The 2025 Restaurant Ticker Scoreboard: Chili's, Toast, Hilton — and the Spread That Should Worry Vendor CFOs

I spent the morning of December 26 staring at a half-frozen ticker scoreboard on the second monitor — the kind of post-holiday tape where volume is thin, market makers are skeleton-crewed, and the year’s narrative finally sits still long enough to be measured. The line that kept catching my eye wasn’t a price; it was a spread. Brinker International, parent of Chili’s, closed the quarter with same-store sales up 21.4%. Sweetgreen, the prestige fast-casual that was supposed to be a generational compounder, printed comps down 9.5%. That is a 30.9-point gap between two operators selling food to roughly the same American consumer in roughly the same calendar quarter.

If you sell AI software into hospitality and you treated 2025 as one market, you spent the year underwriting an average that did not exist.

The contrarian thesis I am going to defend in this column is narrow and mechanical: AI-vendor sales pipelines should be segmented by buyer health, not aggregate market growth. Top-of-funnel dashboards that show “restaurant TAM up” are misleading vendor CFOs into planning quotas, headcount, and discounting policy against a blended consumer that has bifurcated into a winners-and-losers tape sharper than anything I have seen since 2009. The 30.9-point Brinker–Sweetgreen spread is the cleanest single data point of the year, but it is not alone. Toast just printed $2.0B in ARR. Hilton is guiding to $4.0–4.04B in 2026 adjusted EBITDA off a record digital year. First Watch is calmly walking into ~$123M in FY25 adjusted EBITDA at the high end of guide. And Fat Brands is calling $1.3B of debt in November and December as the year ends. These data points don’t average. They diverge. Vendors who don’t segment will mis-price the back half of 2026.

Let me walk the scoreboard.

The Brinker–Sweetgreen spread is a regime indicator, not a quirk

Restaurant Dive’s 2025 winners-and-losers wrap is the single most useful sentence-by-sentence read I’ve done this month, and the headline number — Brinker’s Chili’s at +21.4% comp, Sweetgreen at -9.5% comp — deserves more time than the trade press has given it.

Start with the arithmetic. A casual-dining brand with a check around $20 running +21% comps is doing two things simultaneously: it is winning back lapsed customers from full-service competitors and it is taking trade-down traffic from fast-casual. A fast-casual brand with a check around $16 running -9.5% comps in the same quarter is, almost by identity, losing both directions: customers trading down to QSR and customers trading up to a value-coded casual-dining experience that suddenly feels like a bargain. The 30.9-point spread is not just two operators executing differently. It is the consumer telling you the price-value frontier has moved and that “premium fast-casual” — the category that absorbed roughly $14B in venture and growth equity from 2018 through 2023 — is now structurally on the wrong side of it.

Why does this matter to an AI vendor? Because the buyer profile is inverted from where most 2024-vintage pipelines were built. The Sweetgreen-shaped logo — well-funded, modern stack, eager early adopter, willing to pilot vision systems and forecasting models — is the buyer your AEs were trained to chase. The Brinker-shaped logo — 1,100+ units, legacy POS, conservative IT committee, multi-quarter procurement — is the buyer who actually has the cash to deploy in 2026. If your top-of-funnel scoring model still gives “fast-casual, Series D or later” a higher lead score than “publicly traded casual-dining parent with positive operating leverage,” you are going to spend 2026 burning SDR cycles on logos that cannot sign.

A useful sanity check: I made the case against the AI premium in unit-economics modeling in a forthcoming May piece, and the argument there cuts the same way. When the consumer bifurcates this hard, the operators who can fund AI are not the operators who talk about AI. They are the operators whose comps are funding their capex.

Toast at $2.0B ARR is the back-office tell

The other ticker on my board that won’t stop blinking is Toast. Toast’s Q3 release on November 4 put ARR at ~$2.0B, with location count and gross payment volume both compounding at rates that, for a company this size, frankly should not be possible in a flat consumer year. The headline reads as a Toast story. The deeper read is a buyer-health story.

Toast does not grow ARR by 25%+ in a recessionary consumer year unless its installed base is generating enough cash to fund expansion modules — handheld terminals, online ordering, payroll, marketing, capital. That cash is not coming from Sweetgreen-shaped operators; Toast’s mid-market and SMB independents are punching above their weight precisely because the casual-dining and independent operator cohort is the one taking trade-down traffic. The $2.0B ARR number is the back-office signature of the same consumer rotation Brinker’s +21.4% prints on the front-of-house.

For AI vendors, Toast’s ARR run is a warning shot, not a comp. It is a warning shot because Toast is bundling adjacent capability faster than standalone AI point-solutions can integrate. Every quarter Toast adds another attach module (loyalty, marketing, scheduling, payments-adjacent capital), the surface area available to a standalone AI vendor shrinks. The vendors who will survive 2026 are the ones who either (a) sit clearly above the POS layer — demand forecasting, labor optimization, inventory across multi-unit groups — or (b) sit cleanly below it — vision-based drive-thru, kitchen sensors, voice. The middle, where Toast’s roadmap is converging, is going to get crushed regardless of how the broader market grows.

I wrote about the 12-unit cafe group pricing problem in an upcoming Pass piece — that piece argues the per-unit per-month pricing model breaks for any vendor whose value sits inside Toast’s expanding bundle. The Q3 ARR number is the empirical version of that argument.

Hilton is the cleanest example of “AI funded by operating leverage”

Switch boards from restaurants to lodging for a moment. Hilton’s year-end coverage — running ahead of formal Q4 print but consistent with management commentary through the fall — pegs 2025 adjusted EBITDA at $3.7B, up 9%, with 2026 guide at $4.0–4.04B and Hilton Honors membership at 243M, up 15%. The 2030 tech roadmap is real and well-funded.

Here is the math that matters for vendors. Hilton’s 2026 EBITDA guide implies roughly $300–340M of incremental EBITDA year over year. Even if you assume the chain reinvests only 10–15% of that incremental EBITDA into technology and AI initiatives — a conservative number for a company that has publicly committed to a 2030 roadmap — that is $30–50M of incremental annual AI-adjacent budget at a single operator. The Honors growth of +15% to 243M members is the demand-side feedstock that makes any personalization or recommendation AI investment pencil out.

Hilton is not unique. It is just the cleanest publicly observable example of a buyer who can fund AI through 2026 because operating leverage is paying for it. Marriott will look similar. Hyatt will look similar. The independent boutique segment will not. Two- and three-property regional hotel operators, who anchored a lot of 2024 hotel-tech pipeline decks, will not. This is the same bifurcation logic as the Brinker–Sweetgreen tape, just expressed in lodging.

If you are an AI vendor selling into hospitality, your 2026 named-account list should be heavily weighted to operators whose 2026 EBITDA guide is up, not flat. That list is shorter than your 2025 list was. Plan accordingly.

First Watch and the “boring middle” — the segment vendors keep underweighting

I want to spend a paragraph on First Watch because it is the operator type most AI vendor sales orgs systematically under-prospect, and the Q3 commentary is a clean case study.

In First Watch’s Q3 release on November 4, 2025, CEO Chris Tomasso said: “we are pleased to guide to the high end of our previous range for FY25 adjusted EBITDA at approximately $123 million.” You can read the full filing on the SEC’s First Watch Q3 8-K exhibit. That is a publicly traded daytime-only operator, ~570 units, that just told the market it is calmly walking into the year at the high end of its EBITDA guide. No drama. No turnaround narrative. No “transformation” language.

This is exactly the buyer profile that wins disproportionately in 2026, and exactly the buyer profile that AEs trained on 2021-vintage logo decks dismiss. They aren’t featured on conference panels. They don’t post on LinkedIn about AI strategy. They will, however, sign a multi-year contract with a vendor that can demonstrably reduce food cost by 60 basis points or labor cost by 90 basis points. Their procurement is methodical; their CFOs do real diligence; once they sign, they expand. The ROI threshold is concrete and the contracts hold.

If your 2026 pipeline is overweighted to logos that excite Twitter and underweighted to logos like First Watch, your win rate is going to disappoint your board.

Fat Brands and the credit signal: the right tail of the distribution

The flip side of the scoreboard is the credit signal. Fat Brands called $1.3B of debt in November and December — a large, structurally levered, multi-brand franchisor managing through a refinancing cycle that the rest of the sector watched closely. Whatever you think of Fat Brands’ specific situation, the broader read is that the levered-rollup model is in tighter air.

For AI vendors, the credit signal matters for two reasons. First, levered franchisors are not 2026 net-new buyers. They are renewal-risk accounts. If you have any portion of your ARR booked through highly levered multi-brand franchisors, your renewal forecasting model should haircut those accounts more aggressively than gross retention suggests. Second, the credit signal is the leading indicator of M&A volume in 2026. Distressed franchise rollups will trade. Brands will move. Tech stacks will be re-evaluated by new owners. That is opportunity, but it is transactional opportunity, not subscription opportunity, and most vendor revenue models are not built for it.

I covered the M&A scorecard on December 5, and the Fat Brands data point is the back-end confirmation of the front-end thesis there: deal flow in 2026 will be forced, not strategic. Forced deal flow rewards vendors with portable, easily-redeployed tech stacks and punishes vendors with deep custom integration debt. Know which one you are.

What the segmented pipeline actually looks like

Let me put the scoreboard together as a single mental model for vendor planning.

Tier 1 — Fund-the-roadmap buyers. Hilton, Marriott, Brinker, Texas Roadhouse, Chipotle (despite its own 2025 wobbles), Wingstop, Domino’s, large QSR franchisees with positive comps. These operators have 2026 EBITDA guides going up. AI capex is funded by operating leverage. Sales cycles are long but contract values are large and renewals stick. Weight your enterprise AE coverage here.

Tier 2 — Quietly compounding mid-market operators. First Watch, Cava (off Q3 highs but still positive), regional casual-dining chains with 50-300 units running flat to positive comps, well-run independent groups of 8-30 units. Procurement is methodical. Contracts hold. Weight your mid-market sales motion here.

Tier 3 — Bundle-disrupted SMB. Independent restaurants where Toast and its competitors are absorbing the budget. Standalone AI vendors will lose to platform-bundled equivalents at this tier nearly every time. Re-evaluate whether direct sales is the right motion at all — partner-led or marketplace distribution is almost certainly the better economics.

Tier 4 — Avoid-or-distress. Sweetgreen-shaped fast-casual with negative comps, levered franchise rollups in refinancing cycles, single-unit independents in declining trade areas. These accounts will consume sales cycles disproportionate to their contract value or renewal probability. Negative scoring weight in your lead routing.

If your current pipeline does not break out cleanly into those four tiers, your 2026 plan is built on an aggregate market number that doesn’t describe any actual buyer.

The bottom line for vendor CFOs

The single most useful thing a hospitality-AI vendor CFO can do in the first week of January is re-run the 2026 forecast with buyer-health segmentation rather than market growth assumptions. The mechanics are not complicated. Pull your top 200 named accounts. Tag each one against the most recent public same-store-sales print, EBITDA trajectory, and credit posture. Weight pipeline expected-value by tier. Compare to the original board number. The gap between the two is your 2026 forecasting error.

The Brinker–Sweetgreen 30.9-point spread is the cleanest evidence I have seen this year that the hospitality consumer has bifurcated into two distinct economies. Toast’s $2.0B ARR confirms the back-office signature of that bifurcation. Hilton’s $4.0B+ EBITDA guide shows you who can actually fund AI in 2026. First Watch shows you the under-prospected boring middle. Fat Brands shows you the credit-side risk that vendor renewal models are mispricing. Pull these five data points onto a single page, segment your pipeline against them, and you will enter 2026 with a forecast that survives Q1.

Pull none of them, run your 2026 plan against an aggregate “restaurant market up 4%” assumption, and you’ll spend the back half of the year explaining to your board why pipeline coverage was at 3x and bookings still missed.

The tape always tells you. You just have to stop averaging it.

— Oliver writes The Bottom Line for TableTransfers. Tips: ma@tabletransfers.com.