What is AI tenant concentration risk? AI tenant concentration risk is the exposure an office landlord takes on when a large share of its rent roll comes from artificial intelligence companies, a single industry whose tenants are frequently young, venture funded, and carrying risks a traditional credit file does not capture. On October 13, 2026, Moinian Group founder Joe Moinian put a number on how much he will accept, telling Commercial Observer that his firm sits "somewhere around 2 percent, and will never be over 3 percent, exposed to AI in our office tenant base." For the framework behind this screening, see our guide to AI deal analysis real estate.
Key Takeaways
- Moinian Group, a private landlord with a portfolio exceeding 20 million square feet, holds AI exposure in its office tenant base near 2 percent and will not exceed 3 percent.
- CBRE puts AI companies at 32 percent of tech tenant demand in Manhattan, so the exposure Moinian limits is also the market's fastest growing source of office demand.
- Character.AI's October 13 Chapter 11 filing shows AI tenant credit risk includes litigation liability, not just a funding runway that runs dry.
- One AI tenant at 10 percent of a building's revenue can push DSCR from 1.33x to 1.10x on default, breaching a 1.25x covenant. At 3 percent exposure, the same default leaves 1.26x.
- Derive your own ceiling from your loan covenant rather than copying 3 percent, and price the risk through guarantees and letters of credit before refusing the tenant.
What Moinian Actually Said About AI Tenant Exposure
Moinian set a self-imposed ceiling, not a lender covenant or an industry standard. Speaking to Commercial Observer reporter Mark Hallum on October 13, 2026, he said his firm is "somewhere around 2 percent, and will never be over 3 percent, exposed to AI in our office tenant base." That is a current position near 2 percent with 3 percent as the hard stop, tighter than the flat "caps AI tenants at 3 percent" that several aggregator headlines reported. He is not avoiding AI tenants either: Moinian signed three AI firms at 60 Madison Avenue in June 2026, including insurance workforce developer Pace, and Synthesia expanded at 245 Fifth Avenue. The cap governs how large those tenants may become, not whether they get a lease.
Context explains the caution. Moinian is courting an anchor for a 1.9 million square foot Hudson Yards tower designed by FX Collaborative, and said the prospective tenant "would take half of the building" and that "you need an anchor to get a construction mortgage." A landlord underwriting a tower on one tenant's credit has reason to keep the rest of the rent roll boring. On AI's effect on leasing he was direct: "Based on our experience, yes, for sure there is uncertainty. But it's how much of it can be tolerated."
Why AI Tenants Are a Different Credit Problem
AI tenants break standard tenant credit work in three ways.
1. Runway, not earnings. A venture backed AI firm's ability to pay rent in year four of a seven year lease depends on a future financing round, not trailing cash flow. Metrics built for operating businesses have little to grip. Our guide to AI commercial tenant credit analysis covers the documents and ratios that do work on private tenants.
2. Legal liability can outrun the business. Character.AI filed for Chapter 11 on October 13, 2026 in the U.S. Bankruptcy Court for the District of Delaware, with Walt Disney agreeing to pay $20 million as stalking-horse bidder for select intellectual property. The driver was litigation over teen mental health claims, not a collapse in AI demand. A company once valued near a billion dollars entered Chapter 11 for reasons no trailing financial statement would have flagged.
3. The credits are correlated. Ten AI tenants are not ten independent credits. They draw on the same venture funding market and face the same regulatory environment, so a slowdown hits the cohort in one quarter. Data center landlords repriced this first: Bloomberg reported on September 30, 2026 that AI startups seeking capacity are being pushed to produce credit backstops from large technology partners.
What AI Tenant Concentration Costs When a Lease Fails
Take a 500,000 square foot office building at $75 per square foot: $37.5 million of gross revenue against $17.5 million of operating expenses produces $20 million of NOI. Against $15 million of annual debt service, DSCR is 1.33x.
Now default a 50,000 square foot AI tenant, 10 percent of the building. You lose $3.75 million of gross rent and save roughly $250,000 of variable expenses, so NOI falls about $3.5 million to $16.5 million. Three things happen at once:
- DSCR breaks. $16.5 million divided by $15 million is 1.10x, breaching a 1.25x covenant and potentially triggering a cash trap long before the space is re-let.
- Value drops hard. At a 6.5 percent cap rate, $3.5 million of lost NOI is roughly $54 million of value on one tenant's failure.
- Re-leasing costs land on top. Tenant improvements at $100 per square foot on 50,000 square feet is $5 million of capital, before free rent, commissions, and 12 to 18 months of downtime.
Rerun it at Moinian's ceiling. A 3 percent exposure is about $1.125 million of gross rent, so NOI falls to roughly $18.95 million and DSCR lands at 1.26x. Same failure, covenant intact. A concentration cap is not a view on whether AI companies will succeed; it decides whether one industry's bad quarter can default your loan.
How to Set and Price Your Own AI Exposure Limit
Copying 3 percent is the wrong move; the right limit is what your debt structure can absorb. Measure exposure by contract rent, not square footage: a premium-paying AI tenant is a bigger exposure than its footprint suggests. Then derive the ceiling from your covenant: find the revenue loss that takes DSCR to its limit and cap below that. Test at asset level too: 3 percent portfolio-wide can conceal 25 percent in one building.
A cap should be the last line of defense. Lease structure lets you take AI demand at today's rents while charging for the risk:
- Letters of credit sized to real exposure: unamortized tenant improvements plus expected downtime, not a conventional two months of rent.
- Parent or investor guarantees: what data center landlords now demand, converting a startup credit into a rated one.
- Amortized TI with clawbacks: recover capital through rent instead of funding it upfront.
- Shorter initial terms with renewal options: five years plus two options prices the uncertainty instead of ignoring it.
- Funding disclosure covenants: require notice of financing rounds and cash position, so a tenant six months from trouble is visible.
Owners who want these provisions modeled against their actual rent roll can reach out to Avi Hacker, J.D. at The AI Consulting Network. Re-test quarterly, since exposure moves whenever a lease rolls; our guide to AI stacking plans for office rent rolls covers tracking it alongside rollover risk.
The Caveats Before You Cap Anything
A 2 to 3 percent ceiling is defensible for Moinian and wrong for many owners. It reflects a specific balance sheet: a large private owner with a low land basis. An owner of one Midtown South building has no path under 3 percent when a single floor exceeds it.
It also cuts against where demand is. Per CBRE's 2026 Tech Gateway Office Markets report, tech accounted for 23 percent of U.S. office leasing in Q1 2026, up from 17 percent in 2025, and 8.1 million square feet of AI tenants were in the market across the five largest U.S. tech gateways, with AI at 32 percent of Manhattan tech tenant demand. JLL's U.S. Office Market Dynamics shows technology-led markets leading net absorption. A hard cap can mean turning away the best rent on the table.
Caps are also blunt: a ceiling treats a profitable AI firm with a parent guarantee the same as a seed-stage startup. Where you have underwriting capacity, credit-based screening beats industry-based screening. For more on AI demand and Manhattan absorption, see our analysis of NYC AI office leasing and the hiring gap.
Frequently Asked Questions
Q: What is a reasonable AI tenant concentration limit for an office landlord?
A: There is no industry standard. Moinian Group runs near 2 percent with a 3 percent ceiling, while common concentration frameworks flag single-tenant exposure above 15 percent of revenue. The defensible answer comes from your own loan: find the revenue loss that takes DSCR to its covenant, then cap below that.
Q: Does Moinian's cap mean AI office demand is weakening?
A: No, and the evidence points the other way. CBRE reported 8.1 million square feet of AI tenants in the market across the five largest U.S. tech gateways. Moinian is still signing AI leases. The cap addresses concentration, not a fading market.
Q: How do I tell whether a tenant counts as an AI company?
A: Define it by revenue source, not technology use. If the tenant's product or service is the AI itself, it belongs in the bucket. A law firm or bank using ChatGPT, Claude, or Gemini internally does not, because its ability to pay rent does not depend on AI funding markets.
Q: Which lease terms protect best against an AI tenant default?
A: A letter of credit sized to unamortized tenant improvements plus expected downtime, a parent guarantee, and amortized TI with a clawback. Shorter initial terms and funding disclosure covenants also help, since they provide an exit and an early warning. The AI Consulting Network specializes in exactly this lease-level risk work.