Some thoughts on the data center basket

NBIS, PSIX, TGEN and the broad AI basket

About a year ago I wrote a piece titled Ai is Dead, Long Live Tech:

AI is dead, long live tech

It was written in mid-July right before the market rolled over into the August crash. One of the core points I made in the piece was no one knew what the 2026 AI numbers would be, and that created some selling pressure from the mid June peak leading into my July post. Then add in the yen carry trade blow up and suddenly all the risks of softer 2026 numbers were almost immediately priced in. All breaks and no gas.

That dip provides some general fuel to tech broadly into Q1 of 2025, but it wasn’t led by AI, with various other pieces of the tech trade taking turns with leadership. That lack of leadership by AI/datacenter was justified with part of my 9th point in my ‘100 thoughts about 2025’ came true with the announcement of DeepSeek.

100 Thoughts on the stock market in 2025

“Was it all worthless?” investors were asking themselves in late January and into the April pullback as datacenter stocks broadly preformed poorly, and we had a massive valuation compression event.

Since end of April, really easter, when I wrote up Nebius Group as a long pitch when it was chilling at $20, the AI trade has been all gas and no breaks. Crushing even the strongest bears beneath the might of the long AI/Datacenter trade.

Now, a little more than a year since that first post, its time to reflect and look forward again, and try to figure out where in the AI chain we can find value, and where some of the pitfalls are laying in wait for us.