Capstone: costs and financing

Harbor Point is earning and nothing has paid for it. This chapter funds the deal, and it is deliberately a chapter of mixed construction — chapter 19's promised pattern of pack contracts beside bespoke claims. The permanent loan is standard commercial paper, so the pack states it. The construction facility — advanced in pieces, accruing on a moving balance, repaid in one takeout — is a storyline the pack has no single word for, and it becomes the course's best worked example of fields, prev, and streams cooperating.

The construction facility, bespoke

// The construction lender funds 65% of each draw; interest accrues on the
// drawn balance and the whole facility is repaid at stabilization.
entity asset cloan : Asset.Financial {
  balance init 0
          next if(time.t <= 17, prev + 455000, if(time.t == 30, 0, prev))
}

stream cloan.advances on entity asset.cloan inflow currency USD {
  schedule every month from 2026-02 to 2027-06
  amount = 455000
}

stream cloan.interest on entity asset.cloan outflow currency USD {
  schedule every month from 2026-03 to 2028-06
  amount = prev.asset.cloan.balance * 0.075 / 12
}

stream cloan.payoff on entity asset.cloan outflow currency USD {
  schedule every month from 2028-07 to 2028-07
  amount = prev.asset.cloan.balance
}

Read it as three chapters of this course meeting. The balance is a rule field (chapter 8) on the loan's own entity — a fact about the facility, growing 455,000 per draw month, zeroed at takeout. Interest reads the balance one period back — prev.asset.cloan.balance — because construction interest bills on the balance you carried into the month; the field is declared once and read at both ends, the average-balance machinery from chapter 8 in its simplest form. The payoff reads the same prev at takeout: whatever the facility closed June 2028 with — 7,735,000, and the model derives that rather than stating it, so a changed draw schedule repays itself correctly without anyone updating a payoff number.

One deliberate simplification to name (style rule 4 — comments state claims): interest here is paid monthly, not capitalized into the balance. A capitalizing facility compounds — next prev * (1 + 0.075/12) + draw — and the end-of-part exercises upgrade to it. Write the version the term sheet states; ours states monthly pay.

The takeout and the permanent loan

// The permanent loan closes as operations begin: interest-only to maturity.
stream harbor.perm_proceeds on entity asset.harbor inflow currency USD {
  schedule every month due from 2028-07 to 2028-07
  amount = 9500000
}

contract cre.permanent_debt on entity asset.harbor {
  term 2028-07..2030-12
  terms {
    principal = 9500000
    rate = 0.058
    amort_months = 360
    io_months = 36
  }
}

The proceeds arrive due — closing cash comes at the month's start, funding the payoff that leaves at its end — and the loan itself is one pack contract. Its terms encode the commercial-mortgage grammar from the pack's design: amort_months strikes the payment on a 30-year schedule while the term sets maturity, and io_months = 36 keeps the whole 30-month hold interest-only — a bridge-to-sale structure, which is why chapter 23's exit must repay principal in full. The balloon stays off by default for the reason the pack states: coverage is measured on periodic debt service, and folding a payoff into the final period's service line would make that month's DSCR meaningless. The payoff belongs to the reversion, where the next chapter puts it.

What the numbers say

Advances total 7,735,000 and the payoff repays exactly 7,735,000 — the model's self-consistency, visible as two matching series. Facility interest totals 966,875; the hand anchor is any single month — March 2026 bills 455,000 × 0.625% = 2,843.75 — and the ramp shape of the interest series is the drawn balance, plotted. Permanent debt service is 9,500,000 × 5.8% / 12 = 45,916.67, thirty times: 1,377,500.

And now the pack's derived metrics earn their keep: stabilized NOI against IO service puts DSCR comfortably above two — this deal's risk was never coverage. Notice what the whole-deal total says: −6.2 million. Financing did not create value — it moved the equity check from 17.4 million to about 8, plus carry. Leverage rearranges who funds and who earns; the deal's worth still rests entirely on the next chapter's exit.

What can go wrong

A balance and a payoff maintained separately. The payoff reads the field; the naive version states 7735000 literally, and the first draw-schedule change breaks their agreement silently. Chapter 17's smell — history restated instead of read — with six figures attached.

Interest on the wrong end of the period. asset.cloan.balance (this month's close) instead of prev.… charges interest on money drawn this month — a subtle overstatement that compounds. The convention is a claim; the term sheet says which end, and the expression says it back.

Proceeds without their due. Takeout proceeds at month-end would leave the payoff unfunded for a month inside the model — a fictional cash hole that a reviewer chasing the minimum-balance question would waste an afternoon on. Placement, again, is money.

Exercises

Then, on your own:

1. Upgrade the facility to capitalized interest: the balance compounds and the interest stream disappears into it (delete the stream; the cost now arrives as a larger payoff). Confirm the payoff grows to about 8.16 million and the deal total barely moves — the cost was real either way; only its path changed.
2. Price the takeout differently: raise the permanent loan to 10,500,000 and watch DSCR and the equity check trade against each other. One term, two consequences, both visible in one run — sizing debt is exactly this loop, done until the committee stops arguing.