~75–78% · $180K liquidity
Will US data-center power demand set a record in 2026?
The grid-strain story — near base case
AI & Tech Prediction Markets
The AI race is now a power race. Do data centers set records, do hyperscalers blow past $400B in capex, does Big Tech go nuclear? Prediction markets price each AI-infrastructure milestone as a binary contract on Polymarket and Kalshi. Mantis shows the sharpest cross-venue line in one search.
Live cross-venue odds for power, capex, and facilities. Probability ranges reflect the cross-venue spread as of June 2026 — click any market for real-time quotes.
~75–78% · $180K liquidity
The grid-strain story — near base case
~62–65% · $200K liquidity
Nuclear-for-AI is now a real procurement trend
~47–50% · $160K liquidity
MSFT + GOOG + AMZN + META combined
~37–40% · $140K liquidity
Gigawatt-scale single campus
Electricity is the binding constraint. Nuclear restarts, SMRs, and grid-capacity deals are the catalysts — and the reason power markets dominate this hub.
Microsoft, Google, Amazon, and Meta are spending at unprecedented scale. The $400B market is a direct read on whether the buildout keeps accelerating.
Single sites are crossing into gigawatt territory — a scale that reshapes regional grids. The 1GW market tracks that physical frontier.
AI
The chips inside the data centers — $3T cap & earnings
AI
What the compute is ultimately deployed to run
AI
OpenAI, quantum, robotics and more — full AI prediction market index
As of mid-2026, prediction markets see US data-center power demand setting a record at roughly 75–78% on Polymarket and a Big Tech nuclear power deal near 62–65% — both treated as near base case. Combined hyperscaler capex topping $400B sits around 47–50%, and a gigawatt-scale single campus coming online near 37–40%. Mantis shows the live cross-venue spread.
Training and serving frontier models is electricity-intensive, and data-center demand is now straining grids in several US regions. Power availability — not chips alone — has become the gating factor for AI buildout, which is why nuclear deals (including restarts and SMRs) and grid capacity are the most-watched data-center markets.
The NVIDIA hub is about chips and the company’s valuation; this hub is about the physical buildout — power, capex, and facilities — that turns chips into capacity. They’re complementary: NVIDIA sells the compute, the hyperscalers spend the capex and secure the power. Watching both captures the full AI-infrastructure trade.
Polymarket and Kalshi both list power, capex, and facility milestones as binary contracts. Polymarket carries deeper liquidity on the headline power and nuclear markets. Mantis queries both venues in real time and routes you to the best price with referral codes intact.