Kalshi launches GPU futures forward curve and recorded more search impressions than Coca-Cola during the World Cup final

Kalshi: GPU futures and a World Cup search spike

Tarek Mansour says Kalshi generated more search impressions during the World Cup final than Coca-Cola and Adidas, and briefly spiked above ChatGPT and Instagram in search volume on the day of the final. His read is that most users aren't just trading — they're using the markets to follow what's happening, which is why engagement spikes so sharply around peak events like elections, Fed decisions, and major sports finals.

GPU forward curve

The more consequential announcement is Kalshi's launch of a compute forward curve. The product lists prediction markets on Nvidia H200 hourly prices at weekly intervals for four weeks, then monthly thereafter. Mansour frames it as the first step toward a full futures and derivatives market for compute, built on the same logic as commodity forward curves for oil or interest rates.

Pricing data comes from an index provider called Orin, which aggregates transaction prices across a large number of compute nodes. Mansour is explicit that the underlying benchmark isn't settled yet — the plan is to list several forward curves, watch which one earns trust and volume, and concentrate on the winner. He draws the analogy to WTI versus Brent crude: the standard became the standard partly because it got traction first, not because it was intrinsically the right measure.

The forward curve already has data. As of the segment, the H200 hourly price curve was notably flat, which Mansour reads as the market pricing innovation roughly in line with consumption — supply and demand in rough balance rather than a sharp dislocation in either direction.

“In terms of number of impressions and trends on search, we ended up being number one — we essentially generated more impressions as a brand than essentially all the other consumer brands that were going pretty hard at the World Cup, including Coca-Cola and Adidas... Compute as a commodity is gonna probably be the largest commodity on the planet, and so the derivative market for it will probably be the largest derivative market on the planet.”

Market structure and TAM

The natural participants are anyone who is long or short on compute prices — hyperscalers, AI labs, Neo Clouds, and consumers of GPU capacity who want to hedge their annual compute budgets. Mansour notes that for the first time in roughly a year, compute pricing has stopped its continuous decline and is showing volatility and cyclicality, which is the precondition for a healthy derivatives market.

On sizing, Mansour cites the rule of thumb that derivative markets historically reach 10 to 15 times the size of the underlying spot market. With compute spend approaching $1 trillion and Mansour projecting a roughly 10x increase by 2030, he argues compute derivatives could eventually outpace Treasury futures as the largest derivative market on the planet. He acknowledges that hedgers typically represent only around 5% of liquidity in mature commodity markets — the rest is speculation and arbitrage — so the TAM math depends heavily on attracting a broad base of non-hedging participants.

Kalshi's existing forecaster community is the bridge. The same users pricing election outcomes and inflation figures are now pricing compute. Mansour's argument is that forecast accuracy compounds with practice regardless of the underlying market, and that accuracy is what gives the forward curve credibility.

Manipulation resistance

Mansour points to a recent attempt to manipulate a California election market — a trader put in $2 million to move the price on Spencer Pratt winning, and the price correction took nine seconds. His argument is that manipulation creates an immediate arbitrage opportunity for anyone pricing the underlying correctly, so markets self-correct faster than most alternatives. He contrasts this with polling, where self-interested actors routinely commission biased data with no comparable correction mechanism.

Midterms

Politics remains a volume driver. Mansour expects a material ramp-up heading into October and November, with the midterms likely representing a major market cluster. The overall framing is that prediction market volume tracks whatever is dominating public attention on a given day — and midterm cycles reliably generate that.