Can a data center fund really hold value if the anchor lease burns off before the debt matures

The GP not modeling a re-leasing drag on 18 MW is not an oversight, that is a choice. Six months vacant on a facility that size at $85/kW is somewhere north of $9M in lost revenue and they knew the math before you did.

What I would do from a passive seat is write a single direct question in an email, not a call, something like: what is the operator's current read on renewal probability and has the senior lender been briefed on the year-seven optionality? Email creates a record and forces a written answer, and the quality of that answer tells you almost everything about whether the GP has actually war-gamed this or is hoping the market rate tailwind covers the exposure.

The $110/kW market rate is real but it only closes the gap if lease-up is instantaneous, and data center TI packages for an 18 MW anchor replacement are not small numbers either.

Your math is correct and the assumption doing the most work is the zero-drag re-leasing scenario. A six-month absorption gap on 18 MW at $110/kW/month is roughly $11.9 million in lost revenue before you touch the TI package, and you are right that it compresses year-nine DSCR through the floor. The fact that current market rates are above underwriting does not rescue the coverage ratio if there is any gap at all, it just means the eventual steady-state looks better, which is irrelevant if the lender is testing coverage at the refi date.

The structural issue is a maturity-lease expiry mismatch that the GP should have disclosed and stress-tested in the deck. A nine-year debt maturity sitting two years after a seven-year anchor lease with two optional renewals is not a hidden risk, it is a scheduled one. The GP either modeled it and chose not to show it, or did not model it, and neither answer is comfortable.

The risk you did not mention: TI exposure in data centers is not like office TI. A replacement hyperscale or colocation tenant on 18 MW may require power upgrades, cooling modifications, or connectivity provisioning that can easily run into eight figures depending on what the incoming tenant's spec looks like versus what the departing tenant configured. Your DSCR drop to 0.91x does not include that outlay, and depending on the loan covenants, a TI commitment of that size may require lender consent or trigger a cash trap.

On the GP question, you do not need to frame it as a threat. A direct written request works fine: ask for the sensitivity table on year-nine DSCR under three renewal scenarios, tenant stays, tenant leaves with no gap, and tenant leaves with a six-month re-leasing period plus an estimated TI package. That is a reasonable LP information request and the response tells you a lot. If they have the model, they send the table. If they push back, that is the answer.

What covenants does the senior debt carry around occupancy thresholds or lease notification requirements?

The framing of this as a sequencing problem is right but I think the actual leverage point you have is simpler than you're making it sound. You are a capital partner in a fund with a material structural gap that the GP has not disclosed in twelve months of updates. That is not a question about lease economics, that is a question about whether the GP is running the asset or running the narrative. I had a similar situation on a small industrial fund out of Columbus, not data center scale but same dynamic, anchor lease rolling inside the debt window, and the GP kept publishing occupancy numbers without once mentioning the expiration date sitting two years out. What broke it open was not a threatening letter, it was a direct email asking for the rent roll with lease expiration dates and a copy of the current DSCR covenant reporting to the senior lender. GPs who are managing well send that back in 48 hours. The ones who are managing the story take three weeks and send you a deck instead.

The $110 per kW number is real and I do not want to dismiss it, but data center re-leasing is not like putting a new tenant in a strip center. The fit-out period on 18 MW for a replacement hyperscale or colocation tenant can run 9 to 14 months before you see a dollar, and that is before you negotiate the TI package, which on a facility that size at current construction costs is not a rounding error on a DSCR model. Your 0.91x at year nine is probably generous if there is any overlap with the refi timeline and the lender gets nervous and starts asking for reserves. What is the GP's background, are they operators or financial sponsors who hired an operator? That changes everything about how much they actually see the lease clock ticking.