← back to rankingAUR · Aurora Innovation, Inc.
Information Technology Services · mkt cap $15.1B · calls: Q1 FY2026 vs Q4 FY2025
55.0 conviction · conf-adj 46
conf 4/10 partial
enthusiasm:15.0 · trend:-5 · quantifies:12 · impact:8 · under_radar:0 · credibility:12 · business_impact:10 · disruption:0 · commitment:0 · confirmation:3
Enthusiasm latest 5 / prev 7 (falling)
Aurora’s AI story is autonomy-first: generalized driver software, Verifiable AI–automated Aurora Atlas mapping, and industrialized compute/sensor hardware—not generative AI or LLM copilots. Q4 FY2025 carried the richer AI substance (Verifiable AI, cloud map automation, quantified weather downtime and auto-generated map miles); Q1 FY2026 shifted to scaling/driverless commercialization with only a brief, philosophy-heavy AI answer (rejecting unverifiable models, affirming internal models). Credibility rests on operational metrics (driverless miles, lane-open speed, utilization) and cost-down targets tied to the stack, not on AI-labeled revenue lines.
GROUNDED NEXT-FY IMPACT vs CONSENSUS
Grounded on actual base — revenue $0.0B · net income $-0.8B · net margin -27200.0% · diluted EPS -0.44
These are next-fiscal-year annual uplift estimates, not next-quarter numbers.
Aggregate next-FY est. rev uplift: 750.0% · next-FY EPS uplift: % · vs analysts: inline · priced in: high · confidence: 4/10
| Claim | Figure | Arithmetic | Next-FY Rev % | Next-FY EPS % |
|---|
Weather constrained TX driverless ~40% of 2025
other · soft | ~40% of the time | Operational uptime constraint/drag, not a revenue or cost line; no $/truck or recovery % disclosed to convert 40% downtime into recognized revenue or saving. Unsizeable. | | |
Vast majority of HD map miles auto-generated
productivity · soft | the vast majority | Unanchored ('vast majority'); no mapping-cost $ or FTE base disclosed. Supports faster lane expansion but not quantifiable. | | |
Gen-2 kit cuts Aurora Driver hardware cost 50%+
cost · soft | 50-plus percent reduction | Anchored as a percent but NO dollar hardware-cost/kit-COGS base appears anywhere, so 50%×base is uncomputable (FY2025 gross profit -$14M on $3M rev gives no reliable COGS anchor). Key lever for breakeven gross-margin target but unsizeable in $. | | |
370,000 driverless miles, 0 attributed collisions
other · soft | 370,000 miles | Safety/performance proof point. Y derives implied revenue via disclosed $1.78/mi ($80M / (200 trucks × 225k mi)) → 370k × $1.78 ≈ $0.66M = +21.9% vs $3M base; but those miles are subsumed in the fleet ramp / $80M row, so adding them separately double-counts. Kept null to avoid double counting. | | |
Werner trucks 4,000+ mi/wk (~225k mi/yr); >2x utilization
revenue | >2x utilization; 225,000+ mi/truck/yr | Revenue-per-truck mechanism behind the $80M run rate: 225k mi × ~$1.78/mi ≈ $400k/truck/yr × 200 trucks = $80M. The '>2x' is carrier-side economics. Sized within the $80M TaaS row to avoid double counting; no standalone $ added. | | |
FirstLight 1km lidar range, 34s reaction at highway speed
other · soft | 1 km; 34 sec | Product capability/safety spec; no revenue or cost dollars. Enables driverless operation but not a direct P&L figure. | | |
>200 driverless trucks in operation exiting 2026
revenue | >200 trucks | Fleet-scale enabler of the $80M exit run rate ($80M / 200 = ~$400k rev/truck/yr). Same claim as the run rate from two angles; sized in the $80M row, not additive. | | |
~$80M TaaS revenue run rate exiting 2026
revenue | ~$80M run rate | EXIT-2026 run rate, NOT full-FY2026 recognized revenue. Backloaded ramp to >200 trucks: defensible FY2026 recognized revenue ranges ~$20M (X, near consensus $15.0M) to ~$31M (Y, 35% of exit RR + $3M legacy); midpoint ~$25.5M. Incremental over current $3.0M ≈ $22.5M; rev_uplift_pct = 100×22.5/3 ≈ 750% — a huge figure that is purely a tiny-base artifact. EPS null: company loss-making (NI -$816M), any EPS% meaningless. | 750 | |
Dallas-Laredo lane validated in 6 weeks
productivity · soft | 6 weeks | Generalization/speed-to-market capability; no $ base for engineering saving or incremental lane revenue disclosed. Unsizeable. | | |
Gen-2 kit designed for 1M miles vs ~300k prior
cost · soft | 1M mi vs ~300k mi (~3.3x life) | ~3.3x asset life lowers per-mile hardware amortization (~70% lower depreciation per mile; combined with 50% kit cost ⇒ ~85% lower per-mile hardware burden), supporting gross margin — but no dollar kit cost or per-mile COGS base disclosed, so the $ saving is uncomputable. Bottom-line lever, unsizeable. | | |
Assumptions: Next FY = FY2026 (ending 2026-12-31). $80M is an EXIT-2026 run rate, not full-year recognized; FY2026 recognized revenue assumed ~$20M-$31M (midpoint ~$25.5M) given a backloaded ramp (consensus has $15.0M). rev_uplift_pct = 100 × incremental next-FY revenue / current revenue ($3.0M). $/mile (~$1.78) derived only from disclosed $80M, 200 trucks, 225k mi/truck/yr (DISCLOSED-BASE rule). Tax 21% (irrelevant — pre-profit). No incremental net-margin flow-through to EPS: company deeply loss-making (FY2025 NI -$816M; consensus negative through FY2027), so EPS% off that base is meaningless. Hardware savings expressed as per-mile % only; no kit COGS $ disclosed. Consensus treated as non-GAAP/adjusted basis. All claims adopter-side (AI = Aurora's own self-driving product); no supplier-side compute/chip-sale revenue.
Top line: Entirely adopter-side: AI/autonomy is Aurora's product, not compute it sells. The one hard, sizeable topline figure is the ~$80M TaaS exit-2026 run rate (>200 trucks × ~225k mi/truck/yr at ~$1.78/mi ≈ $400k/truck/yr). Phased to FY2026 (~$20M-$31M recognized, midpoint ~$25.5M), that is ~+750% over the $3.0M base — a huge percentage that is purely a tiny-base artifact (low confidence on magnitude). vs FY2026 consensus revenue $15.0M, the phased path runs ~1.3x-2.1x Street, so near-term revenue may be modestly under-modeled. The 370k-mile proof point (~$0.66M implied) is subsumed in the fleet ramp; the '>2x' customer utilization is carrier economics, not incremental Aurora revenue beyond TaaS.
Bottom line: No computable EPS uplift: deeply loss-making (FY2025 NI -$816M, EPS -$0.44/-$0.36 adj) and consensus stays negative (FY2026 NI ~-$922.5M, FY2027 ~-$917.9M), so est_eps_uplift_pct is null per the loss-making guardrail. The genuine bottom-line levers — 50%+ Gen-2 hardware cost cut and ~3.3x kit durability (1M vs 300k miles), together ~85% lower per-mile hardware burden, aimed at the breakeven gross-margin target against today's -$14M gross profit — are real and qualitatively important but have no disclosed dollar base, so they cannot be sized and remain soft. Operating loss (~-$901M) dominates; fleet COGS savings immaterial near-term vs R&D/opex.
[EPS uplift n/m (loss-making base); rev uplift 750% low-confidence (sub-$1B base)] Management's hard guidance ($80M exit-2026 run rate) phased to FY2026 (~$20M-$31M) sits above FY2026 consensus revenue ($15.0M), so near-term revenue may be modestly under-modeled — but the outer-year scale-up is in the numbers, not above them: consensus FY2026 rev $15.0M is already +400% over the $3M base and FY2027 $192.6M is ~2.4x the $80M run rate. Consensus NI stays ~-$923M FY26 / -$918M FY27 despite the revenue ramp, so management's cost claims (Gen-2 50% hardware cut, 3.3x durability) are not yet embedded in Street losses but also cannot be sized in $. Net: the driverless revenue scale-up is largely priced in; cost levers are real but unquantifiable.
MODEL CONSENSUS (impact)
partial
Loss-making adopter; only hard topline is $80M exit-2026 TaaS run rate. Averaged phasing; conservative verdicts where tied. Cost levers real but unsizeable.
Conflicts reconciled
- est_rev_uplift_pct: X=567 vs Y=933.33 -> used 750 (average of two defensible FY2026 phasings, ~$20M vs ~$31M recognized)
- math[370k miles].rev_uplift_pct: X=null(soft) vs Y=21.93 -> used null because Y itself notes the miles are subsumed in the fleet ramp / $80M row; counting separately double-counts
- vs_analyst_expectations: X=inline vs Y=ahead -> used inline (conservative; outer-year consensus already ~2.4x the run rate)
- priced_in: X=high vs Y=medium -> used high (FY2027 consensus $192.6M embeds the ramp)
- math[Gen-2 1M-mi durability].soft: X=true vs Y=false -> used true (anchored ratio but no $ base obtainable, so unsizeable)
| Field | Opus 4.8 | GPT-5.5 |
|---|
| Rev uplift % | 567 | – |
| EPS uplift % | – | – |
| Priced in | high | – |
| vs analysts | inline | – |
| Confidence | 4 | – |
| Top line | Entirely adopter-side: AI/autonomy is Aurora's product, not compute it sells. The one hard, sizeable topline figure is the ~$80M TaaS exit-2026 run rate (>200 trucks at ~225k mi/truck/yr, >2x utilization). Phased to FY2026 (~$20M recognized, backloaded ramp), that is ~+567% over the $3.0M current base — a huge percentage that is purely a tiny-base artifact. Consensus already embeds this: FY2026 rev $15.0M (+400% vs base) and FY2027 $192.6M, the latter 2.4x the $80M run rate. So the driverless revenue scale-up is in the numbers, not above them. | – |
| Bottom line | No computable EPS uplift: the company is deeply loss-making (FY2025 NI -$816M, EPS -$0.44) and consensus stays negative (FY2026 NI -$922.5M, FY2027 -$917.9M), so est_eps_uplift_pct is null per the loss-making guardrail. The genuine bottom-line levers — 50%+ Gen-2 hardware cost cut and ~3.3x kit durability (1M vs 300k miles), both aimed at the 'breakeven gross margin' target against today's -$14M gross profit — are real and qualitatively important but have no disclosed dollar base, so they cannot be sized and remain soft. | – |
| Reasoning | Management's hard guidance ($80M exit-2026 run rate) sits BELOW where consensus already has the business: consensus FY2027 revenue of $192.6M is ~2.4x the $80M run rate, and FY2026 consensus of $15.0M is consistent with a backloaded ramp. The math points at-or-below consensus, not above it, so the AI/driverless story is priced in. On the bottom line, consensus net income stays ~-$920M through FY2027, leaving no room for an EPS uplift to be meaningful. | – |
Rows highlighted where the two models disagreed.
QUANTIFICATIONS
Weather-related driverless downtime (Texas, 2025): roughly 40% of the time (During 2025, both)
“During 2025, inclement weather of all types constrained our driverless operations in Texas roughly 40% of the time.”
Automated HD map mile generation share: the vast majority (At this point (Q4 FY2025 call), bottomline)
“At this point, the vast majority of the miles that we generate are automatically generated.”
Aurora Driver hardware cost reduction (Gen-2 kit): 50-plus percent reduction / 50% plus reduction (Second-generation commercial hardware kit (2026), bottomline)
“we expect this kit to drive a 50-plus percent reduction in Aurora Driver hardware costs, a key lever supporting our breakeven gross margin target.”
Driverless miles operated (safety/performance proof): 370,000 driverless miles (Through April 2026 (Q1 call), both)
“in April, the Aurora Driver surpassed 370,000 driverless miles with 100% on-time performance and 0 Aurora Driver attributed collisions.”
Truck utilization / revenue per truck uplift: more than doubling utilization; 4,000-plus miles per week (~225,000-plus annual miles per truck) (Current Werner operations; forward expectation for customers, topline)
“the driverless trucks we are operating for Werner are already averaging 4,000-plus miles per week, which translates to an annual run rate of 225,000-plus miles per truck. With the performance we're seeing, we expect Aurora Driver powered trucks will be capable of more than doubling utilization and in turn, revenue per truck for our customers.”
FirstLight lidar range / reaction time at highway speed: 1-kilometer range; more than 34 seconds to react (Second-generation commercial hardware kit, both)
“an extended 1-kilometer range for FirstLight, our proprietary long-range FMCW lidar. This is double the range of the closest FMCW lidar competitor and can give the Aurora Driver more than 34 seconds to react when at highway speeds”
Driverless fleet scale target: more than 200 driverless trucks (Exit 2026, topline)
“We anticipate exiting the year with more than 200 driverless trucks in operation.”
Revenue run rate at fleet scale: approximately $80 million in revenue on a run rate basis (Exit 2026 (Transportation-as-a-Service), topline)
“which translates to approximately $80 million in revenue on a run rate basis for our Transportation-as-a-Service business.”
Lane validation speed (generalization): within just 6 weeks (Dallas–Laredo route; end of March 2026, both)
“At the end of March, we validated driverless operations on the bidirectional route between Dallas and Laredo within just 6 weeks of initiating supervised autonomous runs.”
Hardware kit durability (per-mile economics): 1 million miles of operation (vs. ~300,000 miles prior kits) (Second-generation vs. first-generation hardware, bottomline)
“Designed for 1 million miles of operation”
PAST (realized)
- Q4 FY2025 — Christopher Urmson: "During 2025, inclement weather of all types constrained our driverless operations in Texas roughly 40% of the time."
- Q4 FY2025 — Christopher Urmson: "At this point, the vast majority of the miles that we generate are automatically generated."
- Q4 FY2025 — Christopher Urmson: "In January, the Aurora Driver passed 250,000 driverless miles, nearly tripled the cumulative miles achieved through early October."
- Q1 FY2026 — Christopher Urmson: "in April, the Aurora Driver surpassed 370,000 driverless miles with 100% on-time performance and 0 Aurora Driver attributed collisions."
CURRENT (now)
- Q1 FY2026 — Christopher Urmson: "We're on the cusp of launching our second-generation commercial hardware kit on a new fleet of driverless trucks."
- Q1 FY2026 — Christopher Urmson: "Our forthcoming software release further increases the Aurora Driver's reliability in preparation for scaling, including validation of driverless operations in more severe rain as well as the full spectrum of complex construction scenarios on our highway routes."
- Q1 FY2026 — Christopher Urmson: "We're also in the process of validating our second-generation commercial hardware kit on multiple truck platforms through rigorous on-road, track and lab testing to prepare for our planned second quarter launch."
- Q4 FY2025 — Christopher Urmson: "By leveraging our Verifiable AI systems, our cloud-based algorithms are able to generate semantic components of the Aurora Atlas from collected data automatically building portions of the map with little or no human assistance."
- Q4 FY2025 — David Maday: "And with our generalized AI approach right now, it doesn't take a tremendous amount of effort to take advantage of that next lane."
FORWARD (guidance)
- Q1 FY2026 — Christopher Urmson: "We continue to look for ways we can bring those ideas in and fuse them with our approach to ensuring that we can deliver a safe vehicle on the road."
- Q1 FY2026 — Christopher Urmson: "Our engineering team is also working with AUMOVIO and NVIDIA to develop a first-of-its-kind Super Thor compute configuration an architecture that integrates 2 NVIDIA DRIVE for SoCs into a unified platform optimized to power the Aurora Driver at scale."
- Q4 FY2025 — Christopher Urmson: "we expect the pace of map expansion to continue to increase as we further optimize automation in our cloud mapping software."
- Q1 FY2026 — Christopher Urmson: "we expect this kit to drive a 50-plus percent reduction in Aurora Driver hardware costs, a key lever supporting our breakeven gross margin target."
TRACK RECORD — PROMISE vs DELIVERY
72/100 track record mixed 6 calls reviewed
Aurora has a strong record on core autonomy milestones—April 2025 driverless launch, safety-case closure, API readiness, night ops, lane expansions, and published driverless-mile thresholds— but repeatedly falls short or slips on fleet-scale and hardware-rollout numbers (tens of trucks in 2025, Gen-2 in 2025, “hundreds” to 200+ for 2026).
~90% of commercial loads at 100% API (no on-site support) by April 2025 driverless launch — promised Q4 FY2024
delivered Q1 FY2025 reported 95% of launch-lane loads at 100% API, above the 90% launch target.
First driverless commercial trucks on Dallas–Houston in April 2025 after closing the safety case (ARM) — promised Q4 FY2024
delivered Driverless operations began April 27, 2025; Q1 FY2025 said ARM reached 100% and launch occurred as planned.
Operate up to 10 commercial trucks at launch (starting with 1 driverless), scaling to tens of driverless trucks by end of 2025 — promised Q4 FY2024
partial Commercial launch occurred but fleet stayed small (e.g., 3 trucks in Q2 FY2025, 5 regularly scheduled in Q3 FY2025)—well below “tens” by year-end.
Second-half 2025: validate night driving and rainy/adverse weather; expand driverless lanes Fort Worth–El Paso and to Phoenix — promised Q4 FY2024
partial Night driverless validated ~3 months post-launch (ahead of H2); Fort Worth–El Paso went driverless in Q3; rain/heavy wind tied to Jan 2026 software (Q4 FY2025 cited inclement-weather driverless capability with latest release); Phoenix extension slipped past end-2025.
Introduce second-generation commercial hardware kit later in 2025 to cut hardware costs and support scaling — promised Q4 FY2024
partial 2025 focused on B-sample testing and validation; management pushed driverless deployment on Gen-2 to Q2 2026 (International LT fleet, no observer).
Exit 2026 with hundreds of driverless trucks in operation — promised Q3 FY2025
quietly-dropped By Q4 FY2025 and Q1 FY2026 the target was reframed to more than 200 driverless trucks exiting 2026, with no explicit reconciliation of the lower number.
PRICED-IN (REFINED)
HIGH (already in)Est. revisions rising · Fwd P/E -21.2 · EV/Sales 3,733.8x
Analyst sentiment has migrated up (strongBuy+buy from 5 in Jan–Feb to 7 in May–Jun; holds stable) and price targets step up (lastYearAvg 9.8 → lastQuarterAvg 10 → lastMonthAvg 12.5), while forward revenue consensus jumps from ~$3.6M (2025) to ~$193M (2027), baking in a large autonomy ramp. Valuation is already extreme for current scale (~$15B EV, EV/Sales ~3,734x TTM; negative forward EPS makes P/E uninformative but confirms loss-funded optionality). With no product/geographic segmentation in the feed, AI upside cannot be isolated to a line item—it is effectively priced at the enterprise level. Rising revisions plus a stretched multiple point to AI/autonomy upside largely reflected in the stock.
COVERAGE — ENTHUSIASM TRAJECTORY + CATALYSTS
7Q4 FY20248Q1 FY20257Q2 FY20258Q3 FY20259Q4 FY20258Q1 FY2026
AI enthusiasm across 6 calls — trend ↗ rising
From launch-prep autonomy demos to Verifiable AI, perception wins, Atlas map automation, then scaled NVIDIA compute.
RECENT AI CATALYSTS & NEWS
BUSINESS IMPACT - QUALITATIVE MATERIALITY
10/10 qualitative impact transformational near-term · mixed evidence
Where AI matters: autonomous trucking product and TaaS revenue
Machine-learning perception, planning, and Verifiable AI mapping are Aurora's core product—not internal productivity tools—with hard exit-2026 ~$80M TaaS run-rate guidance, 370k driverless miles, and majority-automated Atlas map generation; Gen-2 hardware cost cuts remain percent-guided without a disclosed COGS base.
Caveats: Fleet-scale targets repeatedly trimmed (hundreds to 200+ exiting 2026) with Gen-2 hardware rollout slipped to Q2 2026; Still deeply loss-making (~$900M+ opex) with unquantified dollar impact from 50%+ hardware cost claims; Weather constrained Texas driverless ops ~40% of 2025; lane expansion pace remains customer- and ops-dependent; Intense AV competition (OEM in-house stacks, Kodiak, capital-rich peers) could compress TaaS pricing before scale economics arrive
AI DISRUPTION / CANNIBALIZATION RISK tailwind · 2/10
Aurora sells AI-enabled autonomous miles, not billable human labor or arbitrage services that GenAI compresses; AI is the revenue engine rather than a deflator of the unit sold, with only indirect risk that commoditized rival autonomy stacks erode pricing power.
OPTIONS / MARKET STRUCTURE
option liquidity: fair
proxy inputs — dollar-ADV $188M · beta 2.586 · px $7.72
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 4/10 measured.
INSIDERS selling 5 open-market sell(s) vs 0 buy(s) — net distribution
INSTITUTIONS (13F) adding as of 2026-03-31: 87 new / 60 closed positions; 247 increased / 94 reduced; institutional ownership +8.30pp; +24 net 13F holders
MGMT LANGUAGE 4/10 measured No AI/ML lexicon; autonomy and Driver scaling use firm ops metrics but commercial and positioning language stays qualified.
commit “we expect this kit to drive a 50-plus percent reduction in Aurora Driver hardware costs”
commit “in April, the Aurora Driver surpassed 370,000 driverless miles with 100% on-time performance”
commit “we are augmenting our driverless network to support real-time dynamic rerouting”
VERBATIM AI QUOTES
“We're excited about the models that we're building at Aurora. We continue to be deep believers in verifiable AI. The idea that you would trust 1 of these giant trucks driving down the road with something where you just kind of hope the output is the right thing given the input. It just doesn't make sense. And so we continue to look for ways we can bring those ideas in and fuse them with our approach to ensuring that we can deliver a safe vehicle on the road.”
— Christopher Urmson, Q1 FY2026
“With the Aurora Driver now sufficiently generalized for us to begin scaling across the Sunbelt aligned with customer demand, we strategically focused our resources on 3 key initiatives: expanding our driverless network, finalizing our latest software release and validating our second-generation commercial hardware kit.”
— Christopher Urmson, Q1 FY2026
“Our forthcoming software release further increases the Aurora Driver's reliability in preparation for scaling, including validation of driverless operations in more severe rain as well as the full spectrum of complex construction scenarios on our highway routes.”
— Christopher Urmson, Q1 FY2026
“We're also in the process of validating our second-generation commercial hardware kit on multiple truck platforms through rigorous on-road, track and lab testing to prepare for our planned second quarter launch and are seeing impressive performance. Designed for 1 million miles of operation and with enhanced sensor cleaning capabilities, this kit meaningfully increases the Aurora Driver's reliability. It also brings exciting performance gains including a more efficient computer and an extended 1-kilometer range for FirstLight, our proprietary long-range FMCW lidar.”
— Christopher Urmson, Q1 FY2026
“Our engineering team is also working with AUMOVIO and NVIDIA to develop a first-of-its-kind Super Thor compute configuration an architecture that integrates 2 NVIDIA DRIVE for SoCs into a unified platform optimized to power the Aurora Driver at scale.”
— Christopher Urmson, Q1 FY2026
“With the performance we're seeing, we expect Aurora Driver powered trucks will be capable of more than doubling utilization and in turn, revenue per truck for our customers.”
— Christopher Urmson, Q1 FY2026
“When it comes to the software that operates on board, that's been a core part of how we've thought about this is how do we ensure that the software will generalize safely over time. It's what leads us to do as much work as we do in testing and validation.”
— Christopher Urmson, Q1 FY2026
“This new release that is going to land with the second-generation hardware really is about making sure that we have a robust platform that's reliable and meets customer needs, increasing the amount of rain we can handle dealing with more complicated construction that we need to deal with on freeways.”
— Christopher Urmson, Q1 FY2026
“With our latest software release, we believe the Aurora Driver is now sufficiently generalized for us to begin expanding across the Sunbelt in 2026.”
— Christopher Urmson, Q4 FY2025
“During 2025, inclement weather of all types constrained our driverless operations in Texas roughly 40% of the time. Our latest software release drives a step change in potential availability and utilization across the Sunbelt, a core component of our value proposition.”
— Christopher Urmson, Q4 FY2025
“The Aurora Driver now has generalizable skills. And we've made meaningful progress automating the creation of new content for the Aurora Atlas, our proprietary high-definition map technology that enhances the safety and computational efficiency of the Aurora Driver. By leveraging our Verifiable AI systems, our cloud-based algorithms are able to generate semantic components of the Aurora Atlas from collected data automatically building portions of the map with little or no human assistance.”
— Christopher Urmson, Q4 FY2025
“This drastically accelerates the production of Atlas content, and we expect the pace of map expansion to continue to increase as we further optimize automation in our cloud mapping software.”
— Christopher Urmson, Q4 FY2025
“On the mapping front, this is a place where the approaches we've taken with AI to be able to detect and understand the road structure to be able to gather data and then feed that back to the mothership are paying huge dividends because we can take that online system, run it offline, verify, validate it and then create the road data product. At this point, the vast majority of the miles that we generate are automatically generated.”
— Christopher Urmson, Q4 FY2025
“And with our generalized AI approach right now, it doesn't take a tremendous amount of effort to take advantage of that next lane.”
— David Maday, Q4 FY2025
“We've reduced the mass of the compute as an example.”
— Christopher Urmson, Q4 FY2025
ANALYST QUESTIONS ON AI
Q (Q1 FY2026, Mark Delaney (Goldman Sachs)): Chris, I was hoping to get your latest thoughts on AI technology and any new innovations that Aurora is looking at one thing that's had more discussion in the investment community and tech community recently has been world models, but curious whether it's that or other newer technologies that you're observing and anything that could be impactful for Aurora?
A: No. We continue to pay attention to what's happening outside. We're excited about the models that we're building at Aurora. We continue to be deep believers in verifiable AI. The idea that you would trust 1 of these giant trucks driving down the road with something where you just kind of hope the output is the right thing given the input. It just doesn't make sense. And so we continue to look for ways we can bring those ideas in and fuse them with our approach to ensuring that we can deliver a safe vehicle on the road.
Q (Q4 FY2025, Chris Pierce (Needham)): I guess, just how should we put the pieces together there? ... maybe you have more OpEx leverage as you lean into AI and on mapping?
A: David Maday did not address AI or mapping technology; he discussed free-cash-flow-positive timing, liquidity (~$1.5B), ATM usage, and planning another Investor Day for outlook beyond 2026.
Q (Q4 FY2025, Colin Rusch (Oppenheimer)): I'm curious about how we can track the cadence of incremental operating domains and environments from a weather perspective throughout the year.
A: Chris Urmson: generalizable skills are largely in place; lane expansion is increasingly a mapping/operational exercise; "approaches we've taken with AI" for road-structure detection feed offline verification and map products; "the vast majority of the miles that we generate are automatically generated"; expansion pace will be driven by customer demand.