# Colliers Links $800B AI Buildout to Real Estate Demand > Source: > Published: 2026-08-12 23:50:24+00:00 # Colliers Links $800B AI Buildout to Real Estate Demand Colliers published an August 12 analysis connecting Morgan Stanley's May forecast of about $800 billion in 2026 AI-infrastructure investment to rising demand for data-center land, power, fiber, and construction capacity. The analysis also highlights more than $850 billion in future cloud-provider lease commitments reported from regulatory filings, signaling that the buildout extends well beyond current capital spending. Colliers published an August 12 analysis connecting Morgan Stanley's forecast of roughly $800 billion in 2026 investment by large U.S. technology companies to growing demand for data-center land, power, fiber, construction, and industrial capacity. Morgan Stanley first published the estimate on May 11, saying the total would be almost double 2025 spending and about three times the 2024 level. ### Two different measures of the buildout The $800 billion figure is an annual investment estimate. Morgan Stanley's Andrew Sheets said the spending covers AI infrastructure including chips, power, and data centers, and argued that demand has remained strong even as input prices and borrowing costs rise. The firm estimated about $1.1 trillion of comparable spending in 2027. Colliers pairs that forecast with a second measure: future lease commitments. It cites a Bloomberg analysis of regulatory filings showing more than $850 billion in future obligations among major cloud providers. Data Center Dynamics separately reported that Meta and Microsoft added about $120 billion of future data-center lease commitments in their latest reported quarters, including roughly $79 billion from Meta and $41 billion from Microsoft. These figures should not be combined. Annual capital investment and future lease obligations measure different commitments across different time horizons. The lease total is not one year's spending or immediately recognized debt; it is evidence that providers are reserving facilities and capacity expected to come online over time. ### Why the real-estate layer matters Colliers' contribution is to translate the technology-spending cycle into constraints for physical infrastructure. AI capacity requires suitable sites, utility interconnections, power generation, fiber routes, cooling equipment, and specialized construction. When multiple hyperscalers pursue those inputs simultaneously, competition can affect project schedules and prices well beyond the companies buying accelerators. For data and AI teams, the practical signal is not that $800 billion guarantees abundant compute. The buildout still depends on facilities becoming operational and on power, networking, and hardware arriving together. Procurement plans should distinguish announced spending from commissioned capacity, and track lease commencements, grid access, and construction milestones alongside chip supply. The forecasts remain estimates rather than realized results. Morgan Stanley's May projection and Colliers' August interpretation describe the scale and direction of the cycle; later company filings will determine how much of the planned investment is actually deployed. ## Key Points - 1Morgan Stanley estimated about $800 billion of 2026 AI-infrastructure investment by large U.S. technology companies, almost double 2025 spending. - 2Colliers connected that forecast to demand for data-center sites, power, fiber, construction, and other physical infrastructure. - 3More than $850 billion in future cloud-provider lease commitments is a separate multi-year measure and should not be treated as one year's capital spending. ## Scoring Rationale The analysis gives practitioners and infrastructure planners a useful, source-backed view of the scale and physical dependencies of the 2026 AI buildout. Its impact is meaningful but bounded because the spending and lease figures are estimates and commitments rather than completed capacity. ## Sources Primary source and supporting public references used for this report. Practice interview problems based on real data 1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with. [Try 250 free problems](/problems)