October 21, 2026 22 min read

Tesla (TSLA) — Q3 FY2026 Analyst Q&A Breakdown

Fiscal period: Q3 FY2026 · Earnings call: October 21, 2026

Twenty-four question units were asked on this call: ten say.com reader questions (numbered Q1–Q10, with two operator-skipped) followed by seven analyst turns, each with a separate follow-up (Q11–Q24). Entries appear in transcript order; say.com questions are attributed to the listener read by Tesla IR.

Q1 — CyberCab beyond Austin and the regulatory path (say.com)

say.com listener, via Travis Axelrod (Head of IR)
The Question: How does the CyberCab fleet scale past Austin, and what is the regulatory path outside Texas — asked specifically against the premise of NHTSA’s open safety review (analyst/listener-provided premise).
Management’s Response: Ashok Elluswamy (VP of AI) said CyberCab launched this quarter by mostly learning from Model Y/Model 3 platform data — “without requiring a ridiculous amount of vehicles to produce the data” — and that “there’s no fundamental reason that we are limited to Austin”: Dallas launches next month, then “the same software will just scale to lots of other cities along with the Model Y robotaxis.”
The “Analyst’s Take”: Expansion-focused and confident, but only half the two-part ask was treated: the NHTSA safety-review premise went completely unaddressed, which is notable given management’s usual insistence on leading with safety data. The Dallas timing marker is the falsifiable piece — check it next month.

Q2 — Optimus training data: crowdsourcing vs Optimus Academy (say.com)

say.com listener, via Travis Axelrod (Head of IR)
The Question: What are management’s thoughts on crowdsourcing training data for Optimus versus the internal “Optimus Academy” approach, and is data currently the primary bottleneck?
Management’s Response: Elon Musk (CEO) said crowdsourcing “is not the issue” — parts of the AI architecture remain to be sorted out — and since Optimus lacks the car fleet’s data flywheel, Tesla must “make… on the order of 10,000 or 20,000 bots” and have them self-play in an Optimus Academy environment to close the loop. Ashok added Optimus learns first by watching humans, then practicing in simulation and the real world — “pretty close to having very good results in this big loop.” Musk’s extended add-on (same unit): a non-canned Optimus video in about a month, believable in “maybe a month or two”; AI5 integration into Optimus “approximately April of next year” with an order of magnitude more compute than AI4; and all pieces converging so external customer shipments — initially industrial, dangerous, repetitive, around-the-clock tasks — arrive about a year from now.
The “Analyst’s Take”: Unusually complete for an Optimus answer: the AI-architecture admission is the falsifiable part, and the response ends with a stacked, testable ladder (video, AI5 date, external-ship window). Data is framed as a self-bootstrapping problem to be solved by scale, not a near-term bottleneck.

Q3 — Optimus scaling (say.com) — operator-skipped

say.com listener, via Travis Axelrod (Head of IR)
The Question: A follow-on say.com ask about Optimus scaling.
Management’s Response: None — Travis announced the skip (“It’s about Optimus scaling, so we can go past that one. Just answered that one, basically”), pointing to Q2.
The “Analyst’s Take”: The skip conflates the training-data question with the scaling question: Q2 covered the learning loop, not a production cadence. Near-term volume specifics therefore moved through this call only via the prepared script.

Q4 — When FSD unsupervised reaches customer cars (say.com)

say.com listener, via Travis Axelrod (Head of IR)
The Question: When can we anticipate FSD unsupervised in customer-owned cars?
Management’s Response: Ashok said the software running CyberCab and Model Y robotaxis is the same and ships to customers “pretty soon” as part of the V15 release; the additional requirement is enrolling customer vehicles in Tesla’s geofences and fleet management. “We suspect somewhere around Q2 next year is when we can turn on some of the first customer vehicles to be part of the robotaxi fleet.”
The “Analyst’s Take”: The cleanest near-dated commitment of the call: customer-owned cars joining the robotaxi fleet now carries a quarter (Q2 2027). V15 is doing triple duty on this call — robotaxi software, mapping/nav, unsupervised enabling — which concentrates schedule risk in one release.

Q5 — CyberCab customer ownership (say.com)

say.com listener, via Travis Axelrod (Head of IR)
The Question: When will the CyberCab be available for customer purchase/ownership?
Management’s Response: Lars Moravy (VP of Vehicle Engineering) said customers get CyberCab “by the end of this year, 2026,” with current Giga Texas output committed to Tesla’s internal robotaxi fleet and CyberCab positioned as the first “fully designed autonomous vehicle” sold to customers. Musk added the framing: CyberCab is “primarily… offered as transport as a service,” mostly Tesla-owned, with Model 3, Model Y and the Roadster remaining the own-car lineup.
The “Analyst’s Take”: Direct, and quietly the most commercially clarifying exchange in the call: Tesla intends to retain most CyberCabs itself, which puts the consumer CyberCab purchases in a learning-cohort frame rather than a volume-statement frame.

Q6 — Robotaxi cost per paid mile (say.com)

say.com listener, via Travis Axelrod (Head of IR)
The Question: What are the unit economics per paid robotaxi mile?
Management’s Response: Vaibhav Taneja (CFO) said utilization is improving week by week and kept the company’s cost target on the table without pretending it’s met: “are we at our goal of $0.30 today? No. But do we see a path to get to that? Very, yes.” Supporting new detail: CyberCab rides are currently Austin-only, more cities arrive before year-end, and Tesla is still “dialing in” fleet-operations costs — pickups, drop-offs, cleaning, learning what managing a fleet costs.
The “Analyst’s Take”: A rare concrete admission structure: goal affirmed, shortfall stated, mechanism (operations scaling) named. The operations-cost framing is the genuinely new data — margins here depend on logistics, not just the car.

Q7 — Optimus unit cost (say.com)

say.com listener, via Travis Axelrod (Head of IR)
The Question: What is the cost per Optimus unit?
Management’s Response: Vaibhav said internal cost tracking is going well as the design matures, but it is “a little too early” to claim arrival given continuing design iterations; the one anchor offered: Optimus cost is “trending closer to where we would say our CyberCab costs would end up being.”
The “Analyst’s Take”: No absolute dollar figure, but the CyberCab-parity anchor effectively cross-links the two robotics cost curves — implying Tesla expects both products to converge toward the same manufacturing economics. Treat as directional.

Q8 — Optimus revenue generation timing (say.com) — operator-skipped

say.com listener, via Travis Axelrod (Head of IR)
The Question: When do you anticipate Optimus will generate revenue?
Management’s Response: None — Travis declared it covered (“we’ve kind of already answered this… So we covered that in the prior question”).
The “Analyst’s Take”: The operator’s judgment is defensible but compresses an important distinction: Q2/Q3 covered cost and data, not revenue. The commercial window implicit in the prepared script is roughly a year out; no one re-stated it here.

Q9 — FSD mapping problems (say.com)

say.com listener, via Travis Axelrod (Head of IR)
The Question: Mapping keeps producing problems — the three named are recognizing proper speed limits, parking, and typical routes; why not build Tesla’s own maps or crowd trusted, verified Tesla owners?
Management’s Response: Ashok conceded the premise outright (“you’re totally right that we shouldn’t be depending on external people’s maps”) and described an in-house replacement: “an entirely new architecture with essentially AI-based mapping and navigation,” voice-integrated, learning both personal and fleet preferences from collective car intelligence. Trip start and end will use it “pretty soon” in V15, and over the next several months it takes over navigation, parking and destination handling.
The “Analyst’s Take”: A calm concession that maps remain a real dependency, answered with architecture rather than a date. Note the load this puts on one release: V15 simultaneously carries the robotaxi stack to customers, the new mapping layer, and unsupervised-capable software — making it the single most important release to track into next quarter.

Q10 — Production rates at the four existing factories (say.com)

say.com listener, via Travis Axelrod (Head of IR)
The Question: What is the current production rate at the four existing factories, and will Tesla build new ones?
Management’s Response: Lars said factories are “operating at record rates this quarter,” running at “the rate of the least lucky process or component,” with last year’s strategic rate investments starting to pay off and expected to continue. New capacity is real and enumerated: CyberCab ramping at Giga Texas, Berlin expansion permits applied for, the new Reno Semi factory just started. Lars closed by recommitting to the 5-million-vehicles-by-2030 ambition.
The “Analyst’s Take”: Confirms the production-constrained story from prepared remarks but without a single output number — investors get direction (rates inflecting up on prior investment) rather than measurement.

Q11 — FSD competitive response and pricing durability — Tom, RBC

Tom, RBC
The Question: Tom’s math puts the paid North American attach rate around 30% to 40% (analyst-provided; “according to my math,” and already described as far above other automakers’ subscription autonomy). With FSD becoming the flywheel driving car sales, could rivals partnering with NVIDIA or Qualcomm on L2+ systems make FSD pricing overly deflationary — Tesla a “victim of your own success”?
Management’s Response: Vaibhav framed the puzzle as simply “hard” — Musk interjected “Super hard problem” — and Ashok supplied the moat argument: the first 90% makes a demo, but the long tail is “so long that… anyone with their sane mind should not attempt” it, and reaching safety “for millions of miles with no injuries” is the real bar; once unsupervised FSD hits the broader customer fleet, “I don’t think anyone can match it in several years.” Vaibhav’s add is the new data: showroom demand has inverted — customers now ask for “FSD drives” rather than asking about Model 3/Y or Cybertruck, recovering independence for buyers with driving limitations.
The “Analyst’s Take”: Deflection-by-moat: the 30–40% figure was neither confirmed nor disputed, and no pricing view was offered — the answer structures competition as an engineering gap rather than a pricing dynamic. The showroom-inversion anecdote is the substantive addition.

Q12 — China: when autonomy is table stakes — Tom, RBC (follow-up)

Tom, RBC
The Question: Follow-up on China, where autonomy arguably reads as table stakes: do Chinese OEMs face the same engineering impossibility, or is a different pricing dynamic at work there?
Management’s Response: Vaibhav said no Chinese OEM has “the leapfrog” in this technology today, while acknowledging Tesla is “working very diligently” — “it is hard” — for broader FSD approval in China, with FSD already operating on certain cars there. Two new assets cited: competitor CEOs have come to the US for FSD drives and “themselves have been blown [away]”; and Tesla has now published its unsupervised robotaxi safety report, so “any regulator who looks at the data… it would be very hard for them to refute it.”
The “Analyst’s Take”: More candid on approval difficulty than the average engagement (“it is hard”), and the safety-report-as-regulator-asset is a genuinely new advocacy lever. Still no timeline for Chinese approval — the CEO anecdotes substitute for progress markers.

Q13 — Robotaxi revenue and profitability triangulation — Emmanuel Rosner, Wolfe Research

Emmanuel Rosner, Wolfe Research
The Question: On the sustained double-digit week-over-week growth in unsupervised miles: Rosner’s math says holding roughly 15% week-over-week growth would cross 10 million cumulative miles by year-end (analyst-provided). How would that translate into revenue and profitability — “how many miles would you need to generate gross profit?”
Management’s Response: Vaibhav’s answer was timing, not economics: “using exponential math, this would become a significant part of our revenue and profitability in ‘27.” Musk sharpened the cadence: the business “is not… going to move the needle this quarter,” becomes “noticeable probably in Q2,” and then “a big deal by Q3 or Q4 of next year.”
The “Analyst’s Take”: The miles-to-gross-profit crossover was never addressed — a quantitative question got a qualitative, quarter-mapped answer. The Q2/Q3/Q4-2027 ladder is itself a falsifiable commitment, but gross-profit breakeven economics remain unpublished.

Q14 — CyberCab line ramp-up curve — Emmanuel Rosner, Wolfe Research (follow-up)

Emmanuel Rosner, Wolfe Research
The Question: CyberCab line capacity is listed above 125,000 units a year (analyst-cited from Tesla’s published listing), and production just started — what does the ramp curve look like?
Management’s Response: Vaibhav praised the engineering result — manufacturing calls it “one of the easiest products… to scale” — but redirected the constraint: “the key thing which we are trying to make sure is that we have all the regulatory approvals where we can start deploying the CyberCabs as fast as we can produce them.” Lars added the mechanical case: 50% fewer parts than any prior Tesla, the line is being “burst[ed]” to find its real rate, and the listed capacity is only what the line can hold today — “the actual capacity of that vehicle line is significantly higher.” Production currently runs at what the robotaxi service needs, with Lars and Ashok “in a bit of a race” between miles growth and vehicle availability.
The “Analyst’s Take”: A qualitative capacity claim without a number or curve; the useful reframe is that regulatory approvals, not manufacturing, are the stated throttle. The race framing turns CyberCab production into a service-metric — watch deployments, not build rates.

Q15 — Energy: distributed power and the Megapod concept — Andrew Percoco, Morgan Stanley

Andrew Percoco, Morgan Stanley
The Question: Megapack demand is clearly strong; how does Tesla see its role in enabling both distributed power and distributed compute — and on last quarter’s “Megapod” allusion, what is the vision and “ultimately how big this could be for Tesla”?
Management’s Response: Musk kept the product deliberately undefined: “It’s still early days on that front. We don’t know the full extent” of how it unfolds; better visibility “probably in… maybe end of Q1 or Q2”; “It does have a lot of potential. It’s just hard to say how much potential right now.”
The “Analyst’s Take”: An honest black box with a revisit clock: no definition, no sizing, but a specific window (around Q1-end/Q2) for the next disclosure. The repetition of “hard to say” twice signals the option is not yet firm enough to defend.

Q16 — Robotaxi bottlenecks: software now, regulatory next — Andrew Percoco, Morgan Stanley (follow-up)

Andrew Percoco, Morgan Stanley
The Question: Having hit an inflection point in Austin since CyberCab launch, are the growth bottlenecks now mostly regulatory rather than software, as Tesla enters incremental regions?
Management’s Response: Vaibhav deferred software specifics to Ashok, who gave the densest operations answer of the call: service hours ladder from 6 AM–9 PM to 10 PM, then 11 PM, about to reach around-the-clock at 2 AM, governed by a zero-incident principle — “even if there’s the slightest possible risk at all, we don’t take on that risk.” By end of 2026 Tesla expects 24/7 operation, higher-speed roads, plus wider weather/night and highway coverage across the breadth network — “We do 6 cities now” — paired with depth in Austin. The pivot is named: “It is technology-driven, but we do see that pretty quickly become regulatory, perhaps in like Q1, Q2.”
The “Analyst’s Take”: Directly responsive and unusually operational: the bottleneck answer is a dated regime shift — technology-gated now, regulatory-gated from around H1 2027. The traffic-light question (importance of the NHTSA-style approvals) is thus answered indirectly: regulation is coming as the binding constraint on schedule.

Q17 — TeraFab capital and milestone path — Edison, Deutsche Bank

Edison, Deutsche Bank
The Question: On the “aspirational goal of 2Q28” for TeraFab: how much capital needs to be invested to get there, and which milestones should investors watch?
Management’s Response: Musk put the first hard frame on the program: “certainly in the tens of billions,” with the SpaceX/Tesla cost split still to be figured out — “it’s… not going to be less than tens of billions in capital equipment… up through early 2028,” and “probably north of $100 billion long-term.” Vaibhav noted most orders for the research fab’s first equipment tranche are already placed; after a garbled exchange over which fab the mid-2027 date belongs to, Musk pinned it: that date is the research fab’s, and “real production would be… late ‘28” — “lightning fast by fab standards.”
The “Analyst’s Take”: The dollar frame is new and large, but the governance question — who funds what — was acknowledged unresolved on air. The Q&A also hardened the prepared aspiration into a schedule (“real production… late ‘28”), which is a subtle but real shift investors should track.

Q18 — Where Optimus design iteration lives — Edison, Deutsche Bank (follow-up)

Edison, Deutsche Bank
The Question: Beyond the promised video, where do the biggest design changes still happen — is it still the hand?
Management’s Response: Musk handed off to Konstantinos (Optimus engineering lead, first appearance on a Tesla call), who answered from the engineering bench: the hand is “one of the most challenging hardware areas,” while most of the electromechanical system “has converged to something that works decently well”; remaining work includes bespoke gears carrying outsized wear and reliability burden. Musk framed the physical problem — humans self-repair, robots don’t — and reiterated AI5 as the intelligence step-change for capability.
The “Analyst’s Take”: The specialists giving the granular read (“converged except the hand”) is high-quality disclosure investors can now track revision over revision. Watch whether “the hand” is still the named constraint next quarter.

Q19 — Robotaxi pricing strategy vs rideshare — Alexander Perry, Bank of America

Alexander Perry, Bank of America
The Question: Robotaxi is priced below competitor rideshare and other robotaxi players in most markets (analyst observation) — is underpricing the long-term market-share plan given the structural cost advantages?
Management’s Response: Vaibhav explicitly repeated the prepared framing — flagging it himself (“like I said in my opening remarks… don’t want to be too repetitive”): the goal is to take over transportation of any kind — plus one new structural point: the rideshare comparison carries a driver while Tesla manufactures its own cars, so “our cost basis overall is different.” Musk escalated: “unless we’re stopped by governments… Tesla will end up doing a majority of all passenger miles on Earth in any country where it is allowed to operate freely.”
The “Analyst’s Take”: The response format honored the question (repeat-script clause included), then Musk converted a pricing probe into a share claim stated as testable (“come back and look at this prediction”). No actual pricing mechanics surfaced.

Q20 — Unsupervised-mile sustainability and fleet composition — Alexander Perry, Bank of America (follow-up)

Alexander Perry, Bank of America
The Question: Unsupervised miles have grown at a high-teens weekly rate since late July (analyst-provided) — will double-digit weekly growth continue through year-end, and what enables it: CyberCab volume or better safety validation on latest-version FSD? Does CyberCab become the majority of the robotaxi fleet over time?
Management’s Response: Ashok: “we will continue the scaling for the foreseeable future. We don’t see any hurdles along the way,” with “a clear path towards scaling to millions of vehicles over the next several months.” Enablers named: multi-year technology investment plus the V15 stack — end-to-end neural networks and learning from humans — with the catch-22 honest framing (“People only see the tip of the iceberg”): years of unseen foundation work, now paying off. The same technology “should scale to millions of vehicles… wherever we are allowed to operate.”
The “Analyst’s Take”: Sustainability answered unequivocally; the fleet-composition sub-question was never addressed, and the primary driver of growth being CyberCab versus FSD-software stayed implicit — the answer credits the stack, not the vehicle. Two of three sub-asks treated.

Q21 — FSD value per car and ARR visibility — Pierre Ferragu, New Street

Pierre Ferragu, New Street
The Question: Building on the at-delivery sign-up rate from prepared remarks, Ferragu’s lifetime math: a car sold with FSD generates roughly $7,000 of gross profit at sale, then more than $1,000 a year from FSD thereafter (analyst-provided) — “it probably at least doubles the profit per car” over the vehicle’s life. Will Tesla disclose FSD’s annual recurring revenue? His own estimate: “probably now approaching $1 billion of ARR. Is that the right order of magnitude?”
Management’s Response: Vaibhav: “the numbers which you shared are correct” — a direct, on-record confirmation of the analyst’s lifetime model — and committed to sharing more FSD metrics as the base scales: “not just the revenue, but also other key metrics around it.”
The “Analyst’s Take”: The most consequential exchange on recurring-revenue quality all call: management attached its credibility to the analyst’s figures rather than offering its own. The disclosure promise creates a new reporting artifact to demand next quarter.

Q22 — FSD subscription retention — Pierre Ferragu, New Street (follow-up)

Pierre Ferragu, New Street
The Question: On retention: how many subscribers stick to the monthly subscription three or six months in, and what share of drivers fall off?
Management’s Response: Vaibhav deferred the numbers — “we will share more metrics around churn rates and whatnot” — with the reason: newly added countries run free-trial cohorts that muddy the read. The interim signal offered: every North American delivery comes with a month of free FSD subscription, “and still people are signing up even though they have that free available.”
The “Analyst’s Take”: A disclosure deferral with a legitimate cohort-confound explanation, plus a usable surrogate (paying against a free month). Retention remains unquantified — watch next quarter for whether the promised churn metric actually arrives.

Q23 — Optimus ramp/training in parallel — Colin Rusch, Oppenheimer

Colin Rusch, Oppenheimer
The Question: Can Tesla ramp Optimus production to start training even before the hands are fully baked with final design — with room to mature simulation in parallel — what is the strategy and CapEx flow?
Management’s Response: Musk confirmed the strategy is exactly that parallel path: “that’s what we’re doing” — building robots without perfected hands, making training progress “and ironing out all aspects of the supply chain and the production system” simultaneously with design and AI work on the giant Fremont line. He added the competitive claim — “I’d be stunned if anyone has a robot that is remotely competitive with Optimus” at delivery, “likely to be towards the end of next year” — and Vaibhav added the capacity proof: an Austin robot factory under construction that “would be able to do a lot more robots as compared to the Fremont one.” Musk’s close: “we’re definitely… betting on the come.”
The “Analyst’s Take”: Confirms deliberate parallel-build rather than sequence-gating on the hand — worth noting because last quarter the Optimus message was manufacturing challenges. The external-delivery timing here hardens the prepared script’s end-of-2027 window into an explicit, twice-stated commitment.

Q24 — Semi supply readiness and the CapEx trajectory — Colin Rusch, Oppenheimer (follow-up)

Colin Rusch, Oppenheimer
The Question: On Semi cell/capacity availability with suppliers: are there more aggressive moves available ahead of the larger ramp, and how should that flow through the CapEx budget — with equipment deposits and deliveries a net positive for TeraFab?
Management’s Response: Musk kept near-term impact limited and reframed spend: “It’s not a massive impact right now. CapEx is actually… ending up being a little lower than we anticipated,” then “more meaningful next year, and then noticeable, like in the tens of billions in ‘28.”
The “Analyst’s Take”: Substantively answers with the capital inflection point rather than supplier mechanics: near-underspend, then a step-change through 2027–2028. No supplier-specific enabling actions were offered; the “tens of billions in 2028” figure is the budget-relevant takeaway.

Synthesis

Clusters

Concern	Analysts (Q#s)	Uniform management answer	Signal
Robotaxi economics and monetization timing	say.com listener (Q6), Emmanuel Q13, Alexander Q19	Cost target affirmed with a named scaling mechanism; monetization deferred to a quarter-mapped ladder in 2027	Economics answers are timing answers — Tesla is buying H2 2027 runway
Optimus reality check	say.com listener (Q2), Edison Q18, Colin Q23	Design except the hand declared converged; parallel build of production, supply chain and AI; external window anchored to end-2027	Consistency across three different askers suggests a rehearsed but honest timeline
Competitive moat vs pricing risk	Tom Q11, Tom Q12, Alexander Q19	Moat framing (long-tail difficulty, structural cost base) substituted for pricing analysis	Management won’t engage pricing as an industry dynamic — treats competition as execution
FSD metrics visibility	Pierre Q21, Pierre Q22	Analyst figures confirmed; new metrics (revenue, churn) promised as the base scales	A disclosure framework is being staged — demand it be fulfilled next quarter
Regulatory dependency	say.com listener (Q1), Andrew Q16, Emmanuel Q14	Technology-gated now, regulatory-gated soon; approvals named as the deployment throttle	The binding constraint on the robotaxi thesis shifts to Washington/Brussels/Beijing around H1 2027
Capital program sizing	Edison Q17, Colin Q24	Near-term spend trending below plan, then a step-change into 2027–2028; ownership split unresolved	Tesla is consciously modeling an inflected capital curve investors haven’t fully priced

Deflections

Topic (Q#)	What was asked	Deflection technique
FSD pricing durability (Q11)	Whether rival L2+ systems compress FSD pricing	Moat restatement (long-tail difficulty) instead of a pricing view; attach figure unaddressed
NHTSA open safety review (Q1)	Regulatory path given the open review	Premise ignored; expansion mechanics substituted
CyberCab ramp curve (Q14)	Quantified ramp from the listed capacity	“Significantly higher” qualitative claim; production-as-service-metric reframe
Megapod sizing (Q15)	How big the distributed-compute product could be	Early-stage black box with a revisit window
Miles-to-gross-profit crossover (Q13)	Economics of unsupervised miles	Quarter-mapped revenue timing instead of unit economics
Retention/churn (Q22)	Subscriber retention rate	Promise-to-disclose deferral; free-trial confound cited
Optimus scaling and revenue (Q3, Q8)	Near-term volume and revenue timing	Operator-side skips on grounds the prepared script covered it

Contradictions

Contradiction	Prepared-remarks claim (source)	Q&A walk-back
TeraFab timing posture	Test chip around May–June 2027 and volume chip around Q3 2028 framed as “aspiration” (Musk, opening remarks)	Q17 hardens it: “real production would be… late ‘28,” and the research-fab vs production-fab division of dates is clarified only under analyst pressure
FSD attach measurement basis	Over 75% of North American deliveries sign up for paid FSD at delivery, up ~20% sequentially (Vaibhav, opening remarks)	Q11 puts a 30–40% North America fleet-level paid attach on record and management neither confirms, disputes, nor reconciles it with the at-delivery figure
Optimus “skip” compression	“At least a year, maybe a little longer” to the 20,000-unit weekly annualized rate (Musk, opening remarks)	Q3 and Q8 operator-skips claim coverage that Q2’s answer (data, not cadence) didn’t actually provide; commercial timing survived only as Q23’s end-of-2027 restatement

Unanswered

Q1 (say.com listener): NHTSA’s open safety review — the regulatory premise was never engaged.
Q11 (Tom, RBC): no view on competitive deflationary pressure on FSD pricing; the 30–40% figure left unreconciled.
Q13 (Emmanuel, Wolfe): no miles-to-gross-profit crossover; no unit economics for unsupervised miles.
Q14 (Emmanuel, Wolfe): no quantified CyberCab ramp curve beyond “significantly higher” than the listed rate.
Q20 (Alexander, BofA): no answer on whether CyberCab becomes the majority of the future robotaxi fleet.
Q22 (Pierre, New Street): no retention/churn number yet — deferred with promised future metrics.
Q17 (Edison, Deutsche Bank): SpaceX/Tesla cost split for TeraFab — acknowledged as unresolved, figures split between two companies without allocation.