Goldman Sachs Maps $7.6 Trillion AI Infrastructure Spending Through 2031

Goldman Sachs projects $7.6 trillion in cumulative AI infrastructure capital expenditure between 2026 and 2031, with Nvidia expected to capture 75% of the $5.1 trillion compute layer. Power emerges as the critical bottleneck despite representing only $358 billion of total spending, with companies like Vistra securing long-term nuclear partnerships to support AI deployment.

Goldman Sachs' $7.6 trillion infrastructure projection underscores the massive capital requirements for scaling artificial intelligence systems globally. The forecast spans six years and encompasses compute, power, cooling, and networking layers essential for training and deploying advanced AI models. This spending trajectory reflects the acceleration in enterprise and cloud provider commitments to AI capabilities, driven by competitive pressures and the need to integrate generative AI into business operations at scale.

The dominance of Nvidia in the compute layer (75% of $5.1 trillion) highlights the company's architectural advantage and the semiconductor industry's centrality to AI infrastructure. However, the disaggregation of spending reveals a critical insight: power infrastructure, though smallest at $358 billion, represents the true constraint limiting deployment velocity. Data center operators face grid capacity limitations, environmental concerns, and lengthy permitting timelines that restrict expansion despite abundant capital. The mention of Vistra's 20-year nuclear contract signals market recognition that renewable and stable power sources are essential for meeting AI infrastructure demands.

For investors and stakeholders, this analysis creates divergent opportunities. While chip manufacturers benefit from sustained demand, the power and cooling segments present underappreciated value creation areas. Energy companies, particularly those with nuclear or stable generation assets, gain leverage in negotiations with hyperscalers. For cryptocurrency markets, reliable power infrastructure implications affect mining operations and blockchain network sustainability discussions, potentially supporting arguments for carbon-aware development.

Monitoring actual capex deployment against these projections matters significantly. If power constraints materialize faster than anticipated, it could slow AI model training cycles and alter competitive dynamics among cloud providers, indirectly affecting broader tech valuations and cryptocurrency market narratives around energy efficiency.

• →Goldman Sachs forecasts $7.6 trillion AI infrastructure spending through 2031, with compute representing the largest segment at $5.1 trillion
• →Nvidia is projected to capture 75% of the compute layer, cementing semiconductor dominance in AI infrastructure buildout
• →Power infrastructure emerges as the critical bottleneck despite being the smallest budget segment at $358 billion
• →Long-term nuclear power contracts like Vistra's 2,600 MW deal demonstrate energy security is becoming a competitive differentiator for hyperscalers
• →The spending projection has implications for energy companies, semiconductor manufacturers, and cryptocurrency mining economics