The Proforma That Never Stops Changing: What Bay Area Developers Are Really Up Against

Real estate development runs on assumptions — about costs, capital, market timing, and demand. The challenge, as one Bay Area multifamily developer laid out plainly in our conversation, is that almost every one of those assumptions has an expiration date. And in today’s environment, that expiration date keeps getting shorter.

Capital is there. But only barely.

On a typical market-rate deal, our interviewee can secure 60–70% LTV from a major lender. On affordable deals, that can stretch to 80%, because lenders view them as lower risk. But that still leaves 25–30% of the capital stack — the equity or mezzanine layer — that has to be raised from private investors.

Right now, that gap is genuinely hard to fill. Equity investors are cautious, and on the affordable side, the public subsidies that traditionally bridge the gap between discounted rents and actual development costs are shrinking fast. The federal government is cutting, state budgets are in deficit, and the tools developers relied on for years are no longer available. “The capital is there,” they said, “but only on the debt side.”

What draws investors in, they explained, is simple in theory but brutally difficult in practice: a good deal, at the right time, in the right market. San Francisco, long written off after the pandemic, is now back in the top three U.S. markets — driven by the AI boom and rising rents. Investors who wouldn’t touch it five years ago are suddenly interested. That cyclical reality means developers are constantly recalibrating what’s feasible, and the proforma is the instrument they’re recalibrating it on.

A living document, not a finished one.

Our developer updates their proformas every one to two months at minimum — sometimes more frequently — across projects that can span five to six years from land acquisition to stabilization. And nearly every line item is in motion.

Hard costs are the most volatile piece. Historically, construction costs have escalated at around 6–7% per year. In the two years after the pandemic, that number spiked even higher — lumber alone saw dramatic swings. The last two years have been an outlier in the other direction: with fewer projects getting built, subcontractors are hungry and GCs are compressing margins to win work, keeping escalation closer to 3%. But no one expects that to last.

The challenge is that GC pricing is only reliable for six months. Beyond that, contractors themselves won’t commit. Permit fees update annually through city-issued master fee schedules. Consultant pricing shifts. Interest rate fluctuations ripple through construction loan sizing, reserve requirements, and total project cost. “I wish I could do it once and have it be the same for four years,” they said. “That’s just not the case.”

The only fixed number, they noted, is the land — paid at close and locked in. Everything else is a moving target.

Prediction is hard. But the window matters.

When we shared that our platform is focused on predicting hard costs — specifically material pricing and construction cost trends — their reaction was immediate and practical: how far out can you go?

It’s the right question. They noted that even GCs, with all their market experience, won’t guarantee pricing beyond six months. Unpredictable inputs — tariffs, pandemic supply shocks, policy shifts like Build America — make anything beyond three to four months deeply uncertain. Our model currently delivers reliable predictions in that three-to-four-month window, which aligns closely with how developers are actually making decisions: in rolling, near-term increments rather than locking in assumptions years out.

That window — even if it’s not a full-cycle forecast — has real value. It gives developers a sharper basis for updating their proformas, a more defensible number to bring to lenders or equity partners, and a faster way to flag when a deal that penciled six months ago no longer does.

Where the opportunity is.

Our interviewee is already using AI to compress the time it takes to process an offering memorandum from four days to two.They are extracting permit fee estimates from city master schedules with a simple prompt. They see clearly where AI saves her time — and they are equally clear about where the hard problems remain.

Predicting hard costs with enough precision to hold up in a proforma, at a time when capital is scarce and every basis point of cost escalation can make or break a deal — that’s the problem still waiting to be solved. That’s where the work is.