Institutional Ground Leases: The Safehold Model and Reversion Economics

Key Takeaways

Ground Leases as an Institutional Asset Class

The institutional ground lease is a niche allocation within commercial real estate, despite a risk-return profile built on contractual rent and long-duration land exposure. Safehold's FY2025 10-K reports estimated portfolio Ground Rent Coverage (the ratio of property NOI to the annualized rent due to Safehold) of 3.4x as of December 31, 2025. The gap between that profile and the size of the market is structural: ground leases are illiquid, require specialized underwriting, and produce returns that institutional allocators struggle to benchmark against traditional property types.

The modern institutional ground lease (as distinct from the legacy ground leases that have existed for centuries) was effectively reinvented by Safehold Inc. (NYSE: SAFE), formerly iStar. Safehold took a dormant legal structure and rebuilt it as a scalable institutional product, complete with standardized documentation, a publicly traded vehicle, and a mechanism for monetizing the long-tail reversion value that makes ground leases so compelling but so difficult to underwrite.

Understanding the Safehold model is not just about understanding one company's strategy. It's about understanding how ground leases function as an investment vehicle: the economic logic, the risk allocation between lessor and tenant, and the embedded optionality that makes a ground lease position fundamentally different from either a bond or a direct property investment.

The Safehold Model

Safehold's model separates the land from the improvements and creates a ground lease on the land component only. The tenant retains full operational control of the building, collects all operating income, and pays a contractual ground rent to Safehold. At lease expiration (initial terms are typically 30 to 99 years; this article uses 99 years as the model case), the property, land and improvements, reverts to Safehold.

The model works because land and buildings have fundamentally different risk profiles. Land appreciates over time (it's non-depreciable, supply-constrained, and benefits from urban density). Buildings depreciate physically and functionally (they require capital expenditure, become obsolete, and eventually need replacement). By separating ownership, each party holds the component best suited to their investment thesis:

• Safehold (ground lessor): Holds the land, which appreciates. Collects contractual rent during the term. Receives the entire property at reversion. Takes no operating risk: no tenanting, no capex, no management. The position behaves like a very long-duration bond with contractual escalators (CPI-linked, capped, or fixed) and equity upside at maturity.
• Property owner (ground tenant): Operates the building, collects NOI, pays ground rent as a fixed cost. Benefits from lower cost of capital (because the land cost is converted from equity to a periodic lease payment, reducing the equity required to acquire the property). Trades reversion value for lower basis and higher current returns.

Safehold prices its ground leases to make both positions economically attractive. The ground rent is set below the property's return on the land component, creating positive leverage for the tenant. As of December 31, 2025, per Safehold's FY2025 10-K, the portfolio had a combined property value (land + improvements) of $15.9B against a ground lease cost of $6.7B.

The 30-45% Cost Basis Math

Safehold's most important structural feature is the relationship between its cost basis (what it pays for the ground lease position) and the combined property value (the total value of land + improvements at the property).

Safehold typically acquires ground lease positions at 30-45% of combined property value. On a $100M property, Safehold's investment (the price of the ground lease) is $30-45M. The remaining $55-70M represents the tenant's leasehold capital (equity plus debt). Safehold's 30-45% cost basis buys three things:

1. Ground rent for the lease term. Paid with periodic escalations (CPI-linked with caps, or fixed). Safehold's stated target initial cap rate is 4.25-5.0%; its FY2025 10-K describes a first-year cash return of 2.5-5.5% and effective yields (including escalators) of 4.5-7.5%.
2. Land reversion at lease expiration. The land reverts to Safehold at the end of the term (99 years in this example). The value of the land at that point depends on location and long-run appreciation, but even modest assumptions produce substantial terminal values over a century.
3. Improvement reversion at lease expiration. The improvements also revert. At 99 years, the original building may have been demolished and replaced multiple times, but whatever improvements exist on the land at expiration belong to Safehold.

The 30-45% cost basis creates embedded upside. If the combined property value is $100M today and the property appreciates at 2.5% annually (a simplification that treats improvements as renewed over time), it will be worth $1.15B in 99 years. Safehold paid $35M for a position that produces current income and eventually controls a $1.15B asset. The gap between $35M and the present value of the total cash flows (ground rent + reversion) is Safehold's embedded value creation.

Reversion Economics

The reversion is both the most valuable and the most difficult-to-underwrite component of a ground lease position. In present-value terms, a 99-year reversion is nearly worthless: discounted at 7.0%, $1.15B in 99 years is worth approximately $1.4M today. But the reversion is not purely a present-value exercise. Its significance lies in three dimensions:

1. The Reversion Creates Asymmetric Upside

Safehold's current income from ground rent produces a baseline return, with a stated target initial cap rate of 4.25-5.0% on cost. The reversion is upside above that baseline. If the property appreciates faster than assumed (3.5% instead of 2.5%), the terminal value at year 99 is $3.0B instead of $1.15B. If the property happens to be in a location that benefits from secular urbanization or infrastructure investment, the upside is even larger. The ground lessor participates in 100% of this appreciation without bearing any operating risk during the term.

2. The Reversion Shortens Over Time

While a 99-year reversion is nearly negligible in PV terms, the reversion on a 30-year remaining term is substantial. At year 69 of a 99-year ground lease, the property is worth about $549M (after 69 years of appreciation at 2.5%). The reversion at year 99 is worth about $1.15B, which discounted 30 years at 7.0% is approximately $151M, about 28% of the year-69 property value. The reversion becomes increasingly material as the lease matures, and this is when Safehold's position becomes most valuable relative to its cost basis.

3. Renewals Reset the Clock

Some ground leases include tenant renewal options, with rent set by lease-specific terms. (Fair market value resets are common in the broader market, but Safehold's website describes its leases as having no FMV resets.) If the tenant renews, the ground lessor keeps collecting ground rent and the reversion pushes further out. If the tenant does not renew, the property reverts. Either outcome keeps the lessor's position intact, and this embedded optionality is easy to miss in standard DCF models.

The CARET Mechanism

Safehold's most distinctive financial engineering is the CARET structure. CARET addresses the fundamental problem of ground lease investing: the reversion value is enormous in nominal terms but nearly worthless in present-value terms because it sits 99 years in the future. Per Safehold's 10-K, Caret units generally entitle holders to a share of the net proceeds from the disposition of ground lease assets in excess of Safehold's investment basis, which separates that embedded value from the ground rent stream.

Safehold tracks its Unrealized Capital Appreciation (UCA), the difference between the combined property value of its portfolio and its aggregate cost basis. As of December 31, 2025:

Table 1: Safehold portfolio metrics as of December 31, 2025 (rounded; exact figures are $15,947M, $6,675M and $9,272M). The $9.3B UCA represents the embedded appreciation that Safehold has not yet realized. The MSD Partners price is one transaction data point: Caret units share in disposition proceeds, not only the year-99 reversion.

CARET units represent interests in this unrealized appreciation. On March 31, 2023, Safehold sold 100,000 Caret units to affiliates of MSD Partners for an aggregate purchase price of $20.0M, effectively monetizing a fraction of its reversion rights decades before lease expiration. Safehold itself owned 83.8% of the outstanding Caret units at December 31, 2025. Because Caret units share in net disposition proceeds rather than only the year-99 reversion, the $200 per unit price is not a market-clearing discount on 99-year reversion claims.

The CARET structure serves two purposes. First, it creates a mechanism for Safehold to share value from its reversion rights with outside capital without waiting for leases to expire. This addresses the core tension in ground lease investing: the best part of the return is locked up for a century. Second, it offers outside investors exposure to long-duration urban land appreciation. The units are held principally by Safehold, and the 10-K notes that public market liquidity was not achieved by February 2024, at which point investors in the February 2022 sale had the right to have their units redeemed.

Why Institutions Don't Own Ground Leases

If ground leases offer contractual rent with 3.4x estimated portfolio ground rent coverage (Safehold's FY2025 10-K) and a modeled 7.6% unlevered IRR on the Figure 1 assumptions (5.0% initial yield on cost, 2.5% rent escalation, 2.5% appreciation), why do so few institutions allocate to them? Several structural barriers explain it:

• Low absolute returns relative to equity real estate. A 7.6% unlevered IRR over 99 years, on the illustrative assumptions above, is attractive on a risk-adjusted basis, but equity real estate allocators often set higher return hurdles. Ground leases sit in an awkward middle ground: too low for equity allocators, too illiquid for fixed-income allocators.
• Duration mismatch. Most institutional investors have 7-10 year hold periods. A 99-year ground lease earns most of its return from escalating ground rent; the reversion adds little at origination (about 0.1 percentage point of IRR on the Figure 1 assumptions) and becomes material only as the remaining term shortens. Selling a ground lease position in the secondary market requires a buyer with the same long-duration thesis, and that buyer pool is thin.
• Lack of institutional expertise. Ground lease underwriting requires specialized knowledge: two-asset valuation, reversion discounting, escalation structure analysis, subordination implications. Most institutional CRE teams lack dedicated ground lease expertise because they encounter these structures infrequently.
• Limited market depth. The investable universe of institutional ground leases is small. Safehold's portfolio is about $6.7B at cost (per its FY2025 10-K), a small fraction of the U.S. commercial real estate market. Funds that need to deploy very large sums into a single strategy can't achieve meaningful allocation to ground leases.
• Accounting complexity. Ground leases create classification questions under ASC 842 (GAAP) and IFRS 16 for both tenants (lessee classification as an operating or finance lease) and lessors, which affects balance sheet presentation and income statement treatment. This adds underwriting and reporting complexity, so check treatment with your accountants before adding ground leases to portfolios that are already difficult to report.

The result: ground leases remain a niche strategy, with Safehold the best-known public platform, despite a theoretical risk-return profile that should attract far more institutional capital.

Ground Lease Funds

Beyond Safehold's publicly traded vehicle, a number of private ground lease investment platforms have emerged:

Safehold (NYSE: SAFE) is the best-known public platform, with $6.7B in ground lease cost as of December 31, 2025. On March 31, 2023, the original Safehold Inc. merged with and into iStar, which continued as the surviving corporation and changed its name to Safehold Inc. Through a subsidiary, Safehold also serves as external manager to Star Holdings, per the FY2025 10-K.

Private ground lease funds have emerged from institutional sponsors targeting a similar risk-return profile. Terms vary by sponsor, so review each vehicle's target portfolio, yield, minimum investment, and hold period against your own horizon.

Family offices and endowments are natural ground lease investors because of their permanent capital base and tolerance for illiquidity. A ground lease purchased at 35% of combined property value with a 5.0% current yield and 99-year reversion matches the investment horizon of a multigenerational family office far better than it matches a 10-year fund.

The ground lease market could grow as more property owners and developers recognize the capital efficiency of separating land from improvements, and as more institutional investors develop the underwriting expertise to evaluate these positions.

Common Mistakes

These errors surface repeatedly when practitioners analyze institutional ground lease positions, either as potential investors, counterparties, or lenders:

• Discounting the reversion at too low a rate. Applying a 5% discount rate to a 99-year reversion overstates its present value dramatically (roughly 6.5x versus a 7% rate). The reversion is speculative at long durations: it depends on assumptions about appreciation, property condition, urban development, and tenant behavior over a century. A common modeling range for 99-year reversions is 7-9% (illustrative, not a market survey), reflecting the compounding uncertainty. Moving from 6% to 8% reduces the PV of a $1.15B reversion from $3.6M to $0.6M.
• Ignoring the improvement reversion. At lease expiration, Safehold receives both the land and the improvements. Some models only value the land reversion, ignoring the building. While the original 2025-vintage building will not exist in 2124, whatever replacement improvements sit on the site will revert. This is additional value, modest in PV terms at origination, but material if the lease is approaching expiration.
• Treating CARET as a simple NPV calculation. CARET units are not simply the NPV of the reversion. They entitle holders to a share of net disposition proceeds above Safehold's investment basis, so value depends on when and how ground lease assets are sold or resolved, not just on the year-99 reversion. Valuing CARET requires optionality analysis, not just DCF.
• Assuming constant appreciation rates. Running a ground lease model with a flat 2.5% annual appreciation rate for 99 years ignores the path dependency of real estate values. Properties go through cycles of value creation and destruction. A Monte Carlo approach with mean-reverting appreciation rates produces more defensible reversion estimates than a deterministic growth assumption.
• Comparing ground lease yields to bond yields directly. Ground leases share characteristics with bonds (contractual income, credit risk, duration) but also with equity real estate (appreciation, reversion, land value participation). A ground lease yielding 5.0% is not directly comparable to a corporate bond yielding 5.0% because the ground lease includes appreciation optionality and reversion rights that the bond lacks.
• Underestimating the tenant's incentive to renew. At year 85 of a 99-year ground lease, the tenant faces a choice: renew on the lease's renewal terms or lose the improvements at expiration. If the building has economic life remaining, the tenant often renews, because walking away means forfeiting the entire building. This renewal dynamic is valuable to the ground lessor (more rent) and should be modeled as a probability-weighted outcome.

How to Model It

Modeling an institutional ground lease position requires a different workbook structure than a standard property acquisition model. The key difference: you are modeling the ground lessor's position, not the property's operating cash flows.

Ground Rent Income Tab

Annual ground rent for the full lease term (99 years). Escalation structure: CPI-linked (project CPI annually or use a long-run average of 2.5-3.0%), fixed (an assumed rate, for example 2.0-2.5%), or FMV resets where the lease has them (model as a step function every 10-20 years, with the reset tied to projected land value). The rent schedule should be buildable year-by-year, with no shortcuts. A 99-year model on the $100M / $35M example with a 5.0% initial yield ($1.75M year-1 rent) and CPI escalations at 2.5% reaches $19.7M in year 99 (1.75 × 1.025⁹⁸).

Reversion Tab

Project the combined property value at lease expiration using an assumed appreciation rate. Run sensitivities on the appreciation rate (1.5%, 2.0%, 2.5%, 3.0%, 3.5%): the range of outcomes is enormous at 99 years. Discount the terminal value back to present using a rate that reflects the uncertainty of the projection. Separately model the renewal scenario: if the tenant renews at year 99, project a new ground rent stream based on the property's fair market land value at that point.

UCA / CARET Tab

Track combined property value and cost basis annually to compute rolling UCA. Apply a discount factor (a percent of UCA) to estimate the CARET value at any point in time. The MSD Partners purchase ($20.0M for 100,000 units) is one data point, but because Caret units share in disposition proceeds rather than only the year-99 reversion, investors should develop their own views on the appropriate factor based on portfolio composition, remaining term distribution, and market conditions.

Returns Tab

Compute unlevered IRR from the ground lessor's perspective: initial investment (cost basis), annual ground rent income, and terminal value (reversion or sale of the position). The IRR should be calculated over multiple hold periods (10, 20, 30, 99 years) because the return profile changes significantly depending on the exit timing. Early exits (10-15 years) depend heavily on the sale price of the ground lease position; long-term holds depend on the rent escalation and reversion.

Related Articles

This article is part of the ground lease underwriting series. Each article covers a specific dimension of ground lease analysis:

• Ground Lease Underwriting: Leasehold vs fee simple valuation and cap rate adjustments.
• Rent Reset Mechanisms: CPI, FMV, and hybrid escalation structures.
• Subordination and Financing: Impact on LTV, loan sizing, and lender requirements.