NextFin News - The global race to build out artificial intelligence infrastructure is hitting a physical wall that no amount of capital can immediately dismantle. According to a report released Thursday by Knight Frank, the global property consultant, the AI boom is projected to demand up to $1.6 trillion in investment by 2030. However, the primary constraint on this growth has shifted from the availability of silicon or funding to the raw capacity of the world’s aging power grids.
Knight Frank, a firm that has historically maintained a bullish outlook on data center real estate as a core institutional asset class, now warns that electricity infrastructure has become the "gating factor" for the industry. The firm’s data center team, led by Stephen Beard, notes that global data center capacity is forecast to nearly double from 62GW in 2025 to over 110GW by 2028. Within that footprint, AI-specific workloads are expected to triple, claiming nearly a quarter of all global data center activity within three years. This projection, while aggressive, aligns with the massive capital expenditure plans recently signaled by the world’s largest technology firms.
The scale of spending is unprecedented. Capital expenditure by the "Big Four"—Microsoft, Amazon Web Services (AWS), Google, and Meta—is expected to exceed $650 billion in 2026 alone, a 73% surge from the previous year. AWS and Google are leading this charge with annual forecasts of $200 billion and $185 billion respectively. Yet, this wall of money is colliding with a utility sector that operates on decades-long cycles rather than the quarterly sprints of Silicon Valley. In London, connection delays for new data centers now stretch to nearly 10 years, while vacancy rates in critical hubs like Frankfurt and Ashburn, Virginia, have plunged below 1%.
This scarcity is forcing a radical "power-first" pivot in corporate strategy. Rather than selecting sites based on proximity to users or tax incentives, investors are now chasing deliverable megawatts wherever they can be found. This has led to a series of extraordinary "behind-the-meter" energy deals. Microsoft recently signed a 20-year agreement to restart the Three Mile Island nuclear facility, while Amazon secured 1.9GW of nuclear capacity through an $18 billion deal with Talen Energy. These moves suggest that the largest tech companies are effectively becoming vertically integrated energy firms to bypass the bottlenecks of public grids.
However, the Knight Frank analysis—which primarily reflects the perspective of a real estate intermediary focused on supply-side constraints—does not represent a universal market consensus regarding the sustainability of this spending. While the firm highlights the rise of "neocloud" operators like CoreWeave, which are scaling on the back of multi-billion dollar GPU contracts, some analysts remain skeptical of the underlying economics. The report itself mentions a $300 billion cloud agreement between Oracle and OpenAI that has already seen revisions, including the cancellation of a major expansion in Texas last month. With OpenAI’s committed cloud spend dwarfing its current annualized revenue, the risk of a massive overbuild looms if AI monetization fails to keep pace with infrastructure costs.
The geographical focus is also shifting toward markets that can offer immediate power availability. Jiya Agarwal, an analyst at Knight Frank, points out that the Asia-Pacific region now accounts for 41% of AI-focused colocation leasing. Markets like India, Malaysia, and Indonesia are seeing a surge in interest not because of their local user base, but because they offer the land and power delivery timelines that established hubs like Tokyo or Seoul can no longer provide. As grid investment requirements reach staggering levels—including an estimated $3.8 trillion needed in China by 2050—the ability to secure energy has become the defining currency of the digital age.
Explore more exclusive insights at nextfin.ai.
