The thesis is one step downstream of the chips: AI data centers consume enormous amounts of electricity, and that power has to come from somewhere. If AI compute keeps scaling, so does demand for the fuels and infrastructure that keep the data centers running. This trade expresses the energy side of the AI build-out rather than the silicon side. It complements the AI/semis trade — same secular driver, a different beneficiary. This post walks through it on Legend, one applied example of the thesis-trade framework.
The Thesis
Training and serving large models is power-hungry. New data centers strain local grids, and operators are signing long-term deals for firm, around-the-clock electricity — which points at nuclear (uranium) and natural gas as the baseload fuels, plus the broad energy sector that supplies them. Alongside the fuels, the AI-infrastructure operators that build and rent the compute capacity capture demand directly. If you believe AI capex continues, the energy and infrastructure layer is a way to express it without owning a single chip stock. Legend lists these as perpetual futures across commodities and equities in one self-custody account.
How to Express It on Legend
Go long the AI fuels (commodities and miners)
- URANIUM — the nuclear fuel for baseload power
- URNM — uranium miners, the equity expression of the same theme
- NATGAS — US natural gas, a primary data-center baseload fuel
- TTF — the European natural gas benchmark
- XLE — the broad energy sector
A basket across fuels and the energy sector spreads exposure so a single market's move hurts less.
Add the AI-infrastructure operators
Beyond the fuels, go long the companies building and renting the compute:
- CRWV — CoreWeave, GPU-cloud capacity
- NBIS — Nebius, AI infrastructure
These let you pair the energy-demand view with the operators that consume the power directly.
Relative value: long power, short the broad market
If your view is that AI energy demand outperforms rather than that everything rises, run a relative-value spread: long an energy name and short SP500, sized notional-balanced so broad-market direction largely offsets and the power thesis is what remains.
Start trading on Legend to put any of these on as real positions.
Sizing and Risk
- Decide your max loss before you size. Energy and uranium are volatile; size for the drawdown, not the dream. See how to manage risk.
- Use leverage deliberately. Available leverage is a ceiling. Use isolated margin to cap the loss per leg.
- Watch funding on commodity perps. You pay or receive funding on each side; over a long hold it is a real carry cost.
- Diversify across fuels. Uranium, gas, and the energy sector respond to different drivers, so a basket is genuinely more spread out than a single commodity.
What Could Go Wrong
The AI power trade depends on both the AI build-out and the energy markets cooperating. The main risks:
- Policy. Nuclear and natural gas are heavily shaped by regulation and permitting; a policy shift can move uranium or gas sharply.
- Rate sensitivity. Capital-intensive energy and infrastructure names can be sensitive to interest rates, independent of demand.
- Weather and storage. Natural gas in particular swings on weather and inventory data, which can swamp the AI-demand thesis in the short run.
- Wrong leg in a spread. A long-power / short-SP500 pair loses if the broad market outperforms energy. See how to avoid liquidation before adding leverage.
This article is educational and is not financial advice.