A new report from Wells Fargo suggests the AI industry is entering a massive physical build-out phase, shifting the entire investment landscape.
At the heart of the report is a staggering forecast: global hyperscaler computing capacity is expected to double from 49 gigawatts (GW) in 2025 to 98 GW by 2027. To fuel this growth, companies like Amazon, Google, and Microsoft are projected to spend a colossal $2.47 trillion on capital expenditures between 2026 and 2028. This isn't just an upgrade; it's a fundamental expansion of the digital world's foundation.
So, why the sudden explosion in spending? The core reason is a shift in what's holding AI back. For a long time, the main challenges were finding top talent or developing smarter AI models. Now, according to Wells Fargo, the primary bottleneck is simply access to compute. In other words, having enough raw processing power, housed in massive data centers and supplied by a stable electric grid, has become the most critical factor for success in the AI race.
This isn't just a theory; we're seeing it play out in real-time. First, look at the companies themselves. Alphabet (Google's parent company) recently announced its 2026 capital expenditure could reach nearly $185 billion, almost double its 2025 spending. Microsoft also reported a quarterly capex of $37.5 billion, stating that customer demand for its AI services is already exceeding its available supply. These figures confirm that Big Tech is aggressively spending to secure its infrastructure advantage.
Second, the physical constraints are becoming undeniable. Reports from energy authorities like PJM, a major U.S. grid operator, warn of potential power shortfalls of up to 60 GW in the next decade, largely driven by data center demand. This aligns with JLL's outlook on a multi-trillion dollar "AI supercycle" where power availability is the main limiting factor. The electric grid wasn't built for this kind of demand, and the race is on to expand it.
Ultimately, this changes how we should think about investing in AI. The focus is broadening from just chip designers like Nvidia to the entire ecosystem that supports them. This includes utility companies providing power, real estate firms building data centers, and networking companies that connect everything. The AI revolution will be built on a foundation of concrete, steel, and immense electrical power.
- Glossary
- Hyperscaler: A term for a massive cloud services provider that can offer computing and storage services at an enormous scale, such as Amazon Web Services (AWS), Microsoft Azure, and Google Cloud.
- Capex (Capital Expenditure): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, buildings, and equipment.
- PJM: PJM Interconnection is a regional transmission organization (RTO) in the United States that coordinates the movement of wholesale electricity in all or parts of 13 states and the District of Columbia.