AI stocks
AI stocks worth thinking about.
Wellow turned a belief about AI infrastructure into a structured thesis. Below: the cleanest beneficiaries, the second-order plays, the risks that would break it, and the live data to watch.
The thesis backing this page is being refreshed. Try generating one yourself with the topic on your mind.
Frequently asked
- What are the best AI stocks to think about right now?
- The cleanest exposure runs through the picks-and-shovels of the AI build-out: GPU vendors, hyperscaler infrastructure, networking and thermal-management plays. Direct beneficiaries scale with model training capex; indirect beneficiaries scale with data-center power and cooling demand. The thesis below ranks them by exposure type and purity.
- Is the AI trade still investable in late 2026?
- It depends on whether you're betting on the model labs or the infrastructure they consume. Model-lab valuations have ridden a steeper repricing than the underlying compute supply. Infrastructure exposure has more durable economics because every model run, regardless of which lab wins, requires GPUs, power, and networking.
- Picks-and-shovels vs model labs — which is the safer AI exposure?
- Picks-and-shovels (semis, networking, power) tend to capture the spend regardless of which model wins. Model labs concentrate the upside but also the downside — a single architectural shift or commoditization wave can compress margins fast. Most diversified AI exposure leans on the infrastructure side.
- What would break the AI infrastructure thesis?
- A meaningful slowdown in hyperscaler capex, a step-change in model efficiency that reduces compute-per-token by 5–10x, or a regulatory move that throttles data-center power buildout. The full thesis below names the specific risks and what to watch.