NewsCryptoGoldman Sachs Projects $7.5 Trillion in AI Capex and What It Could Mean for Crypto

Goldman Sachs Projects $7.5 Trillion in AI Capex and What It Could Mean for Crypto

Author: ICO Bench·

Key Takeaways

  • Goldman Sachs estimates cumulative AI infrastructure spending of about $7.5 trillion to $7.6 trillion between 2026 and 2031.
  • The forecast allocates roughly $5.1 trillion to compute, $2.1 trillion to data centers, and $358 billion to power infrastructure.
  • Annual AI capex is projected to increase from an estimated $765 billion this year to $1.6 trillion by 2031.
  • Goldman expects NVIDIA to capture about 75% of the compute segment, which it says is an unprecedented level of concentration in a technology capex cycle.
  • The forecast is being used by crypto investors as a reference point for AI compute tokens, decentralized GPU networks, and related infrastructure projects.
Goldman Sachs Projects $7.5 Trillion in AI Capex and What It Could Mean for Crypto

Goldman Sachs has projected cumulative AI infrastructure spending of approximately $7.5 trillion to $7.6 trillion between 2026 and 2031. The forecast breaks down into roughly $5.1 trillion for compute, $2.1 trillion for data centers, and $358 billion for power infrastructure, while annual AI capex is expected to rise from an estimated $765 billion this year to $1.6 trillion by 2031.

Goldman’s research framework identifies NVIDIA as the dominant beneficiary of the compute layer, with the chipmaker expected to capture about 75% of that segment. Goldman said that level of concentration has no precedent in prior technology capex cycles.

The annual run rate alone would make this the largest discrete capital deployment in technology history. Goldman Sachs analysts also said consensus estimates have repeatedly lagged the pace of spending, noting that 2026 capex consensus rose from $465 billion to $527 billion within a single earnings season.

“$7.6 trillion in AI CAPEX by 2031? Striking but what really bothers me in Goldman Sachs’ estimates is the tiny share allocated to electricity,” Christophe Barraud wrote in an X post on July 11, 2026. “Over the full period, Goldman expects around $5.1 trillion to be spent on compute, more than $2.1 trillion on data centers and…”

For retail crypto investors tracking AI infrastructure, Goldman’s figures are not background noise. They form the macro framework against which AI compute tokens, decentralized GPU networks, and AI token presales are being evaluated, especially as the spending cycle is concentrated in the same infrastructure layers that crypto-native projects often try to replicate or complement.

The $7.5 trillion forecast represents a capital cycle large enough that even a 0.1% redirection toward crypto-native infrastructure rails would amount to a material token-market catalyst. The key question is not whether the cycle exists, but which token sectors are structurally positioned to capture a share of it.

What Goldman’s $5.1 Trillion Compute Forecast Says About Crypto’s AI Opportunity

The headline number is large, but the internal structure of Goldman’s forecast is what makes it more useful for assessing crypto positioning. The $5.1 trillion compute layer is the largest spending category, more than twice the data center allocation and roughly fourteen times the power budget.

Goldman’s research ties that estimate to specific hardware assumptions, including next-generation GPU nodes priced at approximately $80,500 per unit, drawing 3,000 watts per package, with a power usage effectiveness ratio of 1.2. Goldman said these are not speculative inputs but reflect actual procurement economics for NVIDIA’s current-generation architecture.

The expected concentration of roughly three-quarters of the compute layer at NVIDIA is what makes decentralized alternatives analytically relevant. When one vendor is projected to capture such a large share of a $5.1 trillion spending category, supply-chain risk, pricing power, and export-control exposure create a structural opening for permissionless compute alternatives.

Goldman’s research also notes that insecurity, or fear of missing out on AI infrastructure, is as much a driver of the capex boom as measured return on investment. That suggests the cycle carries duration risk as well. Goldman’s broader AI investment thesis has already begun influencing capital flows across emerging markets and asset classes.

Power is the smallest line item in dollar terms at $358 billion, but Goldman’s analysts identify it as the main bottleneck limiting deployment speed, rather than compute availability. The firm’s data center cost assumptions run at $15 million per megawatt of capacity, with new power priced at $2,500 per kilowatt.

Brownfield data center space currently accounts for 15% of 2026 deployments and is projected to rise to 30% by 2031 as greenfield sites face grid interconnection delays. Goldman expects the power bottleneck to persist throughout the 2026–2031 window, making location-flexible, decentralized AI compute infrastructure a more relevant theme for crypto-native infrastructure projects.

EXPLORE: Best Memecoins Presales to Watch in July

The post Goldman Sachs $7.5T AI Capex Breakdown and What It Means for Crypto appeared first on icobench.com .