What is the Stanford AI Index 2026? The Stanford AI Index 2026 is the annual census of artificial intelligence published by Stanford's Human-Centered AI Institute, and the 2026 edition documents 5,427 AI data centers in the United States, 29.6 GW of dedicated AI power capacity (roughly enough to power New York State at peak demand), and a record 25.4 billion dollars in data center construction starts in January 2026 alone, with construction increasingly concentrated in the South Central US, Southeast US, Pennsylvania, Indiana, Ohio, and Oregon. The report from Stanford's Human-Centered AI institute is the most comprehensive snapshot of where AI infrastructure capital is actually flowing, and it has direct implications for CRE investors weighing data center exposure, land plays in adjacent power markets, and the second-order effects on industrial, multifamily, and retail in nearby submarkets. For broader context, see our AI commercial real estate guide.
Key Takeaways
- The US hosts 5,427 AI data centers, more than 10 times any other country, with 29.6 GW of dedicated AI power capacity equivalent to powering New York State at peak demand.
- January 2026 saw 25.4 billion dollars in data center construction starts, the largest monthly total on record, putting full-year 2026 spending on pace for 116.4 billion dollars.
- An additional 70.8 billion dollars in projects is scheduled to start in the next six months, concentrated in the South Central and Southeast US plus Pennsylvania, Indiana, Ohio, and Oregon.
- Total AI compute capacity has grown 30 fold since 2021, with NVIDIA GPUs accounting for over 60% of global capacity, creating supply chain concentration risk.
- For CRE investors, the data points to clear winners (land near transmission and water in target markets) and clear losers (saturated Northern Virginia, parts of Phoenix where moratoriums are spreading).
What the Stanford AI Index 2026 Actually Reports
Released by Stanford's Human-Centered AI Institute in April 2026, the AI Index is the closest thing the AI industry has to an annual census. The 2026 edition runs over 400 pages and tracks compute, talent, energy, investment, and policy. The chapter most relevant to CRE investors documents the global geography of AI data center deployment, with the United States dominating in raw count and energy consumption.
The headline numbers from the report:
- 5,427 AI data centers in the United States, more than ten times any other country
- 29.6 GW of dedicated AI power capacity, approximately equivalent to peak power consumption in New York State
- 30 fold growth in AI compute since 2021, with capacity tripling annually
- NVIDIA GPUs account for over 60% of global AI compute capacity
- 72,816 tons of CO2 equivalent emissions for training Grok 4 alone, equivalent to driving 17,000 cars for a year
The construction data, sourced from ConstructConnect and analyzed alongside the Stanford findings, shows January 2026 hit a record 25.4 billion dollars in data center construction starts, lifting the trailing 12 month average to 9.7 billion dollars per month.
Where the Construction Is Actually Happening
The geographic concentration is the most actionable finding for CRE investors. The next 70.8 billion dollars of projects (over the next six months) is heavily concentrated in:
- South Central US: Texas (Dallas-Fort Worth, Austin, San Antonio), Oklahoma, Louisiana
- Southeast US: Atlanta, the Carolinas, Tennessee, northern Florida
- Pennsylvania: particularly the Philadelphia and central PA corridors
- Indiana: central and northern Indiana, near transmission infrastructure
- Ohio: the Columbus area and the SoftBank Portsmouth campus
- Oregon: Hillsboro and the Portland metro
What ties these markets together is power availability, water access, land cost, tax incentives, and proximity to fiber backbone routes. The traditional data center markets like Northern Virginia and parts of Phoenix are running out of grid capacity, which is why the next wave of capex is going elsewhere. CRE investors who own land or industrial inventory in these emerging markets are positioned to benefit from rising land prices, build-to-suit demand, and the ancillary multifamily, retail, and hospitality demand that follows construction.
Three Direct Implications for CRE Investors
1. Data Center Land and Power Plays
The single highest leverage CRE play remains owning or controlling power-adjacent land in the South Central and Southeast US. Sites with 100 MW or more of secured grid capacity, water access, and transmission proximity are trading at 3 to 8 times their pre-AI valuations. Investors without the capital to buy entire data center sites can play the adjacent industrial market (cooling equipment, racks, generators, fiber) which feeds the buildout. JLL and CBRE both publish quarterly trackers showing rent growth in industrial submarkets adjacent to active data center construction outpacing the broader industrial average by 200 to 400 basis points.
2. Multifamily and Retail Demand in Construction Hotspots
Each gigawatt of data center capacity under construction supports 2,000 to 4,000 construction jobs at peak and 200 to 600 permanent operations jobs. These workers need housing, food, and services within 30 to 45 minutes of the site. Multifamily operators with assets in markets like Hillsboro Oregon, central Indiana, or southwest Ohio are seeing absorption pickup and rent growth in submarkets directly tied to data center buildout. The same pattern is playing out for select-service hotels and convenience retail.
3. Avoiding Saturated and Resistance Markets
The flip side is equally important. Data center investment is increasingly bypassing markets where grid capacity is exhausted or community opposition has hardened. Northern Virginia, parts of Phoenix, and increasingly suburban Chicago are seeing capital flow to other locations. CRE investors should be cautious about underwriting incremental data center demand in these markets, even where current absorption looks strong.
The Risks the Stanford Report Highlights
The Stanford AI Index 2026 also flags several risks CRE investors should track. First, supply chain concentration: nearly every leading AI processor in US facilities is fabricated by TSMC in Taiwan, creating geopolitical risk. Second, environmental constraints: 29.6 GW of AI power consumption is creating water and grid stress that is starting to drive policy responses. Third, openness retreat: the Foundation Model Transparency Index dropped from 58 to 40, meaning the most capable AI is becoming less transparent, which complicates due diligence on the demand side of the data center thesis. According to Stanford HAI, these risks are accelerating, not diminishing.
What This Means for the Next 12 Months
For CRE investors, the practical takeaways are:
- Underwrite continued data center capex acceleration through at least 2027, with 2026 likely closing near or above the 116.4 billion dollar pace implied by January starts
- Focus new exposure in markets identified as construction hotspots (South Central, Southeast, Pennsylvania, Indiana, Ohio, Oregon)
- Pair data center exposure with adjacent multifamily, industrial, and retail in the same submarkets
- Maintain caution on saturated markets where grid capacity is the binding constraint
- Monitor community opposition trends, which are accelerating and can stop projects mid-cycle
For personalized guidance on positioning your CRE portfolio for the AI infrastructure buildout, connect with The AI Consulting Network. CRE investors who want help translating the Stanford data into specific market and asset selection decisions can reach out to Avi Hacker, J.D. at The AI Consulting Network.
Frequently Asked Questions
Q: How does the Stanford AI Index 2026 compare to other data center forecasts?
A: The Stanford report's findings are broadly consistent with hyperscaler capex forecasts (Moody's projects 700 billion dollars in 2026 hyperscaler spending, with roughly 75% going to AI infrastructure). The Stanford data is more granular on geographic concentration, which is what makes it useful for CRE investors targeting specific markets.
Q: Which CRE markets benefit most from the AI data center buildout?
A: The South Central US (Texas, Oklahoma, Louisiana), Southeast US (Atlanta, Carolinas, Tennessee), Pennsylvania, Indiana, Ohio, and Oregon are receiving the largest share of new construction capital. Markets adjacent to these construction hotspots typically see secondary benefits in multifamily, industrial, and retail.
Q: Is the AI data center buildout sustainable, or is it a bubble?
A: The Stanford data shows compute demand growing 3x annually since 2021, which has so far exceeded supply. Moody's and other rating agencies have warned that hyperscaler capex is straining free cash flow and rising debt levels could trigger a credit reassessment, but the underlying compute demand has continued to materialize. CRE investors should underwrite for accelerating buildout through 2027 while staying attentive to financing stress signals.
Q: What CRE asset types benefit most from AI data center growth?
A: Power-adjacent land, industrial assets serving the supply chain (cooling, racks, generators, fiber, electrical components), select-service hotels in construction markets, multifamily within 30 to 45 minutes of major data center campuses, and convenience retail near worker housing. Data center direct ownership is the highest return play but also the highest capital requirement.
Q: How should CRE investors track AI data center construction trends going forward?
A: Quarterly reports from JLL, CBRE, and Cushman & Wakefield's data center practices, ConstructConnect's monthly construction starts data, and the annual Stanford AI Index are the most reliable public sources. Most institutional investors supplement these with proprietary tracking of specific operator capex announcements and utility interconnection queues, which provide leading indicators 12 to 24 months ahead of construction starts.