though I imagine the barrier to entry is high.

I think the answer to OP is in the combination then? High barriers to entry means there are fewer competitors for talent - leading to stagnating wages. It might all change with news like the above shaking things up..

>I imagine the barrier to entry is high.

Understatement of the last 29 minutes. Invest a few billion with immense risks? Shut up and take the money I wish I had.

I could imagine a collaboration effort from companies like Google, Intel, Nvidia, and Amazon that would reduce the risk to each individual company and result with big benefits. There's been a lot of talk lately about TSMC not being able to produce enough chips for demand. Sounds like the time to take the risk is better than previously.

(I'd also argue that there are benefits to having a fab lab closer to these companies headquarters)

No money in the world would get you near TSMC in less than 10 years.

Source: China throwing all the money in the world, stealing some IP and still not catching up in 20 years.

Playing catch-up is a lot easier than innovation. I'm not sure I buy the claim that another company couldn't catch up. The reason being because if you took all the employees from TSMC and put them in another company (and you weren't worried about IP claims) what would be different? We've seen several companies make up a lot of ground in a short period of time (AMD as an example). But also, 10 years isn't that long.

Actually the major problem for most fabs (Intel/GloFo/Samsung/etc) is not that they can't make 7nm chips, it's they can't make it cheap enough (in high enough yields) to compete with TSMC. However in this case, China can subsidize SMIC to produce at a loss until they eventually catch up compared to the open market.

> subsidize [] to produce at a loss until they eventually catch up compared to the open market.

Isn't this effectively the Silicon Valley model? Corner the market, control, set price.

Your salary is however conditioned on market and how hard your job is