15 min read

Mobileye (MBLY) — Q3 FY2026 Earnings Briefing

Executive Summary

  • A beat-and-raise quarter: Q3 revenue of $492M slipped year over year yet cleared the company’s guided decline, as EyeQ demand ran ahead of plan all year.
  • The full-year adjusted operating income outlook now sits about double what Mobileye budgeted entering 2026 — the Israeli R&D credit and cost discipline, not pricing, did the lifting.
  • Volume is the engine: shipments decoupled from a slightly shrinking vehicle market, with top-10 customers taking more Mobileye content on fewer vehicles and Chinese OEM export programs adding outsized incremental volume.
  • Mix cuts the other way: software-light Chinese export content and a dollar-for-dollar memory-cost pass-through dilute the averaged selling price and gross margin.
  • The quarter’s tension: the base business grows units, not dollars — the content-rich launches (Surround ADAS, Cloud-Enhanced ADAS, Chauffeur) wait for later-this-decade timing, so the investment case hinges on that conversion landing on schedule.

Highlights and Lowlights

Highlights

  1. Shipments decoupled from the vehicle cycle: EyeQ shipments ran up 11% year to date against slightly declining global automotive production, with top-10 customers raising Mobileye volume while their own vehicle production falls.
  2. Q3 volume landed at the top of the plan: 9.5 million EyeQ units, the high end of the 9.3–9.5 million outlook, extending a year of demand ahead of expectations.
  3. SuperVision ran ahead of schedule: deliveries exceeded 15,000 systems in the quarter and sit near 60,000 for the year to date, already beyond the original full-year expectation before a modest raise.
  4. Full-year revenue guidance rose again, to roughly $2.008B at the midpoint, implying 5–7% growth for 2026.
  5. The Israeli R&D credit hit its first clean run-rate quarter, contributing $51M of contra-R&D expense, and management sees the benefit extending into future years with no end date to the rule.
  6. Robotaxi reached the pre-scaling milestone on the driver-out path: the vertically integrated route plans a 100-vehicle, single-city U.S. deployment in 2027, with the Moovit rider-experience and fleet-orchestration stack already in initial testing.
  7. Content wins stacked up: three OEM selections for Surround ADAS — Volkswagen, Mahindra, and a large U.S. OEM — while two additional OEMs chose Cloud-Enhanced ADAS beyond Stellantis.

Lowlights

  1. Q4 is a hard downshift quarter: guidance implies revenue of about $450M — roughly 8.5% below Q3 — with adjusted operating income near $60M as customer orders thin into year end.
  2. Memory inflation leaks margin: the one-for-one customer pass-through carried about $7M of Q3 revenue with zero gross margin and, together with SuperVision mix, shaved roughly 100 basis points from the corporate gross margin — and 2027 pass-through terms are still being negotiated.
  3. Pricing is diluted by design rather than demand: software-light Chinese export content and the pass-through mechanics both press the blended economics even as units grow.
  4. Volume is partly borrowed from the channel: management again described intentional customer inventory-building against component-shortage risk, which it still expects to reverse in Q4 shipments.
  5. The production-over-demos strategy carries a stated cost: years of concentration on launch programs traded near-term nominations for long-term durability, and business development on advanced systems reopens only during 2027.
  6. Forward spending stays opaque: GPU-farm purchases and other 2027 investments are planned but unquantified, described as “not something major” with details deferred to the next call.
  7. Capital return remains token so far: repurchases total roughly $33M since the April authorization, at an average $9.36 — small against the cash accumulating on the balance sheet.

1. The Quarter at a Glance

Metric Q3 2026 Q3 2025 Change
Revenue $492M $504M ~2% below, above the guided 2–5% decline
Year-to-date revenue first nine months — +8% YoY
Year-to-date adjusted operating income first nine months — +42% YoY
Nine-month operating cash flow $303M — sustained generator
Cash, marketable securities and deposits $1.5B — strong balance sheet

The quarter’s arithmetic is two-sided: revenue dollars fell even as nearly every volume metric outperformed, and the profitability jump has an accounting accelerant behind it. The body below separates what came from demand, what came from mix, and what came from the Israeli R&D credit.

2. Volume: Shipments Decoupled From the Vehicle Cycle

  • The defining fact of the year: shipments grew while the vehicle market did not. Q3 EyeQ volume landed at the top of the outlook (see Highlights), volume rose 3% YoY in the quarter, and full-year EyeQ shipments are guided to grow about 10% at the midpoint versus 2025 — against slightly declining global automotive production through Q3.
  • The top-10 Western story crystallizes it: Mobileye expects volume with those customers to grow 7% even as their aggregate vehicle production declines 5% — the fitment trade outgrowing the vehicle cycle.
  • Chinese OEMs are the accelerant: their export push brought in more than 1 million incremental units and four percentage points of growth, chosen to standardize ADAS on cars headed to Latin America, South Asia, and Eastern Europe. Their rapid emerging-market expansion is the multi-year tailwind, and their standardized-ADAS exports pressure legacy OEMs to lift fitment in the same regions.
  • India remains the next high-volume growth market near term, on new safety regulations and strong consumer demand for safety features.
  • Management’s own triad explains the beat against expectations: Mobileye technology adopted faster than its top-10 customers grow production; Chinese automakers preferring Mobileye for export credibility; and the legacy-OEM fitment ripple noted above.
  • On China durability, management pushed back on the recurring competitive claim: such claims have run “at least for 4 years now,” Mobileye’s products have consistently scored highest in European regulatory tests, and Chinese OEMs have picked Mobileye repeatedly across multiple decision cycles. More than 4.5 million EyeQ chips are being delivered into China with demand sustained into Q4 — roughly even across quarters, per the CFO.

3. Pricing: Diluted by Mix, Cushioned by Pass-Through

  • The principal ASP driver is composition. Chinese OEM export shipments carry less software content, so their higher mix alone pulled device-level pricing down about $1 YoY in Q3. Layer on second-chip programs sold this year but not last, and 2026 carries roughly a $2 full-year device-price decline; 2027 is expected to run flat to modestly lower rather than repeat the headwind.
  • Memory costs raised unit cost and were recovered at face value: the customer pass-through added revenue at zero gross margin (quantified in Highlights) — it grows the top line without buying any profit.
  • The 2027 pass-through terms are unresolved. Memory prices are expected to rise industry-wide next year, work with customers on 2027 pricing is “in the midst” of the process, and management committed to the full picture at the next earnings call after a history of successful macro-driven cost claims with customers.
  • A modest counterpoint: Q4 mix expectations include more demand for higher software bundles, so the CFO flagged some expected increase in device pricing in the final quarter.

4. Profitability: The R&D Credit’s First Clean Quarter

  • Adjusted operating income rose to $89M, up 20% YoY, with adjusted operating margin at 18%, up about three points YoY — on a top line that declined.
  • The Israeli R&D credit law contributed $51M of contra-R&D in Q3 (see Highlights). The law took effect shortly after Q1 and applies retroactively to the start of 2026; Q3 was the first quarter reflecting its benefit on a clean run-rate basis.
  • Management’s durability framing: the benefit extends into future years with no end date to the regulation, though it remains subject to legislative change. Recognition is volatile quarter to quarter on thresholds, qualifying Israeli R&D spend, and exchange rates — and the cash benefit lags accounting recognition by roughly two years, landing from 2028.
  • Non-GAAP reporting excludes intangible amortization from the 2017 Intel acquisition, stock compensation, the Q1 goodwill impairment, Mentee transaction costs, and the R&D credit’s offsetting impact. The full-year credit magnitude sits in the guidance table below.

5. Guidance: Raised Again, Still Deliberately Lumpy

Item FY2026 guidance Change
Revenue $1.995B–$2.02B (midpoint ≈ $2.008B) +$13M; now 3.5% above the original expectation; implies 5–7% growth
Adjusted operating income $380M–$420M (midpoint $400M) midpoint raised from $395M
R&D incentive contribution ≈ $190M at the midpoint carried forward
Non-GAAP operating expenses ≈ $910M midpoint, including $180–200M of R&D credits range kept wide pending further incentive-law guidance expected within weeks
SuperVision deliveries ≈ 65,000 systems modestly raised
  • The raise is SuperVision-shaped: the company now expects about 65,000 systems, versus a prior outlook it had already effectively reached through the first three quarters (see Highlights), and Q3/Q4 device pricing runs meaningfully higher on the memory pass-through. Conversion of the incremental revenue into operating income tracks expectations because SuperVision carries below-average gross margin.
  • The EyeQ outlook is unchanged though its Q4 shape is not: EyeQ orders currently point to roughly 8.8–8.9 million units — a visible step below the quarter just reported — with lower SuperVision deliveries and Q4 opex set a bit above Q3’s roughly $225M. On the breadth of the volume story, shipments vary widely quarter to quarter, so the more meaningful measure is the trend across the last twelve months including Q4: volume growth of about 10% (see section 2).
  • The channel-hygiene caveat persists: some of this year’s demand is intentional customer inventory-building to hedge component shortages, and the Q4 volume step-down is management’s forecast reconciliation to that — consistency with the July-call plan management reiterated.

6. The Content Ladder: Where Pricing Re-Rates

The growth arc management sketched for 2027–2030 has two reinforcing drivers: ADAS reaches more vehicles, and Mobileye captures more value inside each one. OEMs are pursuing that along two parallel paths.

  • Path A — low-cost highway hands-free. Mobileye technology already sits in most first-generation L2+ systems on the road, and in the majority of vehicles produced in North America and Europe in 2026 — typically still only as the front-camera perception supplier. Second-generation programs widen the role to surround camera vision, crowd-sourced road intelligence (REM), driving functions, driver monitoring, and automated parking. Three OEMs have picked Surround ADAS on EyeQ6 High — Volkswagen, Mahindra, and a large U.S. OEM — with product launches landing in 2028 and the VW program alone spanning a double-digit number of vehicle models that sell in the millions per year.
  • Cloud-Enhanced ADAS is the faster route. Adding REM to existing EyeQ6 Lite programs earned three selections: Stellantis (announced early in Q3, since expanded to European platforms), plus one top-5 automaker from Asia and one top-10 OEM from Japan. Execution is deliberately simple — same chip, same hardware, mostly software validation — and functions as a bridge to the bigger EyeQ6 High programs.
  • REM got its own call-out. The road-intelligence database crowd-sourced from about 8 million vehicles has accelerated over the last two quarters, catching interest even in multi-supplier point-to-point programs. The enabling shift: OEMs are discovering that turning a good demo into a validated vehicle system across millions of miles is brutal — and Tesla publicly reversing course to build its own crowdsourced version only reinforces that scale-built road databases cannot be duplicated cheaply.
  • Path B — point-to-point systems in the Tesla FSD mold. The main thrust is EyeQ6 High-based SuperVision and Chauffeur with VW Group: SuperVision nearer term (Porsche-led, reaching maturity in early 2027), Chauffeur in 2028 (Audi-led, and treated by the customer as an industry first for consumer Level 3 — the critical selling point behind a 2028 flagship launch).
Product level Pricing economics Gross margin character Program status
Base ADAS (EyeQ6 Lite front camera) ≈ $47–$50 in recent quarters corporate standard; single-chip pricing has eroded for a couple of years run-rate today
Cloud-Enhanced ADAS with REM a few tens of dollars per EyeQ; ≈ single-digit dollars per car per year gross profit per unit similar to Surround ADAS three OEM selections; launches 2028
Surround ADAS (EyeQ6 High) roughly 2× base (2–3× in the same exchange’s framing) similar to base three OEMs; 2028 launches
SuperVision (EyeQ6 High) above $1,000 per car closer to the 40s (percent) VW Group, 2027
Chauffeur (EyeQ6 High) ≈ $2,500 closer to the 40s (percent) VW Group, 2028, industry-first consumer Level 3

Management’s own summary line on the L2+ segment: it is the fastest-growing ADAS segment in North America and Europe on Mobileye’s global projections, which keep rising — and the recent wins shift the company in that segment from essentially base-ADAS content toward next-generation positioning.

7. Robotaxi: Buying the Value Chain

  • The rationale, per management: broader participation is there to accelerate self-driving-system deployment and capture more of the value — emphatically “not to pursue fleet ownership for its own sake.” Owning the cycle gives Mobileye control of product definition, software updates, and hardware selection.
  • The optionality motive: owning its own platform lets Mobileye run the sensor-set experiment it wants. The company says its seven years of imaging-radar investment produced what it believes are the strongest in the industry, and its bet is that long term a robotaxi needs “only cameras and imaging radars” — no lidars — cutting the platform’s most expensive hardware, though the timing horizon is unknown.
  • Buildout pieces are landing: Moovit’s rider experience and fleet-orchestration stack is integrated, with initial testing underway in Tel Aviv; vehicle-platform options are in negotiation, including one that enables rapid scaling after initial commercialization; and the single-city U.S. pilot sized at 100 vehicles is set for 2027 (see Highlights) — Orlando area first, with Miami next — covering the self-driving system, vehicle platform, upfitting, operations, fleet management, the rider app, teleoperation, and back office. Expansion beyond comes from OEM level-4-ready platform purchases at scale, defined in parallel with multiple OEMs from 2028 on.
  • Beep is the operations anchor: the investment formalizes a multi-year collaboration, with Beep as the day-to-day fleet management and maintenance partner for the coming years of the U.S. launch — plus an under-evaluation angle in Beep’s public-transit and micro-transit channels as demand creators for autonomous mobility.
  • Operating economics start manually: one teleoperator per vehicle at launch, with vision-language-model assist technologies (shown at previous CES events) slated to raise vehicles-per-operator when integrated.
  • The VW Group robotaxi lane stays on track without resolving the partner noise: about 100 ID. Buzz prototypes are deployed across geographies accumulating driver-out validation and KPI data, with roles and responsibilities being planned across Mobileye, Moia, the vehicle builder, and upfitters. On CARIAD, management drew the line sharply: it is Volkswagen’s structural second source for autonomy, so its next partner “cannot be Mobileye” — its awards sit in the 2030s-era nomination cycle, separate from the nominated programs Mobileye ships.

8. Mentee Robotics: Simulation-First Humanoids

  • Version 3.2 of the robot is operational and performing well, with the next milestone theatrical and technical: a CES stage demonstration performing useful end-to-end tasks rather than just locomotion. Version 4 is in final design on a production-ready architecture and is slated for build around the end of Q1 2027; the production plan calls for a few hundred robots in 2028.
  • The differentiation claim is the development method: a simulation-first pipeline — synthetic data generation, imitation learning, and reinforcement learning — that trains capabilities hard to reproduce from real-world data alone. At the task level, an agentic architecture combines vision-language models with specialized robot policies to decompose complex tasks, execute them as skill sequences, and verify completion. When asked about in-house world-model capabilities following an industry acquisition, Amnon confirmed Mobileye develops a world model “part of our stack.”

9. Conclusion

Mobileye is winning volume in a market where its customers build fewer cars, and paying for that win in price — the blended economics deteriorate with every software-light China-export car, while memory inflation runs through the P&L without margin attached. Management’s answer — climb the content ladder via Surround ADAS, Cloud-Enhanced ADAS plus REM, and the VW Group’s SuperVision and Chauffeur launches — is coherent, deliberately sequenced, and still largely a promise; the year’s operating-income strength is an accounting accelerant layered on a revenue base smaller than last year’s. What makes the quarter investable is the growing, checkable evidence of that choice: more named OEM selections, an extended REM franchise two classes of customer richer, and a robotaxi route with a dated single-city rollout and an operating partner pinned down. The January call — 2027 pricing, the 2027 spending and opex frame, and the R&D-credit refinement coming within weeks — is the first real checkpoint on whether the content story converts into dollars.

10. Artifacts and Follow-Ups

Key Artifacts — highest priority

  1. Q4 execution versus the guided step-down. Baselines: implied fourth-quarter revenue of about $450M (down roughly 8.5% sequentially), adjusted operating income near $60M, and EyeQ orders pointing to 8.8–8.9 million units. Demand again outrunning orders would beat the guide and reset the 2027 baseline; an in-line landing confirms the inventory-reversal thesis.
  2. FY2027 frame at the January call. Baselines: 2027 device pricing expected flat to modestly lower after 2026’s roughly $2 decline, memory pass-through terms still being negotiated, and 2027 investments (GPU farms among them) flagged but unquantified. Watch for a quantified 2027 opex, ASP, and capex path.
  3. R&D credit refinement and SuperVision cadence. Baseline: FY2026 incentive contribution of about $190M at the midpoint, with detailed guidance expected “in the next few weeks”; SuperVision at about 65,000 systems for the year against a still-expected notable Q4 step-down.
  4. Advanced-program production milestones. Baseline: Porsche-led SuperVision reaching maturity in early 2027, and Chauffeur positioned as an industry-first consumer Level 3 behind a 2028 flagship launch. Confirmation of hardening progress and dates would de-risk the ladder’s core rows.
  5. Robotaxi driver-out chain. Baseline: a 100-vehicle, single-city U.S. pilot planned for 2027 (Orlando area, Miami next), Beep named operating partner, and Moovit-integrated testing underway in Tel Aviv; expansion beyond rides on level-4-ready platform purchases defined with OEMs from 2028.

Secondary Tracking Items

  • Buyback pace after roughly $33M repurchased to date at an average $9.36.
  • R&D-incentive law guidance due in the coming weeks, governing how the $180–200M credit range resolves.
  • Chinese OEM export cadence into Q4, with sustained China demand (more than 4.5 million chips) called out by the CFO.
  • Mentee V4 build completion around the end of Q1 2027 and the CES demonstration of useful end-to-end tasks.
  • Further Cloud-Enhanced ADAS and Surround ADAS nominations as production-grade demos open the 2027 business-development window.
  • CARIAD’s replacement-partner selection for VW’s in-house autonomy program and its 2030s nominations cycle.