Darden trades like a tech company. It shouldn't.

It’s a Friday morning, the kind where Thursday’s print is already a day old and the only thing on my second screen is the DRI tape from yesterday’s close — $204.42, up something like 5% off the open, settled in by the bell. The Q3 release is open in one tab, the transcript-adjacent recap on MarketBeat is open in another, and the investor-relations announcement with the updated FY26 outlook is pinned. The flat white on my desk is the second one. The first one is in the trash because I forgot about it while I was doing the math I’m about to do here.

Here is the math, and here is the trade.

Darden guided FY26 adjusted diluted EPS to $10.57–$10.67 — call it $10.62 at the midpoint. At Thursday’s close of $204.42, that’s ~19.2× forward. That is not a casual-dining multiple. That is a multiple the market reserves for operators it believes have a structural re-rate coming — a platform story, a margin story, a software story. Something that justifies paying a premium today for a cash-flow stream that, in the absence of the story, would clear at 14–15× and nobody would blink.

The contrarian read — and this is INTERPRETATION, marked as such — is that the market is paying the wrong premium for the right reason. Darden deserves a premium. It does not deserve an AI premium. And if you are underwriting it as the latter, you are paying twice for something the company is not selling.

What 19× forward is actually pricing

Casual-dining peers — Texas Roadhouse, Brinker, Cracker Barrel, the rest of the sit-down book — have spent the last eighteen months trading in a band that runs from “punished” (Cracker Barrel) to “deservedly bid” (Texas Roadhouse). The high end of that band, on a clean execution story with no concept-side problems, has been roughly 16–18× forward. Darden is now north of that, on a guide that, by management’s own framing in the Yahoo Finance recap of the quarter, assumes nothing heroic — same-restaurant sales in the previously communicated range, commodity inflation continuing to be manageable, no margin step-function.

So what is the extra two-to-four turns of multiple paying for?

I see three possibilities, in roughly the order an LP would walk through them.

The first is execution and scale durability — the boring, correct answer. Darden’s Q3 print, which Luca covered yesterday, was +4.2% blended same-restaurant sales against an industry running roughly -1.2% on the Black Box index. That is 540 basis points of outperformance, in a quarter when nobody else in the segment cleared zero except Texas Roadhouse. Five-forty bps is not a one-quarter accident. It is the cumulative output of a three-year operating program: retention at record levels, portion architecture refreshed without discounting the check, a culinary recertification at LongHorn, and a portfolio where every brand grew segment profit dollars.

The second is capital-return mechanics — Darden bought back $127M of stock in Q3 alone, and the FY26 outlook implies the program continues. EPS denominator math at this pace is real. A buyback-driven re-rate is legitimate, and at ~19× it explains maybe one of the four turns.

The third is AI optionality — the assumption that Darden’s tech stack is going to deliver an unannounced margin step somewhere in FY27 or FY28. This is the one that is wrong, and it is the one I want to spend the rest of the column on.

The execution-and-scale thesis, marked as such

Before I knock the AI leg over, let me make the bull case on the legs that actually hold weight.

Darden’s moat — and this is INTERPRETATION — has three components, and AI is not one of them. The first is retention, which Rick Cardenas named on the Q3 call as a record. In casual dining, hourly turnover is the single largest controllable cost-driver after food, and retention compounds: a server in year three runs a five-top better than a server in week three, full stop. The second is portfolio, where Olive Garden, LongHorn, Yard House, Cheddar’s, and the rest of the book give Darden a price-point spread that no single-concept operator can replicate. The third is supply chain — Darden’s protein purchasing alone moves enough volume that it can absorb commodity shocks the smaller operators have to pass through to the guest, and the guest does not like that.

Those three things, compounded over a decade, are why Darden’s Q3 beat the industry by 540 bps. The 19× multiple, to the extent it is paying for those three things, is paying for the right asset.

The problem is the part that is not paying for those three things.

The tech stack walk — where AI actually shows up

Here is what Darden’s tech stack is, as best I can reconstruct from public material and my own conversations with operators in the space. This is interpretation in the sense that Darden does not break it out cleanly; it is fact in the sense that the components are well-known to anyone covering the segment.

For off-premises, Darden runs on Olo. This is not a secret. Olo’s brand-network deck has historically named Darden’s brands among its enterprise customers, and the integration is the same one most large casual-dining operators have — digital ordering, channel management, dispatch handoff to the third-party aggregators where the brand decides delivery is worth it. Olo is a competent off-premises stack. It is not a differentiated one. Brinker has Olo. Cracker Barrel has Olo. Half the segment has Olo.

For on-premises, Darden runs in-house mobile and digital — guest-facing apps, payment, loyalty plumbing — which it has built and iterated over years rather than outsourced to a marquee vendor. This is, again, competent. It is also, again, not a differentiated story. The Olive Garden app is fine. The LongHorn app is fine. Nobody has ever said “I went to Olive Garden because of the app.”

For labor, Darden runs internal scheduling and demand-forecasting tooling — the kind of system that most segment leaders have at this scale, often built on top of a workforce-management vendor with a forecasting layer on top. This is where AI actually shows up in the Darden P&L, and the place where it earns its keep. Better demand forecasting on a Saturday-night Olive Garden push means fewer over-staffed shifts, fewer under-staffed ones, less overtime. Call it 30–50 basis points of margin assist at maturity, across the book. Real money — and not a re-rate.

For kitchen operations, voice ordering, drive-thru AI — the things the AI-premium narrative actually trades on — Darden has named nothing. Cardenas spent zero seconds of Thursday’s call pitching a vendor, a deployment, or a roadmap. The absence is the data.

Why the absence is the data, and the case against the AI premium

I have been writing variations of this argument for six months. The cleanest articulation of the general thesis is the forthcoming buy-side note Oliver has been drafting on the AI premium, which applies the case against paying for AI optionality to indie acquirers and the strategic buyers; the version I want to write today is the public-market version, and it is narrower.

The narrow version is this: when retention is at a record, same-restaurant sales beat the industry by 540 bps, and segment profit dollars grow in every brand, AI is, at best, a margin assist — not a sales engine. Darden’s Q3 was not won by a forecasting model. It was won by retention, portion architecture, and a recertification program. The lift those things produced is already in the comp number. It is already in the guide. There is no second bite. There is no AI margin step waiting on the other side of FY26 that is not already in the $10.62 midpoint, because if there were, Cardenas would have said so. He did not. He is too disciplined to leak.

Compare this to the operators where the AI premium is, arguably, defensible. Toast trades on a take-rate on a network of small operators where any incremental software lift compounds across thousands of customers. DoorDash trades on a commerce-stack expansion that genuinely changes the unit economics of every restaurant on the platform, which is the case our colleague has been making through the commerce-platform repositioning and adjacent work. Those are stories where AI is a sales engine because it changes the size of the addressable opportunity, not just the cost of serving the existing one.

Darden does not have that shape. Darden’s addressable opportunity is constrained by physical seats, by trade-area density, by the willingness of a guest to drive to an Olive Garden on a Tuesday. AI can help Darden serve the existing demand more profitably. It cannot create more Tuesdays.

Which means the AI leg of the 19× multiple is paying for a re-rate that, mechanically, cannot happen the way the market is imagining it.

The bet, and what would change my mind

If you are long DRI at $204 because you believe Darden is the best-run multi-concept casual-dining operator in the country, that retention will hold, that the portfolio is durable, and that the buyback program will continue to grind the share count — that is a coherent trade, and at 16–17× forward I would not argue with you about it. At 19×, I think you are paying two-to-three turns for something the company is not selling and would, frankly, prefer not to be charged for.

If you are long because you believe Darden has an AI margin step coming that justifies the spread to the casual-dining band — that is the trade I think is wrong, and it is the trade I would not put on at this level.

Mark this: the most telling thing in Thursday’s communication was not what was in it, but what was not. There was no named vendor. There was no roadmap slide. There was no “we are investing in” boilerplate that public-company CEOs deploy when they want the sell side to imagine optionality. Cardenas is too good an operator to pitch a story he is not running, and he is too good a steward to charge his shareholders for one. The absence is the integrity.

So the working line on my desk this morning is this. Darden trades like a tech company because the market does not have a clean place to file an old-line full-service operator that is compounding execution while everything around it stalls. The market is reaching for the AI premium because it is the premium it knows how to underwrite. That does not make the premium correct.

Two things I will be watching, through the Q4 print and into FY27, that bear on whether this column ages well or badly.

The first is whether the next leg of the Darden portfolio — the Bahama Breeze situation Luca walked through on Wednesday and the rest of the small-brand book — gets resolved with concept-level discipline or with a strategic transaction. If it is the former, the execution-and-scale thesis gets stronger and the AI premium gets harder to defend. If it is the latter — if Darden ends up buying or selling a meaningful concept in FY27 — the conversation changes, and the forthcoming Q1 M&A roundup will be where to track it.

The second is whether anyone on the Darden bench introduces a named AI deployment in the next two cycles. My read is the same as Luca’s: they will not, and the absence is deliberate. If I am wrong about that — if Cardenas walks onto a Q4 call and names a kitchen-AI vendor — then either I have misread the operator or the operator has misread his own shareholders, and one of us has to update.

Until then: underwrite Darden as scale plus retention plus buyback, or pass. Do not pay the AI premium on a business that is not running on an AI story. The company is too well-run for the price, and the price is for the wrong reason.

— Marcus edits The Bottom Line for TableTransfers. Tips: [email protected].