← back to ranking

DASH · DoorDash, Inc.

Internet Content & Information · mkt cap $66.2B · calls: Q1 FY2026 vs Q4 FY2025
47.0 conviction · conf-adj 47

conf 2/10 partial

enthusiasm:24.0 · trend:8 · quantifies:0 · impact:0 · under_radar:5 · credibility:5 · business_impact:8 · disruption:-6 · commitment:0 · confirmation:3

Enthusiasm latest 8 / prev 7 (rising)

DoorDash frames AI as defense of the full stack—agentic discovery only matters if proprietary physical-world catalog and fulfillment win the end-to-end job, echoing prior top-of-funnel lessons (Google Food Ordering, Amazon search). Credibility improved sequentially: Q4 cited coding-agent adoption and Smart Campaigns as an "agent"; Q1 added ~two-thirds AI-written code, merchant/support agents with stated P&L benefit, and a sharper autonomy (Dot) scaling narrative—still light on revenue or margin dollars tied explicitly to AI.

GROUNDED NEXT-FY IMPACT vs CONSENSUS

Grounded on actual base — revenue $13.7B · net income $0.9B · net margin 6.8% · diluted EPS 2.13

These are next-fiscal-year annual uplift estimates, not next-quarter numbers.

Aggregate next-FY est. rev uplift: % · next-FY EPS uplift: % · vs analysts: unclear · priced in: medium (model's call-read: high; verdict above is the hard-data one used for ranking) · confidence: 2/10

ClaimFigureArithmeticNext-FY Rev %Next-FY EPS %
~2/3 of code written by AI
productivity · soft		well north of half; ~2/3 of code AI-writtenEngineering-productivity input, not a dollar figure. No AI-attributable engineering/R&D cost base is disclosed (financials give revenue 13,717M, NI 935M, op income 723M only — no eng headcount or cost base), and management leaves the P&L benefit unquantified. No cost base obtainable from any claim, so no $ saving = (undisclosed base) x ? is derivable → rev/eps not computable.
90%+ engineer daily active use of coding agents
productivity · soft		90-plus percent DAUAdoption/engagement metric for internal tooling. No revenue line or cost saving attached; management ties it to 'made them productive' but gives no FTE count, salary base, or output-uplift %. No anchored base anywhere in the claims → no saving_$ = usage × base; no path to rev or EPS.
OpEx ~2% of GOV (near-term/2026 model)
cost · soft		roughly 2% of GOVA steady-state target ratio, not an AI-specific delta. GOV (Gross Order Value) is not provided anywhere in the inputs, no prior OpEx/GOV % is given, and no AI-attributable change in the ratio is stated → cannot compute opex_$ = 0.02 × GOV or after_tax_saving. Stays unanchored.
Already seeing benefits to P&L from AI work
other · soft		benefits to the P&L (unspecified)Explicitly unquantified ('benefits ... from some of the AI work'). No magnitude, no base. Per vague/unanchored rule → pcts null, soft=true. Do not invent a number.

Assumptions: None of the four quantified claims carries a dollar amount OR a percentage with a recoverable base: engineering/R&D cost is not broken out in the financials, and GOV (the denominator for the 2%-of-GOV claim) is not disclosed in any claim. So every claim is unanchored and no aggregation is possible. Earnings basis: consensus uses a lightly-adjusted EPS (~2.25 vs 2.13 GAAP; NI ~993M vs 935M) — would have sized any EPS% off the ~993M/2.25 base; default 21% tax and incremental net margin = current ~6.8% (935/13,717) would apply if any dollar figure existed. GAAP NI/EPS not distorted. All claims are adopter-side (DoorDash improving its own engineering throughput, support, ordering UX, opex efficiency); none sell AI capacity/compute, so there is no supplier-side revenue. Forward agentic/AV statements excluded as unquantified.

Top line: No quantified topline AI impact is computable. The forward AI statements (agentic ordering, agentic customer support, catalog/agentic-commerce advantage, AV orchestration) are directional product bets with no revenue figure, conversion ratio, or take-rate attached. Nothing to add to the 13,717M base. Consensus already models +28.55% FY26 revenue (17,633.6M vs 13,717M = +3,916.6M); that path is analyst-driven, not derivable from mgmt AI metrics.

Bottom line: Zero hard EPS uplift from quantified claims: the bottom-line claims are real in direction (~2/3 code AI-written, 90%+ engineer agent usage, opex targeted at ~2% of GOV, 'already seeing benefits to the P&L') but every one is unanchored — no FTE/eng-cost base to dollarize coding-productivity gains, and GOV undisclosed. With NI 935M (GAAP)/~993M (consensus), any guessed saving would be undisclosed-base, so eps_uplift is left null rather than invented. Consensus FY26 NI +18.64% vs base; FY27 EPS 4.47 vs FY26 2.56 implies net margin climbing ~7% to ~9.2% with no mgmt $ tie to AI.

[impact n/m (all claims soft/unanchored)] No measurable AI gap to compare — all four claims are soft, so adopter aggregate est_rev/eps_uplift_pct are null and ahead/inline/behind cannot be tested on AI math alone. Consensus already bakes in strong leverage: FY26 revenue +28.55% and EPS +14% (2.56 vs 2.25), then FY27 EPS +74% (4.47 vs 2.56) — net margin expanding from ~7% to ~9.2% by FY27. Management's 'already seeing P&L benefits' and 2%-of-GOV opex discipline are consistent with that embedded margin expansion, so the (unquantified) AI efficiency most plausibly already sits inside the curve rather than above it. Because no AI claim is dollar-sized, the position is genuinely indeterminate, but the efficiency/margin narrative appears largely captured in the FY27 EPS ramp.

MODEL CONSENSUS (impact)

partial

Near-identical answers: all four claims soft/null, adopter-only, est uplifts null, verdict unclear. Only priced_in differed (high vs medium); chose more conservative high.

Conflicts reconciled
FieldOpus 4.8GPT-5.5
Rev uplift %––
EPS uplift %––
Priced inmedium–
vs analystsunclear–
Confidence2–
Top lineNo quantified topline AI impact is computable. The forward AI statements (agentic ordering, agentic customer support, catalog/agentic-commerce advantage, AV orchestration) are directional product bets with no revenue figure, conversion ratio, or take-rate attached. Nothing to add to the 13,717M base.–
Bottom lineThe bottom-line claims are real in direction — ~2/3 of code AI-written, 90%+ engineer agent usage, opex targeted at ~2% of GOV, and 'already seeing benefits to the P&L' — but every one is unanchored: no FTE/eng-cost base is disclosed to dollarize the coding-productivity gains, and GOV (the 2%-of-GOV denominator) is not given. With NI at 935M (GAAP) / ~993M (consensus), even a guessed saving would be undisclosed-base, so eps_uplift is left null rather than invented.–
ReasoningNo measurable AI gap to compare. Consensus already bakes in strong leverage: FY26 revenue +28% (17,633M vs 13,768M) on EPS +14% (2.56 vs 2.25), then FY27 EPS +74% (4.47 vs 2.56) on revenue +21% — i.e. consensus assumes net margin climbing from ~7.2% (993M/13.77B) to ~9.2% (1,961M/21.25B) by FY27. Management's 'already seeing P&L benefits' and the 2%-of-GOV opex discipline are consistent with that embedded margin expansion, so the (unquantified) AI efficiency most plausibly sits inside the curve rather than above it — hence medium, not low. But because no AI claim is dollar-sized, the position is genuinely indeterminate.–

Rows highlighted where the two models disagreed.

QUANTIFICATIONS
Share of code written by AI: well north of half; probably closer to 2/3 (today (Q1 FY2026), bottomline)
“about well north of half of our code, as an example, probably closer to 2/3 of our code is written by AI today.”
Engineer daily active usage of coding agents: 90-plus percent (current (Q4 FY2025), bottomline)
“we see 90-plus percent daily active usage, something like that across all of our engineers when it comes to these coding agents, which certainly has made them productive.”
OpEx as % of GOV (in response to AI/productivity/headcount question): roughly 2% (near-term model / 2026, bottomline)
“Look, I would expect from a near-term model perspective for OpEx to roughly be in the 2% range that I've talked about before.”
P&L impact from AI initiatives (qualitative, no dollar amount): benefits to the P&L (unspecified magnitude) (current (Q1 FY2026), bottomline)
“we are already seeing benefits to the P&L from some of the AI work that we're doing.”
PAST (realized)
CURRENT (now)
FORWARD (guidance)
TRACK RECORD — PROMISE vs DELIVERY

68/100 track record   delivers  6 calls reviewed

DoorDash makes very few hard, quantified AI promises—mostly 2026 autonomy commercialization and multi-hundred-million AI-native platform spend—and generally avoids numeric AI ROI targets. What is judgeable so far looks partially on track (limited Dot launches, replatform execution underway) with no clear beats on AI KPIs and one soft walk-back on near-term LLM product milestones.

Ready to commercialize autonomous delivery (DoorDash Dot/platform) in 2026 — promised Q3 FY2025
partial By Q1 FY2026 management reported Dot launched in a couple of markets with good early results and scaling/manufacturing work underway, but framed the program as still early rather than broad commercial rollout.
Invest several hundred million dollars in AI-native global tech platform rebuild in 2026 — promised Q3 FY2025
partial In Q1 FY2026 Ravi confirmed the spend quantum was unchanged versus prior guidance, domain mapping was complete, production traffic was moving onto the new stack, and early velocity/cost benefits were starting to appear.
Majority of global tech-platform (AI-native) investment concentrated in 2026, with tail-end engineering-efficiency gains — promised Q3 FY2025
too-early Through Q1 FY2026 the replatform was still in execution with parallel-stack costs and only early benefits cited; full efficiency payoff was not yet demonstrated in the transcript set.
Ship LLM-driven consumer-product iterations (search/personalization/ranking) over the next 6–12 months — promised Q2 FY2025
quietly-dropped Later calls cited personalization and product improvements qualitatively but never reported discrete 6–12 month AI UX metrics or confirmed a defined launch milestone.
Engineering productivity from LLMs/coding assistants to keep increasing after several years of use — promised Q2 FY2025
too-early Q3 FY2025–Q1 FY2026 tied replatforming to future engineering capacity and early gains, but no quantified productivity uplift was ever reported.
Agentic ordering experiences over the next 12–24 months — promised Q1 FY2026
too-early Management affirmed agentic ordering is coming and discussed discovery/support use cases, but the 12–24 month window had just started and no delivery metrics were available in this transcript set.
PRICED-IN (REFINED)
MEDIUM

Est. revisions flat  ·  Fwd P/E 67.6  ·  EV/Sales 4.4x

AI claim maps to Reportable Segment, UNITED STATES

Analyst ratings are stable (35 buys, 9-10 holds for six months) and recent price-target averages are drifting lower (265 vs 272.58 over the past year), so estimate-revision momentum is flat, not rising. Valuation is rich on hard multiples (67.6x next-FY EPS, 4.4x EV/Sales) and consensus already embeds strong growth (FY26 revenue +28%, FY27 EPS +75%), which prices in much of the AI efficiency narrative. AI-driven gains would most plausibly flow through the consolidated Reportable Segment, dominated by UNITED STATES revenue (~84%). Mixed signals—stretched multiples plus flat/falling revisions rather than rising estimates on a premium—support a medium priced-in verdict.
COVERAGE — ENTHUSIASM TRAJECTORY + CATALYSTS
3Q4 FY20244Q1 FY20257Q2 FY20258Q3 FY20257Q4 FY20258Q1 FY2026

AI enthusiasm across 6 calls — trend ↗ rising

Silent early; then search/ops AI, AI-native platform, autonomy, and agentic commerce with catalog moat.

RECENT AI CATALYSTS & NEWS
BUSINESS IMPACT - QUALITATIVE MATERIALITY

7/10 qualitative impact   material  medium-term · mixed evidence

Where AI matters: agentic discovery, physical-world catalog, and ops productivity

Real deployment (~2/3 AI-written code, 90%+ engineer agent use, merchant/support agents, Smart Campaigns) with claimed P&L help, but no dollar-sized revenue or margin uplift and agentic ordering remains a forward bet.

Caveats: Agentic intermediaries may capture discovery and compress take rates; AI efficiency gains are unquantified and may already be embedded in consensus margin expansion; High code velocity has not yet proven proportionally faster customer outcomes; AI-enabled fraud and support abuse add operational cost

AI DISRUPTION / CANNIBALIZATION RISK  two-sided · 4/10

Personal AI agents could sit between consumers and the marketplace and commoditize DoorDash toward a logistics/API layer, but local fulfillment, retention vs prior top-of-funnel players, and a structured physical-world catalog make the core marketplace durable.

OPTIONS / MARKET STRUCTURE

option liquidity: good

ATM IV
—
TYPICAL BID-ASK
—
OPEN INTEREST
—

proxy inputs — dollar-ADV $680M · beta 1.871 · px $152.32

source: proxy (no options chain on FMP)
FMP /stable/ exposes no options-chain endpoint on this key, so ATM IV, bid-ask spread and open interest are unavailable. Liquidity below is a PROXY from dollar-ADV, beta and price level (a stand-in for option depth), not measured option-market data.

CONFIRMATION — INSIDERS · 13F · LANGUAGE
Mixed — insiders selling, institutions adding, management language 5/10 measured.
INSIDERS selling 69 open-market sell(s) vs 0 buy(s) — net distribution
INSTITUTIONS (13F) adding as of 2026-03-31: 133 new / 209 closed positions; 526 increased / 390 reduced; institutional ownership -3.14pp; -83 net 13F holders
MGMT LANGUAGE 5/10 measured Agentic AI is discussed at length but mostly as product vision; firm will/see language is offset by if/then, I think, and no metrics.
commit “I think that we absolutely will have agentic ordering experiences”
commit “you will see all of these things play out in the DoorDash product experience”
commit “we're not just trying to build aigentic ordering experiences on DoorDash”
VERBATIM AI QUOTES
“I think that we absolutely will have agentic ordering experiences in which it will be a lot easier for customers to do many things that they do today with much lower friction to discover things that they perhaps didn't know, existed on DoorDash to formulate complicated queries and solve those in the best possible way.”
— Tony Xu, Q1 FY2026
“customer support, which, I think, also is having an agentic revolution in and of itself.”
— Tony Xu, Q1 FY2026
“especially in the world of agentic commerce.”
— Tony Xu, Q1 FY2026
“we're not just trying to build aigentic ordering experiences on DoorDash to make the discovery or the search experience easier, kind of echoing what I said in the previous question, but we're also building a catalog, a digital catalog of structured information for the physical world, collecting where every banana sits or every ripe or unripe avocado to every size shoe in whatever color and style that a customer is looking for.”
— Tony Xu, Q1 FY2026
“about well north of half of our code, as an example, probably closer to 2/3 of our code is written by AI today.”
— Tony Xu, Q1 FY2026
“The second priority is to make sure that I think everyone in the company, not just the engineers, but everyone in the company, I think, is as AI capable as anyone else.”
— Tony Xu, Q1 FY2026
“a lot of what I found AI to be helpful, especially now with more powerful models that can reason in a multi-turn kind of fashion is that you can start looking at repetitive processes that kind of are stitched together and actually get them done with perfect quality every single time, and you can actually do that with just an agent.”
— Tony Xu, Q1 FY2026
“Some of it is on our own products like the AI ordering agent stuff I mentioned on a separate question and some of it is on tools that you're talking about, whether it's related to merchants or customer support or Dashers. So we're already seeing that. And with respect to like things we still have to do, you were mentioning about like all the inventory inside of a city.”
— Tony Xu, Q1 FY2026
“we are already seeing benefits to the P&L from some of the AI work that we're doing.”
— Tony Xu, Q1 FY2026
“the vision for us is we are building an autonomous delivery platform because ultimately we think different formats are needed for different types of deliveries.”
— Ravi Inukonda, Q1 FY2026
“this year for us, it's really climbing that curve for the autonomy program and making sure that we can harden our -- I mean, it's not just the autonomy, it's the autonomy, the hardware, the remote operations, all the work on regulatory with the different cities so that we can do this at scale and truly be, again, best-of-breed.”
— Tony Xu, Q1 FY2026
“ads are just a means to connect consumers with merchants who are hoping to be discovered and making sure that you do that in the perfect possible way. So with respect to agentic commerce, I mean, that's just one way of shopping.”
— Tony Xu, Q1 FY2026
“when it comes to any new technology, whether it's autonomous vehicles or agentic commerce, and kind of how that interacts with LLMs, it's really how well does it solve the end-to-end job for a customer, kind of becomes the lens in which I approach all of these things.”
— Tony Xu, Q4 FY2025
“You got to map the physical world, all of which that information does not exist anywhere on the Internet. That's data that DoorDash has to collect in a proprietary way.”
— Tony Xu, Q4 FY2025
“the autonomous delivery platform is probably the most valuable part of what we're building.”
— Tony Xu, Q4 FY2025
“we actually have real live deliveries happening right now with AVs and we're very excited about the future.”
— Tony Xu, Q4 FY2025
“one of the biggest agents that actually we've shipped last year was really our Smart Campaigns product in which we are helping restaurants buy on their behalf, always ROI positive ad campaigns effectively.”
— Tony Xu, Q4 FY2025
“we see 90-plus percent daily active usage, something like that across all of our engineers when it comes to these coding agents, which certainly has made them productive.”
— Tony Xu, Q4 FY2025
“there's also a lot that we're also doing in cataloging all of the physical information that exists nowhere on the Internet that we're also partnering with Dashers to do as well.”
— Tony Xu, Q4 FY2025
ANALYST QUESTIONS ON AI
Q (Q1 FY2026, Shweta Khajuria (Wolfe Research)): Could you please talk about how you envision your product develop over the next 12 to 24 months as you integrate more of agentic and AI capabilities... voice... cart together... execute a transaction... better search and discovery?
A: Tony Xu: Agrees on agentic discovery/search; will have agentic ordering with lower friction and complicated queries; stresses end-to-end fulfillment (stock, personalization), catalog accuracy/selection, speed, and agentic customer support; cites proprietary physical-world catalog as advantage in agentic commerce. (Lyft/travel partnership question was not AI-related.)
Q (Q1 FY2026, Michael Morton (MoffettNathanson)): As AI platforms become more capable, concern that personal agents could sit between marketplaces and consumers — risk of becoming an API/logistics layer; strategic view and willingness to work with third-party AI platforms?
A: Tony Xu: Uses Google Food Ordering and Amazon product-search history — top-of-funnel traffic did not retain vs. DoorDash; customers judge end-to-end outcomes; building agentic ordering plus proprietary structured catalog of the physical world; top-of-funnel/agent players may be small-traffic partners by choice. Ravi/Tony on Dot: building autonomous delivery platform (land and air), early scaling; benefits speed, quality, range; Tony adds 2026 is about hardening autonomy, hardware, remote ops, regulatory work at scale.
Q (Q1 FY2026, Nikhil Devnani (Bernstein)): In a world with AI workloads and a more productive workforce, is your mental model for headcount growth and organizational structure changing?
A: Tony Xu: Yes — seeing productivity gains; ~2/3 of code written by AI but unclear how workflows/teams should change; prioritizing single tech stack then company-wide AI capability before workflow redesign; shipping faster but still proving faster customer outcomes. Ravi Inukonda: AI used across the board; channel productivity into more features; near-term OpEx still ~2% of GOV, disciplined leverage over time.
Q (Q1 FY2026, Brian Nowak (Morgan Stanley)): New tools streamlining merchant onboarding and inventory/catalog (bananas, avocados) — progress and remaining technological advancements?
A: Tony Xu: Multi-turn reasoning models enable agents to complete repetitive onboarding/catalog/photo/metadata tasks with consistent quality; friction removal increases activity; already seeing P&L benefits from AI (ordering agent, merchant tools, support, Dasher tools); still must document and structure messy, changing in-store inventory across channels.
Q (Q1 FY2026, Justin Post (Bank of America)): How do you think about integrating advertising with agentic capabilities on your own platform, and generating ad revenues on agentic platforms elsewhere?
A: Tony Xu: Ads connect merchants to discovery; agentic commerce is one shopping form, won't block advertising and may expand service-area opportunities; ideal agentic UX won't be chat-only; for third-party agentic ad revenue, "you'll have to ask them."
Q (Q4 FY2025, Nikhil Devnani (Bernstein)): Agentic commerce longer term — risk search/discovery/transaction compress; economics may differ; integrate with third parties vs. own vertical solutions?
A: Tony Xu: Judges tech by end-to-end job (Amazon vs. search, Google Food Ordering vs. DoorDash retention); DoorDash positioned via physical-world mapping, operations execution, metadata/personalization; will treat AI systems as channel partners like Facebook/Google historically.
Q (Q4 FY2025, Deepak Mathivanan (Cantor Fitzgerald)): Strategy for autonomous delivery platform in 2–3 years (Dot vs. third-party partnerships)?
A: Tony Xu: Future fleet mix (land/air, build + partner); orchestration and handoffs between Dashers and AVs matter most; 2026 goal is pragmatic use cases; live AV deliveries in a couple of markets.
Q (Q4 FY2025, Justin Patterson (KeyBanc)): Evolving ads product; separately, efficiencies from agentic coding and fit with replatforming?
A: Tony Xu: Ads growing fast; Smart Campaigns described as an agent buying ROI-positive campaigns; grocery/retail ads earlier stage. On coding agents: 90%+ engineer daily active usage, productivity up, still figuring out sustained org design amid rapid change.
Q (Q4 FY2025, Bernard McTernan (Needham)): Will delivery AV use cases be broader than robotaxis (e.g., suburbs vs. dense cities only)?
A: Tony Xu: Yes — Dot built for suburbs (form factor, assignment logic); drones for more rural/long-distance cases.