Goldman Sachs now projects that the annual spending on AI infrastructure will surpass $800 billion by the end of 2026.
This isn't just a minor adjustment; it's a significant upgrade driven by a confluence of powerful signals that have emerged in recent months. The most pivotal development came from the 'hyperscalers'—tech giants like Alphabet, Amazon, Microsoft, and Meta. During their late-April earnings calls, they collectively announced 2026 capital expenditure (capex) plans nearing $725 billion, a staggering 77% increase from the previous year. This massive commitment from the biggest players provides a solid foundation for the $800 billion forecast.
Further validating this trend is the performance of key suppliers. First, Nvidia reported record-breaking data center revenue, demonstrating that the demand for AI chips and systems is not just hype but a tangible reality. This signals that hyperscalers are actively purchasing the hardware needed for their ambitious build-outs. Second, supply chain partners like TSMC have also raised their own capex guidance to expand production capacity, ensuring they can meet the surging demand.
However, the story has evolved beyond just chips and servers. A new, critical factor is the enormous strain AI data centers are placing on power grids. Energy has become the primary bottleneck, forcing a significant portion of AI-related spending to shift towards infrastructure. This includes investments in upgrading electrical grids, building new power generation facilities, and even exploring long-term solutions like small modular nuclear reactors (SMRs). This spending on power and structures is now a core component of the AI capex boom, turning it into a first-order macroeconomic driver.
Finally, the financial runway for this expansion is clear. These tech giants have been aggressively tapping into the bond markets, raising tens of billions of dollars with ease. This access to capital ensures that funding is not an obstacle to their multi-hundred-billion-dollar infrastructure projects. When combined, these factors—hyperscaler ambition, supply chain validation, infrastructure bottlenecks, and open capital markets—create a coherent and data-backed picture of an AI investment supercycle well on its way to $800 billion.
- Capex (Capital Expenditure): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, buildings, an industrial plant, technology, or equipment.
- Hyperscaler: A large cloud service provider that can provide computing and storage services at a massive scale. Examples include Amazon Web Services (AWS), Google Cloud, and Microsoft Azure.
- PPA (Power Purchase Agreement): A long-term contract between an electricity generator and a customer, usually a utility, government, or company. They are often used to finance and de-risk new power generation projects.