So I've been watching the semiconductor space pretty closely, and my biggest position in this sector was the VanEck Semiconductor ETF - it's up 49% last year which honestly felt wild to watch happen. We're already four months into 2026 and it's gained another 12%, so the momentum is still there.

The thing is, everyone's basically betting that AI demand for chips isn't going anywhere. The numbers back it up - global semiconductor sales are projected to hit $975 billion this year, which is 26% growth over 2025. That's pretty healthy. When you've got every major tech company dumping billions into AI infrastructure, and we're probably years away from anything resembling a mature product, there's real fuel for this rally.

But here's where I'm getting a bit cautious. The SMH ETF is trading at 45x trailing earnings right now. Even if we hit that 26% growth target, the forward P/E drops to around 35. That's still pretty stretched when you think about it. And then Nvidia reported solid numbers - beat on revenue, beat on earnings, raised guidance - and the stock still dropped 5% the next day. That's a warning signal if I've ever seen one. The market might have already priced in all the good news, and any disappointment could get ugly fast.

There's also this concentration risk that's hard to ignore. About 30% of the SMH portfolio is just Nvidia and Taiwan Semiconductor. So your whole thesis basically depends on these two companies executing flawlessly. If either stumbles, you're going to feel it.

Do I still think semiconductors are the right play for AI exposure? Yeah, probably. The software story looks way more vulnerable to disruption as AI coding gets better. Chips are harder to replace. But I'm watching valuation pretty carefully at this point. The sector could definitely keep running in 2026, but I'd be lying if I said I wasn't thinking about taking some profits or waiting for a better entry point.