Fluidstack Repeatable Campus Economics
Fluidstack
This clause turns one project into a repeatable campus development engine. The 168 MW Abernathy build is not just a single long dated lease, it also gives the venture rights to add more phases on the same site, where transmission, land work, and design are already in place. That is how a target near 70% NOI margins becomes plausible, because each added phase can reuse expensive early infrastructure while rent compounds over a 25 year term with escalators.
-
The economics look like infrastructure, not ordinary cloud resale. TeraWulf disclosed about $9.5B of contracted revenue over 25 years, average annual revenue of roughly $380M, project costs of $8M to $10M per MW, and project level financing backed by Google support for about $1.3B of Fluidstack lease obligations.
-
Subsequent phase rights matter because the first 168 MW already represents the full initial build of the Abernathy campus, and the joint venture can still develop future phases beyond that using existing transmission and development infrastructure. In practice, that means the next megawatt should be cheaper and faster to bring online than the first megawatt.
-
This is also different from the models used by other GPU infrastructure players. CoreWeave has used large long term contracts plus GPU backed financing to add capacity quickly, while Crusoe has pursued more self built energy linked campuses. Fluidstack is pairing long term pre leased demand with partner owned sites, which can support high real estate like margins without taking full campus ownership risk alone.
The next step is a portfolio of copy pasted AI campuses where one signed customer unlocks debt, one powered site unlocks expansion, and each added phase lifts returns on the original development work. If Fluidstack keeps stacking these rights across TeraWulf, Cipher, and Hut 8, it moves from renting GPUs to controlling a growing share of the economics of AI data center buildouts.