← back to rankingCVNA · Carvana Co.
Auto - Dealerships · mkt cap $71.1B · calls: Q1 FY2026 vs Q4 FY2025
26.0 conviction · conf-adj 26
conf 3/10 partial
enthusiasm:18.0 · trend:-5 · quantifies:0 · impact:0 · under_radar:5 · credibility:0 · business_impact:8 · disruption:0 · commitment:0 · confirmation:0
Enthusiasm latest 6 / prev 8 (falling)
The AI thesis is credible but narrower in Q1 FY2026 than in Q4 FY2025: Carvana moved from explicit AI-brain/Sebastian framing to operational tooling, algorithms, data integrations, automated pricing, and AI-related technology spend. The strongest quantified evidence remains customer self-service automation at 30% of retail buyers and 60% of sellers, plus operational efficiency claims in reconditioning. Management’s enthusiasm is falling because Q4 framed AI as a strategic competitive advantage, while Q1 treated it mostly as one tool inside overhead and reconditioning execution.
GROUNDED NEXT-FY IMPACT vs CONSENSUS
Grounded on actual base — revenue $20.3B · net income $1.4B · net margin 6.9% · diluted EPS 1.69
These are next-fiscal-year annual uplift estimates, not next-quarter numbers.
Aggregate next-FY est. rev uplift: % · next-FY EPS uplift: % · vs analysts: inline · priced in: medium (model's call-read: high; verdict above is the hard-data one used for ranking) · confidence: 3/10
| Claim | Figure | Arithmetic | Next-FY Rev % | Next-FY EPS % |
|---|
30% of retail buyers complete purchase with no human contact
engagement · soft | 30% | Adoption/labor-leverage metric; mgmt disclosed no $ or FTE saving. Cannot convert to revenue or cost without inventing a number against $20.322B rev / $1.407B NI. | | |
60% of car sellers complete sale with no human contact
engagement · soft | 60% | Same: adoption metric for existing seller workflow, no anchored $ saving or incremental supply disclosed. | | |
4.8-day fastest fully-automated retail cycle
productivity · soft | 4.8 days | Single best-case (record) cycle time, not an average; no baseline days, affected volume, or margin figure attached, so revenue/EPS not calculable. | | |
~$200 GPU spread, top vs bottom quartile recon facilities, centralized tools may close
cost · soft | $200/unit | Would be after-tax saving = $200 * affected_units * capture * (1-21%) / $1.407B NI. Y illustratively assumed ~550k retail units (not in base) closing bottom quartile fully: 0.25*550k*$200=$27.5M pre-tax, $21.7M after-tax, ~1.5% EPS. Affected units/capture undisclosed, so no hard figure is defensible. Topline ~0 (cost, not volume). | 0.0 | |
Assumptions: Tax 21%; incremental margin n/a (no incremental revenue claims). The recon-spread EPS sizing depends on an assumed ~550k retail units NOT in the financial base, so it is illustrative only and reported null. All four claims are adopter-side operational/engagement metrics; none are supplier-side AI revenue.
Top line: Effectively zero direct topline from the quantified AI claims — the 30%/60% no-human adoption rates and 4.8-day record cycle are internal efficiency/engagement metrics, not new revenue streams; mgmt attached no incremental revenue, so modeled rev uplift is 0%.
Bottom line: The only dollar-like claim is the ~$200/unit recon GPU spread. Closing the bottom quartile could yield ~$27.5M pre-tax / $21.7M after-tax (~1.5% of $1.407B NI, ~+$0.02 EPS), but this rests on an undisclosed/assumed unit count and capture rate, so no hard EPS uplift is defensible.
[impact n/m (all claims soft/unanchored)] Consensus already models 2026 revenue ~$27.76B (+36% vs $20.32B) and EPS ~$1.51, rising toward ~$2.06 by 2027 — growth driven by volume and ongoing GPU/SG&A efficiency. The quantified AI gains here are either unsizable or, at most, ~1.5% EPS (~$0.02) — exactly the operating-leverage already embedded in consensus margin assumptions; nothing in the calls points the math clearly above what analysts assume.
MODEL CONSENSUS (impact)
partial
All claims adopter-side and soft; no defensible hard revenue/EPS uplift. Recon spread illustratively ~1.5% EPS but unsized, so reported null.
Conflicts reconciled
- recon eps_uplift_pct: X=null vs Y=1.5 -> used null because Y's 1.5% rests on an assumed ~550k retail units not in the base (invented fact); kept Y's calc in basis as illustrative
- est_eps_uplift_pct: X=null vs Y=1.5 -> used null (more conservative; undisclosed units/capture)
- est_rev_uplift_pct: X=null vs Y=0.0 -> used 0.0 since claims create no revenue stream
- supplier_rev_uplift_pct: X=0 vs Y=null -> used 0 (no supplier claims exist)
- vs_analyst_expectations: X=unclear vs Y=inline -> used inline (better justified: gains embedded in consensus leverage)
- claim types: minor X/Y differences -> used Y's (adoption=engagement, cycle-time=productivity)
| Field | Opus 4.8 | GPT-5.5 |
|---|
| Rev uplift % | 0.0 | – |
| EPS uplift % | 1.5 | – |
| Priced in | high | high |
| vs analysts | inline | unclear |
| Confidence | 3 | 3 |
| Top line | Effectively zero direct topline from the quantified AI claims — every figure is an internal efficiency/engagement metric, not a new revenue stream. The 30%/60% no-human adoption rates support volume scalability without proportional headcount, but management attached no incremental revenue, so the modeled rev uplift is 0%. | No quantified revenue uplift is calculable. The 30% buyer automation, 60% seller automation, and 4.8-day fastest-cycle claim describe workflow adoption or anecdotes, not incremental revenue dollars or volume. |
| Bottom line | The only convertible claim is the ~$200/unit recon GPU spread: closing the bottom quartile yields roughly $27.5M pre-tax / $21.7M after-tax, ~1.5% of $1.407B net income, i.e. ~+$0.02 EPS. The no-human-contact metrics plausibly add further SG&A leverage but are unsized, so the defensible bottom-line uplift is ~1.5% and soft. | The only dollar-like claim is a 'couple hundred dollar' reconditioning spread, but without affected units and capture rate the after-tax saving cannot be sized against $1.407B of current net income. |
| Reasoning | Consensus already models 2026 revenue $27.76B (+36% vs $20.32B) and EPS $1.51, rising to $2.06 by 2027 — growth driven by volume and ongoing GPU/SG&A efficiency. The AI gains here (~1.5% EPS, ~$0.02) are small relative to those moves and are exactly the kind of operating-leverage already embedded in consensus margin assumptions; nothing in the calls points the math clearly above what analysts assume. | Consensus FY2026 revenue is $27.757B versus the current $20.322B base, implying +$7.435B or +36.59% revenue growth. Consensus FY2026 EPS is $1.51333 versus current EPS of $1.69, implying -10.45% EPS change, while consensus net income of $1.758B implies +$351.4M or +24.98% net income growth. The AI claims provided do not support a hard uplift above those consensus numbers. |
Rows highlighted where the two models disagreed.
QUANTIFICATIONS
retail customers completing process without talking to a person until vehicle handoff: 30% (Q4 FY2025 call, current at time of call, both)
“We have 30% of our retail customers now go through the entire process without talking to a person until they get the car.”
vehicle sellers completing process without talking to anyone until drop-off: 60% (Q4 FY2025 call, current at time of call, both)
“We have 60% of our customers that are selling cars to us who go through the process without talking to anyone until they drop off their car.”
fastest retail car cycle through automated system: 4.8 days (Q1 FY2026 call, both)
“And another customer finds it, decides they want to buy it, they go through the entire purchase process, schedule their delivery. We put it on a truck, deliver to them, and it's theirs, and that took 4.8 days, which is pretty exciting.”
performance spread between top and bottom quartile reconditioning facilities that centralized planning/tools may address: couple of hundred dollar spread (Q1 FY2026 call, referencing prior quarter, bottomline)
“But I think last quarter, we talked about there being a couple of hundred dollar spread between our top quartile and bottom quartile performers.”
PAST (realized)
- We have 30% of our retail customers now go through the entire process without talking to a person until they get the car.
- We have 60% of our customers that are selling cars to us who go through the process without talking to anyone until they drop off their car.
- And I think a major set of tools that contributes to that is Sebastian and other tools that emerge from that AI brain.
- Those customers who go through the experience in that way have a higher NPS than customers that call us.
CURRENT (now)
- For example, we're making some investments now in additional technology, including AI-related technology that would be in that overhead expense number.
- Over the last couple of months, they built additional data integrations, developed tools to help managers make faster, higher quality decisions, and how they staff their lines, and how they optimize flow through their paint lines and implemented a productivity tracker to ensure feedback reaches the right groups quickly.
- We inspected. We run it through the reconditioning process after figuring out what needs to be fixed on the car, photograph it, put it up on the site, price it in an automated way.
- But on the other hand, it's also -- that's a problem that can be solved with algorithms data and pairing those two things together, very strong quantitative focus via software and via making even better use of all the data that we're collecting in the centers and then pairing that more effectively with the teams on the ground.
FORWARD (guidance)
- I think if you look at other parts of the business, including just the speed at which we're developing new products, that continues to get better all the time.
- But we do believe that we're fundamentally extremely well positioned to benefit from these things because we have a big deterministic system that's vertically integrated that has access to all the information and that brain has every system feeding it so we can give customers very simple answers to any questions they've got in really any software interface that we choose to put on top of it.
- So I think that will continue to be our primary focus for the foreseeable future, but we're always paying attention.
- And then we think as we continue to build out Carli that makes the systems inside Carli that make the individual operators more efficient and as we continue to build out these manager tools that make manager decision-making more straightforward.
TRACK RECORD — PROMISE vs DELIVERY
—/100 (no quantified promises) no-quantified-promises 6 calls reviewed
Carvana is a substantive AI/automation adopter — deterministic no-touch transactions (~30% of retail buyers and ~60% of sellers complete with zero human contact), the Sebastian/Carli model stack, ML-driven pricing and underwriting, and reconditioning automation — but across all six calls management discussed these as reported current-state results and qualitative excitement, never as a prior quantified AI promise with both a number and a future timeframe/milestone that can be audited. The only quantified forward target (3M units / 13.5% adjusted EBITDA margin by 2030-2035) is an operational growth goal, not an AI-capability commitment, so there is no judgeable quantified AI promise to score.
PRICED-IN (REFINED)
MEDIUMEst. revisions falling · Fwd P/E 65.0 · EV/Sales 3.1x
AI claim maps to Used Vehicle Sales, Product and Service, Other
Analyst ratings have migrated down from 19 strongBuy/buy in January 2026 to 16 in June 2026, and price targets are falling with lastMonthAvg far below lastQuarterAvg and lastYearAvg. Forward revenue and EPS estimates still embed fast growth, but that is growth already in consensus rather than upward revision momentum. Valuation is rich at about 65x forward EPS and 3.1x EV/Sales, so AI-driven upside tied mainly to Used Vehicle Sales and Product and Service, Other is partly reflected despite falling revisions, making the priced-in verdict medium.
COVERAGE — ENTHUSIASM TRAJECTORY + CATALYSTS
2Q4 FY20242Q1 FY20252Q2 FY20258Q3 FY20256Q4 FY20256Q1 FY2026
AI enthusiasm across 6 calls — trend ↗ rising
AI/automation went from essentially absent to concrete data, model, workflow automation tied to delivery, self-service, and reconditioning efficiency.
RECENT AI CATALYSTS & NEWS
BUSINESS IMPACT - QUALITATIVE MATERIALITY
7/10 qualitative impact material near-term · mixed evidence
Where AI matters: self-service transactions, pricing, reconditioning workflow
AI/automation is already embedded in core retail and seller workflows, with 30% of buyers and 60% of sellers completing no-touch processes, plus automated pricing and reconditioning tools that can improve throughput and unit economics. The upside is material but not clearly transformational because management has not tied it to hard revenue or EPS uplift, and the latest framing is more operational than strategic.
Caveats: Benefits may already be part of normal operating leverage rather than incremental AI alpha; Management disclosures are mostly soft and current-state, with limited auditable financial targets; Competitors can adopt similar AI pricing, customer-service, and workflow tools; Execution risk in reconditioning and logistics remains more important than AI narrative
AI DISRUPTION / CANNIBALIZATION RISK two-sided · 3/10
AI does not automate away the need for vehicle inventory, inspection, financing, logistics, and fulfillment, so Carvana's physical/operational moat is durable. The risk is that better AI pricing, sourcing, shopping agents, and dealer tools become widely available and compress some digital-retail differentiation or margin, but this is not a direct cannibalization of the core model.
OPTIONS / MARKET STRUCTURE
option liquidity: good
proxy inputs — dollar-ADV $864M · beta 3.55 · px $65.60
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
Undercutting — insiders selling, institutions flat, management language 5/10 measured.
INSIDERS selling 53 open-market sell(s) vs 0 buy(s) — net distribution
INSTITUTIONS (13F) flat as of 2026-03-31: 106 new / 191 closed positions; 409 increased / 293 reduced; institutional ownership +1.18pp; -90 net 13F holders
MGMT LANGUAGE 5/10 measured No explicit AI discussion; automation/data-tooling language is real and owned, but magnitude is repeatedly qualified.
commit “Over the last couple of months, they built additional data integrations, developed tools to help managers make faster, higher quality decisions”
commit “implemented a productivity tracker to ensure feedback reaches the right groups quickly.”
commit “We'll continue to iterate on these tools, and we'll roll them out to the rest of the facilities over the coming months.”
VERBATIM AI QUOTES
“Over the last couple of months, they built additional data integrations, developed tools to help managers make faster, higher quality decisions, and how they staff their lines, and how they optimize flow through their paint lines and implemented a productivity tracker to ensure feedback reaches the right groups quickly.”
— Ernest Garcia, Q1 FY2026
“For example, we're making some investments now in additional technology, including AI-related technology that would be in that overhead expense number.”
— Mark Jenkins, Q1 FY2026
“We inspected. We run it through the reconditioning process after figuring out what needs to be fixed on the car, photograph it, put it up on the site, price it in an automated way.”
— Ernest Garcia, Q1 FY2026
“But on the other hand, it's also -- that's a problem that can be solved with algorithms data and pairing those two things together, very strong quantitative focus via software and via making even better use of all the data that we're collecting in the centers and then pairing that more effectively with the teams on the ground.”
— Mark Jenkins, Q1 FY2026
“And then we think as we continue to build out Carli that makes the systems inside Carli that make the individual operators more efficient and as we continue to build out these manager tools that make manager decision-making more straightforward, so they can focus on the other parts of management, making sure they're identifying their best performers and keeping people motivated and keeping the system moving.”
— Ernest Garcia, Q1 FY2026
“But it's still hard work and we still have significant room to continue to push more of the complexity of managing cars through these locations into systems with the goals of continually improving consistency across locations and of making scaling easier.”
— Ernest Garcia, Q4 FY2025
“We have 30% of our retail customers now go through the entire process without talking to a person until they get the car. We have 60% of our customers that are selling cars to us who go through the process without talking to anyone until they drop off their car.”
— Ernest Garcia, Q4 FY2025
“That's only possible because of the systems that we've built and those systems being intuitive and automated and straightforward. And I think a major set of tools that contributes to that is Sebastian and other tools that emerge from that AI brain.”
— Ernest Garcia, Q4 FY2025
“So I think that that's a very clear place where we're getting more scalable, where we're reducing costs. And I think very importantly, where we're improving customer experience.”
— Ernest Garcia, Q4 FY2025
“Those customers who go through the experience in that way have a higher NPS than customers that call us.”
— Ernest Garcia, Q4 FY2025
“But we do believe that we're fundamentally extremely well positioned to benefit from these things because we have a big deterministic system that's vertically integrated that has access to all the information and that brain has every system feeding it so we can give customers very simple answers to any questions they've got in really any software interface that we choose to put on top of it.”
— Ernest Garcia, Q4 FY2025
ANALYST QUESTIONS ON AI
Q (Q4 FY2025, Sharon Zackfia): I guess as a follow-up, I know you have your AI brand, I think, as well in the shareholder letter, and it seems to me you would be maybe the most uniquely poised to benefit from what's happening in an AI. Can you talk about what the early kind of nascent uses are that you're implementing AI to do? And then if you're seeing anything in the competitive set or if it's just business as usual there?
A: We have 30% of our retail customers now go through the entire process without talking to a person until they get the car. We have 60% of our customers that are selling cars to us who go through the process without talking to anyone until they drop off their car. That's only possible because of the systems that we've built and those systems being intuitive and automated and straightforward. And I think a major set of tools that contributes to that is Sebastian and other tools that emerge from that AI brain.
Q (Q1 FY2026, Rajat Gupta): You mentioned some investments around AI and stuff. Any way you could double click on that, give us a little more detail around what's going on.
A: There are ongoing investments, things that I wouldn't think of as seasonal or one time, including technology investments, some incremental investments in facilities. That, I think, will have us operating at a higher level on overhead expenses than we were in 2025.
Q (Q1 FY2026, John Healy): Obviously, we continue to see mobility and autonomous offerings rolling out in more cities. Obviously, you have a really good asset, and we've talked about capacity at the reconditioning centers. Have you guys game planned out any more that you could talk to us about maybe how you see yourself maybe facilitating those, that business potentially as being a service provider there?
A: I would say we're always paying attention, and I think we try to always be thoughtful about what opportunities exist out there given the assets that we've built. But I think we try to balance that with where is the best place to put our focus.
Q (Q1 FY2026, John Babcock): I don't know if you could talk about that a little bit, but that would be useful.
A: So for example, if you have a given number of people after reconditioning center, on any given day, with a given distribution of skill sets, what's the optimal way to distribute that team that you have on the ground that day across the various stations and the reconditioning process. And you can do that by hand, and you can do it on the ground manually. But on the other hand, it's also -- that's a problem that can be solved with algorithms data and pairing those two things together.