A 'Market of Extremes': RFDC Just Made Plain What Q3 Earnings Tried to Hide

I was eating a cold chicken sandwich in a hallway at the Restaurant Finance and Development Conference when a private-equity associate, two name-tags deep into his second espresso, leaned over and asked me what I thought “the number” was. He meant the spread — the gap between the best- and worst-performing public restaurant brands on same-store sales. I guessed twenty points. He smiled like a man who had just heard a child mispronounce a French wine. “Thirty-point-nine,” he said. Then he walked off toward a panel on debt restructuring.

That number — 30.9 points — is the only one that matters from RFDC this year, and it’s the one that operator-AI vendors should staple to the inside of their pitch decks. The headline from the conference, as Restaurant Dive reported, was that the U.S. restaurant industry has become “a market of extremes.” My read is sharper than that: aggregate restaurant data is now actively misleading, and any vendor still selling against an industry-average story is going to learn that lesson the expensive way in 2026.

The 30.9-point spread is a buyer-segmentation problem, not a tailwind

Here is the bifurcation in a single breath. Brinker — Chili’s parent — posted +21.4% same-store growth in Q3. Sweetgreen posted –9.5%. That is not a sector. That is two different economies operating under the same NAICS code. And the macro substrate underneath it is uglier than the operator press releases admit: Moody’s Analytics chief economist Mark Zandi has noted that top earners drove 49.2% of all consumer spending in Q2 2025, the highest share since 1989. Restaurants are downstream of that wealth concentration, and the bifurcation at RFDC is what it looks like when half the dining room is paying with house money and the other half is paying with a tapped-out credit line.

The vendors I spoke with — voice-AI companies, scheduling-AI companies, the entire AI-for-BOH cohort — were almost uniformly pitching “industry recovering” decks. That is the wrong instrument. If your buyer is Brinker, your problem is throughput and labor optimization at peak. If your buyer is Sweetgreen, your problem is unit-economics survival and same-store recovery. Those are different products. They require different ROI math, different pilots, different reference customers. A forthcoming Pass case study on Sweetgreen’s Infinite Kitchen rollout will dig into exactly this asymmetry — what “AI as margin lifeline” looks like versus “AI as growth accelerant” — and I’ll link it here once it runs.

Debt is the part nobody onstage wanted to name

The other thing RFDC kept gesturing at without quite saying out loud: the debt cliff. Fat Brands defaulted in October. Across Q4 so far, more than $1.3 billion in restaurant debt has been declared immediately payable across various credit facilities and bond covenants — a quietly remarkable figure given how thinly the trade press has covered it. Mark interpretation: when distressed-debt people show up to a development conference in numbers, the conference is no longer really about development. It’s about who buys the bones.

For operator-AI vendors, the debt overhang changes the sales motion in a way most decks aren’t built for. Brinker can write a check for a multi-year AI platform. A leveraged mid-cap with covenant pressure cannot — not because the ROI math doesn’t pencil, but because every dollar of operating cash flow is spoken for by the credit agreement. I watched at least three AI vendors at RFDC pitch into rooms full of CFOs who were doing covenant math in their heads while nodding politely. The polite nod is not a buying signal. The polite nod is the sound of a CFO calculating which quarter the lender steps in.

This is also where the AI-premium thesis starts to wobble. Public-market multiples for AI-forward operators have priced in a story about category-wide adoption. A coming Pass piece will dig into the AI-premium counter-thesis in detail, but the short version is: if 60% of the addressable buyer base is debt-constrained for the next four to six quarters, the adoption curve is going to be slower and lumpier than the multiples assume. The 30.9-point spread is the leading indicator. The debt declarations are the lagging one. The middle — where vendors are pitching today — is where the bodies will be.

What I’d actually do

If I ran sales at an operator-AI company right now, I’d do three unglamorous things. One: re-segment the pipeline by balance sheet, not by logo. Two: build two distinct ROI narratives — one for the Brinkers, one for the Sweetgreens — and stop trying to triangulate a “middle” customer who increasingly does not exist. Three: assume the debt cliff bites in the back half of 2026 and design pilot terms that can survive a buyer’s covenant renegotiation.

The market of extremes is not a phase. It is the shape of the next eighteen months. RFDC just said the quiet part out loud, and the vendors who heard it are the ones who’ll still be standing at the next one.

— Luca covers restaurants for TableTransfers. Tips: [email protected].