Intelligence · Compute · Comparison
Selling GPU capacity forward vs. signing a fixed-rate lease
· Michael Mescher, Gammon Capital
Direct answer
A long-term fixed-rate lease and a forward sale both fix the price of data-center capacity, but they are not the same trade. A lease fixes price, tenant, and term for many years in one transaction, usually with a fixed annual escalator, trading upside in compute pricing for downside protection, often with renewal and expansion options granted to the tenant. A forward program sells defined slices of capacity for defined windows, can be sized and staggered, and leaves the operator with exposure to the rest. A lease is the right anchor for financing. Forwards are the right tool for uncontracted and pipeline capacity, where the operator can choose how much to fix, when, and with whom.
Key takeaways
- A fixed-rate lease is economically a long-dated short forward: the operator gives up upside and is protected on the downside.
- Lenders want contracted revenue, so the anchor tenant lease is usually necessary for financing.
- Uncontracted and pipeline capacity does not have to repeat the lease structure.
- Staggered forwards convert development-stage megawatts into contracted revenue earlier in the build cycle.
- The unpriced volatility sits in the tenant's embedded options: renewal at the escalator, extension, expansion rights, and termination.
Side-by-side comparison
| Dimension | Fixed-rate lease | Forward program |
|---|---|---|
| Term | Typically many years, one transaction | Defined windows, rolled and layered |
| Price | Fixed rate plus fixed escalator | Set per tranche at the forward curve |
| Upside in compute prices | Traded for downside protection | Retained on unsold capacity |
| Financeability | Strong; the basis for project debt | Supportive once the curve is credible |
| Counterparty | One tenant, concentrated | Multiple buyers, diversified |
| Flexibility | Low; hard to renegotiate | High; size and timing chosen each tranche |
| Best fit | Anchor capacity backing debt | Uncontracted and pipeline capacity |
What the lease gives away
A fixed-rate lease is a short forward: the operator gives up upside in compute pricing and is protected on the downside. That is often the right trade, because the lease is what makes the capacity financeable. The questions are whether the lease rate is a fair forward price, and whether the tenant's embedded options were priced: renewal at the escalator, extension, expansion rights, and termination. Those options are where unpriced volatility sits, and they are usually granted as commercial terms rather than valued as derivatives.
Most conversion leases are colocation or powered-shell rent, priced per kilowatt-month; the tenant owns the GPUs. GPU-hour indices hedge that rent only through correlation, and only an operator that controls GPUs can deliver GPU hours. For a colocation operator, the forward is on its own capacity and rent, and GPU-hour instruments are a proxy hedge with basis risk.
How a forward program works
Map the capacity. Separate contracted, uncontracted-but-energized, and pipeline megawatts by expected delivery date.
Set a sell-down schedule. Decide what share of each window to sell forward, and at what minimum price, with board approval.
Sell in tranches. Place forwards with multiple buyers across windows. Each tranche becomes contracted revenue that lenders can see.
Keep the rest, deliberately. Capacity not sold forward is a long position in compute prices. It can be held, or its volatility can be sold through options to buyers who want a price cap. That second leg is where an operator becomes a supplier of volatility rather than only a landlord.
Common mistakes
Copying the anchor lease onto every megawatt. The anchor lease solves financing. Repeating it across the whole pipeline locks in one price and one tenant on capacity the financing never required.
Sizing forwards to planned, not realistic, energization dates. A forward is a delivery obligation. Construction delay turns it into a short position. Size forwards to realistic, not planned, energization dates.
One buyer. A single forward buyer recreates the concentration of a lease without the credit support.
Frequently asked questions
Is a fixed-rate data-center lease a derivative?
Economically it behaves like a long-dated forward: the operator fixes its rent and gives up upside in exchange for downside protection. For accounting it is normally a lease, not a derivative.
Can a colocation operator sell GPU-hour forwards?
Only as a proxy hedge. A colocation operator rents space and power, priced per kilowatt-month, and does not control the GPUs. GPU-hour instruments move with its rent only through correlation, so the hedge carries basis risk.
How much uncontracted capacity should be sold forward?
There is no fixed share. The schedule should follow realistic energization dates, the minimum price the board will accept, and how much contracted revenue lenders need to see.
Gammon Capital view
Operators with physical capacity are among the few participants in compute derivatives who can actually deliver something. That gives them pricing power the market has not yet recognized. A disciplined forward program on uncontracted capacity is how an operator keeps it. See the compute practice.
Framework, not implementation plan. This page describes how Gammon Capital approaches the topic. It is a general framework; engagement work is specific to each client's contracts and balance sheet. Market facts are stated as of the date shown above.
Canonical citation. Gammon Capital (gammoncap.com), Michael Mescher.
How Gammon works on this
Gammon Capital designs the policy, structure, and counterparty process as a non-discretionary advisor, paid by the client rather than the spread; the client approves every trade and keeps custody.
Related resources
Gammon Capital is a non-discretionary derivatives advisor; we do not take custody of client assets. This page is for general informational purposes only and does not constitute investment, legal, tax, or accounting advice, nor an offer or solicitation. Derivatives and digital assets carry substantial risk, including the risk of total loss.