Hyperscalers Set to Invest $1.15 Trillion in AI Infrastructure by 2027

• Goldman Sachs estimates total hyperscaler capex will reach $1.15 trillion between 2025 and 2027, reflecting a major infrastructure push to support AI compute growth.
  • Annual spending could hit ~$400 billion by 2027, driven by Big Tech firms including Microsoft, Amazon, Alphabet, Meta, and Nvidia.
• Data center power demand is forecast to increase 50% by 2027, rising from ~55 GW in 2023 to ~84 GW.
  • AI workloads are expected to represent over 25% of total demand, highlighting the strain on existing capacity.

💰 Utilization & Power Infrastructure

• Utilization rates at hyperscale data centers are projected to climb from ~85% in 2023 to ~95% by late 2026, nearing operational limits.
  • This may lead to capex moderation in 2027, as supply begins to catch up and efficiency improves.
• Utility sector capex is projected to grow 36% between 2024 and 2027, focused on expanding power generation, transmission, and distribution to support new data centers.
• Over the longer term, Goldman sees total grid-related investment exceeding $700 billion by 2030, primarily to support the AI infrastructure boom.

🔍 Key Drivers & Implications

• AI is the catalyst: Training and deploying AI models are significantly increasing compute intensity, leading to massive investment in chips, power, and cooling infrastructure.
• Capex normalization ahead: While spending remains elevated through 2026, Goldman expects a flattening in 2027 as hyperscalers shift from buildout to optimization.
• Wider participation: Beyond the Big 4, second-tier cloud and AI infrastructure providers are expected to account for a growing share of capex.

🧠 Summary: Hyperscaler Capex Outlook

Bottom Line:
Goldman Sachs sees AI as the main driver of a historic capex cycle in cloud and utilities. While investment intensity may ease by 2027, the structural demand from AI remains intact—and will continue reshaping digital and energy infrastructure for years to come.