Broadcom's AI Chip Story Might Be Telling Us Something We Don't Want to Hear

I should know the semiconductor supply chain better than I do, tbh. It's not my core beat. But I've talked to enough robotics founders to know that chip costs are one of the biggest constraints on scaling production. When Figure or Apptronik or any of these humanoid startups talk about getting to mass production, the conversation always comes back to silicon. Can they get enough? Can they afford it? Will the next generation be cheaper or more expensive?

Shalett's broader point was that the entire market has basically become one trade. Everything is correlated to AI optimism, which is correlated to semiconductor performance, which is correlated to a handful of companies like Broadcom and Nvidia. That's not healthy. It means if the chip story wobbles, everything wobbles.

Now, I'm not saying Tan is wrong about demand. The guy runs Broadcom. He knows his order book. When he talks about AI scaling and revenue outlook, he's not making it up. The question is whether the demand at current prices will persist once the initial buildout phase ends, or whether we're looking at a classic boom that overshoots.

Honestly, I don't know the answer. Anyone who claims certainty here is either lying or selling something.

What I do know is that the robotics industry is planning as if chip abundance is coming. The roadmaps I've seen assume falling costs, improving performance, wider availability. If those assumptions are wrong, if we're actually in a pricing power bubble rather than a productivity revolution, a lot of business plans are going to need rewriting.

There's also a timing question that remains unclear. Even if Shalett is right about the underlying dynamics, being early on a market call is the same as being wrong. The AI buildout could have years left to run. Hyperscalers are still spending aggressively. The demand signals are still strong.

But "warning flag" is the phrase that sticks with me. Not "sell everything," not "the sky is falling." Just a warning. A suggestion that maybe the story is more complicated than it looks.

I talked to a robotics hardware engineer last month who put it differently. She said the chip situation feels like 2021 real estate: everyone knows the prices are kind of insane, but everyone also thinks they can time the exit. The problem is timing exits is really hard, and the people who get hurt are usually the ones who can least afford it.

For humanoid companies specifically, this creates a weird strategic tension. Do you lock in chip supply now at high prices, guaranteeing your costs but also guaranteeing high costs? Or do you bet on prices falling and risk getting caught without supply if demand stays hot?

I don't think there's a right answer. I think it depends on your runway, your timeline, your risk tolerance. But I do think anyone in this space should be paying attention to what people like Shalett are saying, even if you ultimately disagree.

The semiconductor story and the robotics story are the same story now. That's either very good or very dangerous, depending on which parts of the narrative turn out to be true.

Aisha Patel · 16 hours ago · 6 min