---
title: AI Spending Reaches the Scale of Global Oil and Gas for the First Time in 2026
description: AI capital spending reaches the scale of the global oil and gas industry in 2026, and Wall Street is building compute futures to trade computing power like a commodity.
author: Darie Nani (Editor-in-Chief)
updated: 2026-08-25T13:23:39.648Z
canonical: https://www.sovereignmagazine.com/article/ai-capex-oil-gas-compute-futures-2026
image: https://cdn.nanimediahouse.com/ai-capex-oil-gas-compute-205551.webp
categories: Markets
content_type: Analysis
region: Global
publication: Sovereign Magazine
schema_type: Article
---

The world is now putting as much capital into building artificial intelligence as it puts into finding and producing oil. Goldman Sachs projects roughly $765 billion of total AI capital spending in 2026, and the piece of that coming from the largest technology companies, around $680 billion, matches the entire global oil and gas industry's capital spending for the year. Big Tech's AI outlay is up about 78 percent from 2025. For investors and operators who allocate capital, the shift is concrete: an industry that barely existed a decade ago is drawing money on the scale of the sector that has powered the global economy for a century.

Financiers are now building a way to trade computing power the way they trade crude, wheat or natural gas. They are creating a hedging tool for an asset the market has only just started spending on.

## Goldman Sachs puts total AI capital spending at $765 billion this year

The $765 billion figure comes from Goldman Sachs and covers total AI capital spending in 2026. The roughly $680 billion hyperscaler slice is the part that reaches parity with energy. Oil and gas capital spending sits at about the same $680 billion, so for the first time the money going into GPUs, data centers and the power to run them equals the money going into wells, rigs and pipelines. The comparison is the one investors keep reaching for because it puts a familiar number on an unfamiliar buildout.

## Chamath Palihapitiya calls it the moment compute gets its first futures contract

Chamath Palihapitiya said in a post on X on 24 August 2026: "AI capex is projected to reach $765B in 2026, passing oil and gas for the first time. Compute is now getting its first futures contract. Here is what has to be true for it to work…" The post links to his own essay laying out the case.

His figure tracks the reporting. His "first futures contract" line is a simplification worth correcting. Computing power is genuinely being turned into a tradable, hedgeable asset, but through several venues at once rather than a single first contract.

## The first regulated US compute futures are due to launch in October

CME Group and Silicon Data announced on 11 August 2026 plans to launch compute futures, priced off a GPU-rental index, on 5 October 2026, pending approval from the Commodity Futures Trading Commission. Described as the first regulated compute futures, they are not yet live as of late August. They are not the first compute contracts of any kind: unregulated GPU perpetual contracts have traded since January 2026, run by Architect Financial on an exchange in Bermuda, and ICE has announced regulated GPU futures of its own with a venue called Ornn. What Palihapitiya is pointing at is real even if the "first" is loose. A standardized, exchange-traded price for a unit of compute is arriving, and the first regulated US version has a launch date.

## No one has yet shown the spending pays off

The energy comparison is not fully apples to apples. Oil and gas capital spending is recurring, the cost of drilling, maintaining and producing year after year. Most of the AI number is one-time buildout of data centers and chips. The returns on that buildout remain unproven. Only a small share of companies have quantified AI productivity gains or tied AI to earnings, and hyperscaler capital spending is running high against revenue, with Meta near 54 percent, Microsoft near 47 percent and Alphabet near 46 percent in 2026. Morgan Stanley has flagged the main risk of the year as AI capital spending failing to lift productivity. The hardware itself is a wasting asset: GPUs lose much of their value in the first year and are running below their projected utilization.

Compute is also harder to trade than a barrel of oil. You cannot store it or ship it, so a standardized futures contract carries real basis risk, the gap between the index price and what any given buyer actually pays for capacity. Building a way to hedge compute, before the returns are in, is the financial system pricing its own doubt about whether the biggest capital build of the decade will pay off.

## FAQ

**Q: How much is being spent on AI in 2026?**
Goldman Sachs projects roughly $765 billion of total AI capital spending in 2026. The largest technology companies account for about $680 billion of that, up roughly 78 percent from 2025.

**Q: Is AI really outspending oil and gas?**
It reaches parity. The roughly $680 billion of Big Tech AI capital spending matches the entire global oil and gas industry's capital spending of about $680 billion, the first time AI infrastructure has reached the energy sector's scale. The comparison has a caveat: AI spending is largely one-time buildout, while oil and gas spending recurs every year.

**Q: What are compute futures?**
They are contracts that let buyers and sellers lock in a future price for computing power, priced off an index of GPU-rental costs, the same way oil or wheat futures work. CME Group and Silicon Data plan to launch the first regulated US version on 5 October 2026, pending CFTC approval. Unregulated GPU contracts have traded since January 2026.

**Q: Why does it matter that the returns are unproven?**
Because the case for spending at this scale rests on gains that have not yet shown up in earnings. Few companies have quantified AI productivity improvements, hyperscaler spending is high relative to revenue, GPUs depreciate quickly and are running below expected use, and Morgan Stanley has named weak productivity payoff as the main risk of the year. That is why a market to hedge compute is telling: it prices the doubt as much as the boom.