Every thesis Dave Lee has voiced on TSLA — 45 in all, newest first, with the arguments and risks behind each. Reads as the evolution of their view: the bottom is where they started, the top is where they landed.
The description discloses Tesla ownership, but the transcript offers no Tesla-specific investment thesis. Lee's optimism about AI and humanoid robots concerns societal benefits and adoption rather than Tesla shares or investment returns.
Key arguments
AI could produce substantial productivity gains and discoveries that improve society.
Capable humanoid robots could reduce household burdens and help care for elderly people.
The host hopes humanoid robots become widespread within ten years, without identifying a company or predicting a stock return.
Risks acknowledged
Widespread humanoid robot adoption could take longer than ten years.
Society must address financial pressure and technological displacement before AI's benefits become broadly tangible.
MixedTesla and SpaceX Need to Merge. Here’s Why. w/ Emmet PeppersOct 18, 2026
Dave views Tesla's Cybercab experience and improving FSD as promising, but considers access to leading general intelligence essential to Optimus's long-term competitiveness. He advocates a SpaceX merger and a broader driverless vehicle lineup, while expressing concern about rollout delays, hardware limitations and the durability of Tesla's manufacturing advantage.
Key arguments
The Cybercab experience is compelling enough that large Tesla shareholders could benefit from experiencing it firsthand.
The slow rollout may reflect model and hardware limitations, with V15 and AI 5 potentially enabling more confident expansion.
Tesla's tightly integrated driving hardware, software and real-time requirements provide a defensible competitive position.
Optimus needs advanced intelligence alongside capable hardware; access to a publicly available external model may be insufficient against competitors using superior private models.
Frontier AI laboratories could increasingly apply their intelligence to hardware engineering and manufacturing, weakening Tesla's existing advantage.
A SpaceX merger could combine Tesla's resources with a potentially distinctive route to expanding AI compute through space-based data centers.
Improved FSD could increase vehicle appeal, sales, margins and profits, providing a potential basis for a stock rerating.
Tesla's dependence on Model 3 and Model Y limits its addressable consumer demand.
Dave hypothesizes that Tesla is developing a broader lineup of fully driverless vehicles, but rejects the assertion that autonomy creates literally infinite demand.
Risks acknowledged
Tesla's manufacturing expertise, supply chain and installed production capabilities remain meaningful advantages.
Avoiding the frontier model race has insulated Tesla from some risks associated with rapidly changing leadership.
AI cannot yet independently execute the complex manufacturing projects discussed, and the transition could take years.
SpaceX's compute strategy offers a plausible opportunity rather than a guaranteed victory over leading AI laboratories.
Demand for autonomous vehicles remains sensitive to price.
Dave conditions his response to Emmet's investment thesis on robotaxi scaling and financial results, without adopting the guest's numerical return forecasts.
MixedI Saw Tesla’s Flying Roadster In Person w/ Emmet PeppersOct 16, 2026
Lee connects Tesla's stalled stock performance to stalled revenue and argues that new vehicles generating substantially higher sales could drive a positive valuation reassessment. He praises the company's engineering while emphasizing uncertain robotaxi revenue timing, weak Cybertruck demand, and the intelligence challenges facing Optimus.
Key arguments
Large investors want demonstrated revenue and business results before assigning substantial value to emerging products.
New vehicles and doubled vehicle sales could create operating leverage, improve margins, and support a higher stock valuation.
Robotaxi deployment may progress too slowly to generate the revenue acceleration needed for an immediate positive re-rating.
Optimus requires generalized frontier-level intelligence in addition to manufacturing expertise, capable hardware, and real-time physical-world AI.
Broad household usefulness could take three to four years, conditional on solving those technical requirements; this is an operational estimate rather than a stock-price forecast.
The Roadster demonstration showcases engineering talent and may support recruiting or technology reuse, but the flying version faces major safety and availability constraints.
Cybertruck illustrates the risk that substantial engineering investment does not translate into sufficient demand.
Risks acknowledged
Robotaxi rollout could accelerate beyond the historical deployment pattern.
Tesla's hardware and manufacturing capabilities are significant strengths for humanoid robotics.
Once humanoids become broadly useful, Lee sees very large demand and rapidly expanding applications.
Apparently excessive engineering can prove strategically useful through recruiting or applications in future products.
The guest argues that robotaxi revenue and Optimus optionality could substantially increase Tesla's valuation, but Lee challenges the timing and assumptions.
Dave Lee shares his firsthand experience riding Tesla's Cybercab in Austin, arguing that removing the human driver eliminates a 'zero-sum game' of competing preferences (temperature, music, cleanliness) between rider and driver in traditional ride-hailing. He believes this creates a fundamentally different and superior product category than Uber, not just a cheaper alternative, and predicts most people underestimate this shift until they experience it themselves.
Key arguments
Cybercab removes the tug-of-war over temperature, music, volume, and cleanliness that exists between Uber drivers and passengers
The zero-sum game inherent in ride-hailing (rider comfort vs driver comfort) disappears with an autonomous vehicle
Consistency of experience (same car, same climate, same setup every ride) is an underrated value-add
Beyond lower cost and shorter wait times, the actual in-car experience is a differentiated product, not just cheaper Uber
Scaling depends on regulatory approval and Tesla achieving FSD/robotaxi software (v15) reliability
Risks acknowledged
Scaling is gated by how fast Tesla can get past regulators and achieve required safety/reliability with the software
Dave Lee argues Tesla's robotaxi business is not really competing with Uber but with whichever company can deploy autonomous vehicles at massive scale first. He believes Tesla has a double competitive advantage—AI/software safety plus dramatically lower manufacturing costs—that will let it flood cities with cheap, low-wait-time robotaxis and instantly create its own dominant network, making Uber's driver-based network largely irrelevant for most point-to-point rides.
Key arguments
Robotaxis eliminate the scarcity of human drivers that gave Uber's network its value.
Tesla can manufacture robotaxis (Cyber Cab) far more cheaply than competitors can retrofit or manufacture their own vehicles, creating a durable cost advantage.
First-mover advantage at scale creates instant network effects (lower wait times, lower cost) that are extremely hard for competitors to replicate.
Tesla's existing fleet of millions of privately owned vehicles could be leveraged to extend robotaxi reach into suburbs and outskirts that pure robotaxi fleets can't easily serve.
Risks acknowledged
FSD (specifically V15) may not reach the required safety/reliability bar, delaying the robotaxi rollout by a year or more.
Regulatory approval and safety incidents pose real risks to scaling timeline.
A Chinese competitor or new joint venture could potentially solve the robotaxi problem and challenge Tesla's position, though Dave views this as less likely in the near term.
Tesla's slower growth is interpreted as the cost of concentrating on a rider-focused autonomous platform instead of introducing a cheaper vehicle with driving controls. The host expects consumer sales and a broader affordable lineup eventually, with renewed growth dependent on software, safety, manufacturing, and regulatory execution.
Key arguments
Eliminating driving controls permits a more fundamental redesign and reduces component complexity.
Unboxed assembly, modular construction, and fewer parts could lower manufacturing costs.
Concentrating the company on Cybercab and FSD creates engineering focus and urgency.
An initially controlled robotaxi rollout can validate safety, software, and manufacturing before consumer expansion.
A proven platform could support cars, sedans, and SUVs across several sizes.
The host rejects the idea that Tesla will permanently abandon consumer vehicle sales.
Risks acknowledged
A cheaper conventional Tesla could have supported near-term volume and revenue while autonomy developed.
Autonomous software timing and regulatory approval remain uncertain.
Cybertruck did not provide the expected volume bridge because of cost and demand shortcomings.
Current products have not met Tesla's high-growth expectations.
An all-in strategy increases risk and can expose the stock to substantial losses if execution slips.
Cybercab renews Lee's confidence in Tesla by demonstrating manufacturing and engineering improvements that he expects to extend into future autonomous vehicles. He anticipates a durable passenger-experience advantage, substantial autonomy improvements from AI5, and stronger competitive economics as Tesla's robotaxi network grows.
Key arguments
Six separately assembled modules simplify access, allow concurrent assembly, and reduce manufacturing complexity.
Colored plastic exterior panels eliminate the paint shop and may reduce production costs and factory footprint.
Integrated thermal management, drive units, electronics, and serviceable battery modules simplify assembly and maintenance.
The purpose-built cabin prioritizes comfort and spaciousness for autonomous passengers.
Cybercab innovations provide reusable systems for future vehicle models.
FSD on AI4 has improved substantially despite slower-than-expected robotaxi deployment.
Integrated vehicle hardware, autonomy software, and AI chips create a difficult engineering challenge for competitors.
A larger driverless network could improve availability and reduce prices without human-driver labor costs.
Risks acknowledged
Tesla remains below two million annual vehicle sales despite earlier much larger production ambitions.
FSD and robotaxi rollout have repeatedly disappointed expectations.
AI4 computing limitations restrict the size of deployable models.
Hardware progress alone cannot make a steering-wheel-free vehicle useful without autonomous software.
Aggressive Chinese manufacturers and autonomy developers could compete.
Future product releases and AI5 deployment timing remain uncertain.
Tesla's new affordable model is a shorter-wheelbase Model Y variant, not a stripped trim — analogous to iPhone 16 vs Pro vs Pro Max. Combined with the longer Model Y L (China-launched, six-seat, more upscale), Tesla now has three Model Y SKUs sharing one manufacturing line. Cost reduction $6-$8k unlocks a new SUV demand segment and partly offsets US federal tax-credit expiration, while FSD progress provides a secondary demand pull.
Key arguments
Tesla announced Q2 production start of the new affordable model; launch delayed to Q4 2025 to time around tax credit expiration
The Model Y L (longer) was telegraphed on previous earnings calls as one of multiple new models — and a longer car requires hundreds of design changes, signaling Tesla is willing to invest in real model variants not just trims
Tesla's manufacturing strategy: 'use existing lines, drive extra volume' — a pure cheaper trim doesn't justify multi-quarter engineering work, a shorter platform does
FSD unsupervised rollout (per the companion video) creates compounding demand pull: family members of the speaker bought Teslas because FSD became 'good enough'
Long-term: post-robotaxi, Tesla can revisit the manufacturing platform entirely — sub-$30k unboxed-platform car at 10-20M/year
Risks acknowledged
Skeptics: a $6-10k discount alone may not be enough to drive needed volume
Federal tax credit expiration applies only to ~33% of Tesla sales (US), so impact is bounded
FSD-driven demand uplift won't fully offset credit expiration in the first 1-2 quarters
Unsupervised FSD for retail Tesla owners is closer than the consensus thinks. Tesla's robotaxi program in Austin runs on a more advanced FSD build than the public release — likely thanks to extensive simulation training using LIDAR-mapped streets — and once unmanned robotaxis prove out safely in a region, retail unsupervised follows in that same geofenced area. The Q2 earnings call already telegraphed this for end-2025.
Key arguments
Tesla's robotaxi FSD is materially better than v13.2.9 retail FSD; Dave attributes this to localized simulation training (LIDAR-captured streets fed into Tesla's simulator, hundreds of millions of synthetic rides per region)
Elon stated on the Q2 earnings call that unsupervised FSD in some US regions is the year-end goal
Path: safety-monitored robotaxi → unmanned robotaxi → retail unsupervised in the same geofenced area
Even without regulatory approval, Tesla can collapse nags/monitoring to an effectively unsupervised UX
Demand-side flywheel: family members of Dave actually purchased Teslas because FSD got 'good enough'
Risks acknowledged
Localized simulation could overweight training for specific regions and cause regressions elsewhere — Tesla has to validate before general release
Tesla's robotaxi approach (camera + neural net + per-region simulation) will outscale Whimo's LIDAR+HD-maps stack, with the divergence becoming evident in 1-2 years. Combined with new Model Y variants (long for China, short coming for entry-level), XAI/Grok integration, and methodical India market entry, the multi-year fundamentals remain bullish even amid political-backlash and US tax-credit headwinds.
Key arguments
FSD has improved 'mind-blowingly' since v12 endto-end neural nets; the rate of improvement compounds with data + compute + larger nets
Tesla's robotaxi build runs a *separate, locally-trained* FSD stack vs the retail v13.2.9 — Dave evidence is direct riding experience: 'just feels confident, smooth, a step change'
Whimo's expansion in Austin (90 sq mi, 1500 vehicles, 100M autonomous miles) is irrelevant to long-run scaling because the per-city economics + sensor suite don't generalize the way Tesla's stack does
Post-Tesla-DOGE political cooling reduces TSLA brand-backlash risk; America Party is narrow-scope, not a national majority play
XAI valuation has already skyrocketed — Tesla should still invest a few billion just to access Grok in vehicles + Optimus
Long-term India: a domestic gigafactory deal is the right shape; current $70k import-tariff Model Y is a market-establishing play, not a volume play
Risks acknowledged
Whimo has scale advantage today (100M autonomous miles, 1500 vehicles) — Dave concedes this is 'decent in some ways'
Tesla executive turnover (Troy Jones, Omead Afshar, Drew Baglino) raises stability questions — Dave reframes as Elon's high-bar performance management
Tax-credit expiration + politics are real near-term headwinds
The host interprets Tesla's rapid Austin robotaxi geofence expansion as evidence of confidence in its initial operations. He expresses optimism about the robotaxi service's future but makes no specific forecast or clear investment conviction claim.
Key arguments
The service area expanded into downtown Austin and farther north only three weeks after launch.
The host says the expanded area is approximately the size of Waymo's Austin service area.
Rapid expansion suggests Tesla is satisfied with its initial service area.
Increasing density within a manageable area could improve pickup speed and service quality.
Risks acknowledged
Expanding across more of Austin could lengthen vehicle arrival times and reduce service quality.
Keeping coverage centralized could limit demand from passengers who want to travel anywhere in Austin.
Strongly BullishThe Dangerous Push to Control Elon MuskJul 8, 2025
The host expects Musk's long-term contributions to Tesla to outweigh the costs of his outside activities. He sees FSD and Optimus as central to long-term value creation and argues that shareholders should assess ambitious business milestones rather than annual stock performance or hours worked.
Key arguments
Musk's prior performance-based compensation plans coincided with substantial increases in Tesla's market capitalization.
The host considers Musk's engineering leadership in FSD and Optimus difficult to replace.
Musk remains deeply involved in engineering, design, and strategic decisions.
The host believes Musk changes course quickly when an approach fails.
Long-term performance incentives better align executive accountability with value creation than time tracking.
Annual fund-manager performance incentives can encourage excessive attention to temporary volatility.
Risks acknowledged
Political involvement, including DOGE activity, may have harmed Tesla's brand or demand.
Outside projects can introduce distractions and short-term costs.
The board has a responsibility to address significant harm to Tesla from Musk's outside activities.
A sufficiently innovative replacement CEO is possible, though the host considers finding one unlikely.
Tesla's annual investment returns may be inconsistent.
The host views Tesla's robotaxi as a functioning service with unusually smooth, humanlike driving and strong cost advantages. He expects its rapid software improvement and fleet expansion to produce a substantial competitive lead over Waymo.
Key arguments
Tesla's end-to-end driving system has improved much faster since V12 than the host previously expected.
Tesla's driving feels smoother and more humanlike than Waymo's in these rides.
Cheaper vehicles and hardware, plus operating without an Uber partnership, support lower service costs.
Greater fleet density should improve availability and shorten pickup waits.
Safety monitors provide customer reassurance and a bridge toward fully unoccupied operation.
Risks acknowledged
An attempted left turn stalled and required apparent support intervention.
Pickup availability was limited at one point by high demand.
The robotaxi system may have special software or geofenced optimization that consumer FSD lacks.
Safety monitors remain onboard, and the host does not know Tesla's expansion plans.
Some ride-quality differences could reflect vehicle characteristics.
The host believes Tesla appears to be making sound strategic choices for Optimus and endorses the relevance of its mechanical design, manufacturing and AI experience. The investment narrative is favorable but remains qualitative: the host does not make an attributable stock-price forecast or affirm the guest's specific robot-production projections.
Key arguments
The host highlights Tesla's emphasis on hands, hardware design and appropriate AI software decisions.
He agrees that useful humanoid robots must provide customers with economic value exceeding their cost.
He identifies capable hands as a gateway to using existing tools and performing useful human tasks.
He says Elon Musk appears to understand actuator design, hand performance and the requirements for scaling production.
He compares Tesla's ability to combine existing capabilities for robotics with Apple's combination of software and device experience for the iPhone.
The guest's supporting thesis emphasizes purpose-built actuators, manufacturing scale, internal testing and durable low-cost bodies.
Risks acknowledged
The host notes that investors do not know competing robot hands' actual costs or durability.
The discussion distinguishes attractive demonstrations from reliable hardware suitable for sustained real-world work.
The guest cautions that software advantages and data network effects are uncertain.
The guest stresses that human-equivalent hand capabilities require technologies and supply chains that are still immature.
Specific million-unit forecasts originate with the guest, while the host's dates and quantities are hypothetical questions.
Strongly BullishIs Tesla Robotaxi Ready for Primetime? - with James DoumaJun 28, 2025
The host finds Tesla's Austin Robotaxi service unusually smooth and capable, with everyday driving appearing close to solved despite uncertainty about rare failures. He expects fleet density, low costs, and network effects to support a strong long-term business, conditional on Tesla successfully establishing and expanding the service.
Key arguments
The host reports smooth, confident driving, effective speed-bump handling, and successful navigation through challenging parking lots.
He sees the transition to end-to-end FSD as a major acceleration in Tesla's rate of improvement.
Short, convenient rides suggest broader everyday use for errands, meals, coffee, and recreational trips.
He argues that Tesla's vehicle supply and cost structure could support dense fleets with low prices and short wait times.
He views early market saturation and growing customer demand as a flywheel that raises the hurdle for competing services.
He suggests that accessing owners' cars could improve suburban availability.
Risks acknowledged
The host observes an unusual episode of Robotaxi indecision and notes inconvenient pickup locations.
He explicitly says the available experience does not establish whether rare, unusual driving situations are solved.
The guest challenges the host's emphasis on owner cars as a uniquely suburban advantage, arguing that fleet size is the underlying constraint.
The discussion acknowledges differentiation, lower prices, subsidies, and expanding demand as possible openings for competitors.
The host recognizes that Tesla accidents may attract unusually intense public attention.
Lee sees Tesla robotaxi as a compelling new AI software product whose smooth driving, manufacturing capacity, data, and vertical integration support a substantial competitive advantage. He endorses the guest's broad expectation of eventual appreciation while arguing that expansion should account for rare edge cases and additional safety improvements.
The host views Tesla's robotaxi experience as smoother and more humanlike than Waymo's, with a cost structure and AI development approach suited to large-scale deployment. He expects substantial FSD improvements, expansion beyond the initial service area, and eventual competitive leadership.
Key arguments
Personal rides showed nuanced driving, cautious speeds, and smooth handling of traffic.
Larger neural networks, more and better data, and increased compute should support continued FSD improvement.
Camera-based hardware and an efficient software architecture could enable inexpensive rides and dense fleet coverage.
Tesla controls both hardware and software and has further improvements planned through simulation and AI5.
Better prices and shorter waits could attract usage without acquiring Uber or Lyft.
Risks acknowledged
An onboard safety monitor remains present, and its intervention capabilities are undisclosed.
The initial service is geographically restricted and cannot use freeways.
New software builds can introduce bugs and temporary regressions.
Rare safety interventions require extensive mileage to evaluate.
Deployment timing remains uncertain and depends on Tesla's confidence in safety.
The host views Tesla's early robotaxi service as impressive, with smooth driving, capable parking-lot navigation, and an integrated vehicle and app experience. He expects its lower hardware costs to enable competitive pricing and eventual rapid scaling.
Key arguments
The host reports ten Tesla rides without observed mistakes.
Tesla navigated tight parking lots, performed multi-point turns, and accommodated other vehicles.
Tesla drove into the coffee shop parking lot while Waymo dropped him on a nearby street.
The host considers Tesla's driving slightly smoother and more natural than Waymo's.
Tesla's integrated hardware, software, vehicle, and app create a seamless experience.
The host argues that Tesla's hardware is substantially less expensive than Waymo's.
Risks acknowledged
Tesla still uses an onboard safety monitor.
Waymo is further ahead in operating without a person inside the car.
The host remains uncertain about Tesla's remote assistance and current scaling capability.
The comparison is based on a limited number of rides within Tesla's geofenced launch area.
The host reports five smooth robotaxi rides without interventions and believes Tesla has assembled the core elements of a competitive ride-hailing service. He is bullish on scaling in 2026 but expects safety improvements and removal of the in-car monitor to determine the pace.
Key arguments
All five rides were smooth and required no interventions from the monitor or passenger.
The service uses a standard Model Y without special external equipment.
The app and in-car experience are convenient and largely polished.
The vehicle handled a tight parking lot through a multi-point maneuver.
More vehicles could shorten pickup times, while competitive pricing could support market-share gains.
Launching with monitors allows Tesla to learn from real operations and resolve service issues.
Risks acknowledged
The host does not yet feel fully confident sleeping in a vehicle running his current FSD version.
The actual robotaxi software version and its relative capabilities remain unknown.
An in-car monitor limits the scalability of a fully driverless service.
Drop-off locations can require walking or crossing the street.
Limited fleet size can produce longer waits.
Safety edge cases and safeguards against passenger interference remain works in progress.
The host views Tesla's robotaxi launch as a major milestone following years of investment in full self-driving. He expects gradual deployment and safety-dependent expansion, without making a stock-price forecast.
Key arguments
Tesla has invested billions of dollars and substantial manpower in full self-driving.
The launch marks a transition toward unsupervised FSD.
The Austin service provides a major test run for the robotaxi network.
Risks acknowledged
The initial rollout will be extremely cautious and slow.
Robotaxis will initially have remote oversight.
Tesla must test and resolve problems before broader deployment.
The host views Tesla as a company capable of expanding beyond a mature, low-margin automotive industry into much larger AI transport and labor markets. He expects these businesses to drive future value creation through innovation, data advantages, scalable hardware, and demonstrated execution.
Key arguments
Traditional automotive markets have limited growth, high capital requirements, and thin margins.
AI transport offers innovation-driven competitive advantages and a reinforcing cycle of driving data and model improvements.
Autonomous transport removes human-driver constraints and could substantially reduce transport costs.
Tesla's standardized camera-based hardware could support faster scaling than lidar-dependent approaches.
Optimus addresses a potentially larger market because driving represents only a subset of human labor.
Investors' focus on near-term deliveries during bear markets may leave expansion opportunities underpriced.
Risks acknowledged
The host explicitly states that these outcomes are not guaranteed.
Intense competition from major technology companies could challenge Tesla's expansion.
A compelling vision must be backed by successful execution in new markets.
Tesla's increased vehicle deliveries and record energy storage deployments should improve its financial results and challenge the narrative that it is becoming a low-margin automaker. The host sees these developments as a positive setup for the second half of 2024, subject to execution, demand, and economic conditions.
Key arguments
Q2 deliveries of roughly 444,000 vehicles exceeded production by approximately 33,000 and rose about 57,000 from Q1.
Higher deliveries spread fixed production and operating expenses across more vehicles, supporting stronger margins.
Energy storage deployments reached 9.4 GWh, more than double the previous quarterly record of 4 GWh.
The upcoming robotaxi reveal and ongoing FSD development provide additional areas of investor interest.
Risks acknowledged
Gross and operating margins had declined substantially over preceding quarters.
The prevailing bearish narrative argues that Tesla's growth has stalled and it is becoming a low-margin automaker.
A favorable second half requires strong execution, healthy demand, and a relatively strong economy.
Lee views Tesla's vision-based physical AI and robotaxi development as the foundation for a highly profitable business over the next five years. He expects a potential adoption tipping point once unsupervised driving becomes substantially safer than humans, while acknowledging uncertain timing and current limitations.
Key arguments
Tesla's video-first architecture may offer latency and processing advantages for real-time driving.
Wall Street's stalled-growth narrative could reverse if Tesla restores growth and profitability.
Robotaxi adoption could accelerate sharply once safety exceeds human driving by a substantial margin.
Separately reported robotaxi revenue and margins could make the business opportunity visible to investors.
Optimus could achieve mainstream adoption through sufficiently broad practical usefulness.
A more affordable Cybertruck and additional SUV variants could broaden demand.
Risks acknowledged
Current unsupervised FSD safety has not been proven superior to human driving.
FSD 12.4.1 still exhibits unusual behavior and needs refinement.
Robotaxi rollout, safety improvements and investor recognition could take longer than anticipated.
Automotive margin weakness can constrain near-term stock performance.
Optimus's early-adopter phase could last years.
Cybertruck pricing currently places it beyond many consumers' budgets.
The host is optimistic that Tesla's Optimus robot, if executed well, could leverage over-the-air updates to continuously add new physical skills, making it exponentially more useful and valuable over time, similar to FSD improvements but applied to a generalized humanoid form factor. He believes this could drive massive future demand for the product.
Key arguments
Over-the-air software updates have already radically improved Tesla's FSD, and a similar dynamic could apply to Optimus
Unlike LLMs which only add knowledge, a humanoid robot can add new physical skills and abilities with each update
As the robot's usefulness grows exponentially while hardware cost stays flat, demand could skyrocket in a way rarely seen in other products
A generalized humanoid robot can capture and translate broad AI improvements into real-world physical utility, a gap non-physical AI like ChatGPT cannot fill
Risks acknowledged
Tesla must execute extremely well and has not yet shown a full rollout or production ramp, which is far from guaranteed
Initial adoption could be slow if the robot's usefulness doesn't yet justify its cost
Significant additional work is needed before the robot can perform enough useful tasks to justify its cost
BullishTesla Shareholders Back Elon Musk - Why It MattersJun 21, 2024
The host interprets the landslide shareholder vote as a mandate for Elon Musk to continue leading Tesla aggressively toward an AI and robotics future centered on FSD and Optimus. He sees this governance outcome as removing an overhang and enabling bolder long-term execution, even though near-term financial results have been weaker.
Key arguments
Over 70% of shareholders reapproved the 2018 compensation plan, a landslide showing strong backing for Musk's leadership
Shareholders are betting on the future potential of FSD and Optimus despite minimal current revenue contribution
The vote signals shareholders prioritize long-term growth over short-term pricing, margin and profit headwinds
Reinvigorated backing allows Musk to be more aggressive with growth plans, including a rumored 'master plan part 4'
Texas reincorporation and the vote open the door to a new incentive-based compensation plan and a new share class with super-voting rights, potentially increasing Musk's voting power toward his preferred 25%
Risks acknowledged
Tesla has faced short-term headwinds around vehicle pricing, margins and profit
FSD and Optimus upside depends on reaching unsupervised FSD and Optimus production milestones, which have not yet occurred
Both speakers see Tesla's recent shareholder vote as reinforcing the market's confidence in Elon Musk, FSD, and robotaxi as the company's future, and argue Tesla has a multi-year, possibly insurmountable lead in self-driving and humanoid robotics due to its chip, compute, and data advantages.
Key arguments
The shareholder vote reaffirming Elon's compensation and reapproving him as CEO removes a major overhang and signals confidence in the FSD/robotaxi/Optimus roadmap
Tesla's vision-only, intelligence-first approach to self-driving is seen as the only genuinely scalable solution versus HD-map/geofenced competitors like Waymo
Tesla likely launches a small, controlled robotaxi market within about 12 months, with a possible human-in-the-loop remote monitoring approach initially
Tesla's first-mover advantage, app, and vertically integrated cost structure (lowest cost per mile) could let it dominate ride-hailing once robotaxi scales
Optimus represents a potentially enormous long-term value driver once it becomes useful enough to hit the P&L, compounding through falling manufacturing costs and improving AI capability
Tesla had a roughly six-to-eight year head start by installing neural-net inference hardware in cars years before the software stack was ready
Risks acknowledged
Wall Street has refused to assign valuation to FSD progress despite real-world improvements, wanting proof in financial results first
There could be an unknown technical limiter that caps miles-per-intervention scaling, raising doubts about true scalability
Sandy Munro and others argue stainless steel construction for Cybercab would be too complex and costly to manufacture at scale
The exact timing of both the robotaxi safety crossover and Optimus's path to usefulness remains highly uncertain
Dave Lee uses Tesla's FSD software and future Optimus humanoid robot as his key examples of how embedding human-level intelligence into a product can massively increase its value, from a $220,000 car that can't drive itself to one with transformative added value once it can. He frames being an early tester and observer of this technology as useful for personally evaluating investment opportunities.
Key arguments
A car without human-level intelligence is 'just a car,' but adding self-driving intelligence completely changes its value proposition and what buyers will pay.
Tesla's FSD software is cited as a product he can personally test and evaluate to gauge AI potential.
A future Tesla humanoid robot (Optimus) is described as eventually being able to cook, clean, and farm, illustrating the scale of value AI can add to physical products.
Early identification of companies embedding AI into products/services can yield 10x to 100x returns due to timing.
Risks acknowledged
The competition to bring AI into products and services will be immense, and many companies pursuing this will fail.
There is always risk in investing and one can be wrong in stock selection, requiring contingency planning.
BullishWhy Tesla's Robotaxi Lead is a Game-ChangerJun 16, 2024
Tesla's roughly four-to-five-year lead in generalized autonomous driving, combined with winner-takes-most network effects in ride-hailing, would let it rapidly saturate markets with a cheaper, driverless robotaxi fleet. The host argues this advantage is underappreciated by analysts and creates an extremely high barrier for competitors to overcome once established.
Key arguments
Tesla probably has a four or five-year lead in generalized autonomous driving over competitors
Ride-hailing exhibits winner-takes-most network effects tied to driver/vehicle supply density, similar to Airbnb's property network effect
A driverless Tesla robotaxi fleet can undercut Uber and Lyft on price because it removes human driver costs
Being first to market lets Tesla build fleet density and sunk-cost infrastructure that later entrants cannot economically replicate
Most Tesla-covering analysts are auto analysts who evaluate the current auto business but don't model the robotaxi platform's network-effect potential
Risks acknowledged
Some investors argue robotaxi shouldn't be included in Tesla's valuation because other companies will likely replicate unsupervised FSD within two to three years
The size of Tesla's lead could change year to year depending on which competitors enter the space
The whole thesis depends on Tesla solving unsupervised FSD and the robotaxi being significantly safer than a human driver
Dave Lee argues that common biases toward a slow Tesla robotaxi rollout may be wrong, citing Tesla's massive existing fleet, Tesla Insurance as an enabler of seamless vehicle enrollment, and the fast pace of FSD/AI capability gains as reasons the rollout could surprise to the upside in speed once a safety threshold is crossed. He discusses Cathy Wood's $2,600 five-year price target as context for why the robotaxi timeline question matters, but does not state his own price view or explicit buy/sell stance.
Key arguments
Tesla already has millions of vehicles in major cities worldwide that could be instantly eligible for a robotaxi network once FSD is ready
Tesla Insurance could enable seamless, instant enrollment of consumer vehicles into the robotaxi fleet without third-party insurance friction, and robotaxis should be cheaper to insure than human-driven cars due to lower accident rates
FSD capability has reportedly been improving radically every 6 to 12 months, and once a safety crossover point versus human drivers is passed, adoption could accelerate rapidly rather than grow incrementally
Upcoming hardware and training compute buildout (Hardware 4 maximization in 2025, AI5 chip in 18-24 months) should further boost FSD capability
Risks acknowledged
Tesla has previously vocalized robotaxi expectations that have taken much longer than stated to materialize
Waymo's autonomous taxi rollout in cities like San Francisco and Los Angeles has been slow, which biases observers toward expecting the same for Tesla
It remains unclear exactly when the critical FSD safety crossover point will occur
Enrolling privately owned vehicles into a shared autonomous robotaxi fleet on a flexible basis has never been done before and carries real execution uncertainty
BullishWhy Being First Matters for Tesla OptimusJun 14, 2024
Tesla's combination of from-scratch robot hardware, a multi-year lead in efficient AI inference computing, real-world AI software experience, and scale manufacturing expertise uniquely positions it to build the best and cheapest humanoid robot. Being first to market lets Tesla compound advantages in data collection, cost reduction via economies of scale, and brand mind share, making it very difficult for startups or narrow-use-case competitors to catch up.
Key arguments
Tesla has had to design every part of Optimus (motors, gearboxes, sensors, power electronics) from scratch because no supply chain exists for humanoid robot components.
Tesla's multi-year lead in efficient AI inference computers from its FSD hardware (HW2-HW4, soon HW5/AI5) gives it a hardware brain advantage for robotics.
Tesla's real-world AI experience from full self-driving training transfers directly to Optimus's software/AI model.
Tesla's manufacturing expertise allows it to design for high-volume, low-cost production, which is necessary to make a humanoid robot affordable.
Being first to market compounds advantages: more data improves AI models faster, economies of scale lower prices and increase demand, and market mind share accrues to the leader.
A generalized humanoid robot like Optimus could eventually outpace narrower, specialized robot competitors as its capabilities improve with more data.
Elon Musk is described as an accelerator of innovation whose leadership and ability to push teams is central to Tesla's ability to execute on this opportunity.
Risks acknowledged
Narrow, specialized-use-case robots could reach profitability in smaller markets even without Tesla's generalized approach, at least in the near term.
There are risks involved and it is an immense amount of work required to realize the full potential of Optimus, per Elon Musk's own caveat.
Being first to market alone is insufficient; a company must also sustain a fast pace of innovation or competitors can catch up and overtake it.
Both the host and guest are committed long-term Tesla shareholders who support re-approving Elon Musk's 2018 pay package and moving the company's incorporation to Texas, viewing Musk's continued leadership and expanded voting influence as critical to Tesla's long-term success. The conversation focuses on shareholder-vote mechanics, legal precedent and investor turnout rather than a price or return forecast for the stock.
Key arguments
The guest liquidated essentially all other holdings to concentrate in Tesla in 2020 and considers herself in it for the long term (5-10+ years).
Both speakers believe Elon Musk is the right person to lead Tesla and that his incentive alignment via the compensation plan and increased voting power protects the company from being taken over by short-term or 'bad' actors.
They expect the compensation plan re-approval vote to pass with 50%+ (guest estimates 'optimistic around 60%') of cast votes.
They view the Texas redomestication vote as far harder to win because of the Trip Advisor precedent requiring 50% of all outstanding shares, not just cast votes.
Four of Tesla's ten largest institutional holders have reportedly already indicated support for both measures.
Significant effort has gone into helping retail shareholders (which make up ~40-47.5% of voting power net of insiders) actually cast proxy votes, which the guest sees as decisive to the outcome.
Risks acknowledged
The presiding Delaware judge is viewed as unfavorable and has already ruled against the original 2018 package.
A plaintiff's attorney is seeking roughly $5-6 billion in fees and may oppose the ratification process.
Low awareness and voting-access problems among retail shareholders (especially in Europe and Asia) could cause the Texas vote to fail even if sentiment is broadly supportive.
If the Texas vote fails, Tesla would have to pursue a new Delaware-based compensation plan, which the guest frames as a less desirable 'Plan B'.
The hosts describe FSD v12's shift to an end-to-end neural planner as a surprisingly polished, major capability jump that could reach near- or super-human reliability within the year, and view this as foundational to a robotaxi business they expect to be highly profitable during an early fleet-scaling phase before margins compress as the market commoditizes. They also frame Optimus and Tesla's broader AI/software stack as a longer-term value driver alongside the core vehicle business.
Key arguments
FSD v12's end-to-end neural network removed ~300K lines of heuristic code yet eliminated most prior planning failure modes, surprising both speakers with its polish
Human-mimicry training lets the system generalize nuanced, previously unprogrammed driving behaviors (e.g., not blocking intersections, pulling over for visibility) at a level of subtlety heuristics could never capture
Interventions have flipped from being mostly justified (v11) to mostly being the driver's own fault (v12), suggesting rapid approach toward a threshold where human oversight may no longer be needed
Robotaxi economics are expected to be very profitable in an initial production/regulation-constrained phase, with gross margins compressing over 5+ years as the fleet scales and the market commoditizes, while overall profit dollars grow
Tesla is expected to maintain its retail vehicle business in parallel with robotaxi scaling to hedge against uncertainty in how quickly consumers shift away from private vehicle ownership
Optimus humanoid robot progress is viewed as promising but gated by the timing of industrial scale-up relative to software maturity, with meaningful production likely still some years away
The discussion frames Tesla's long-term identity shift toward being perceived as an AI/software company, analogous to how OpenAI's product launch (ChatGPT) crystallized its identity, though the hosts are unsure whether public/investor perception of Tesla will ever fully shift away from 'car company'
Risks acknowledged
FSD v12 remains feature-incomplete (no parking lot navigation or smart summon at time of recording), leaving open the possibility of undiscovered edge-case problems
A fully driverless, no-safety-driver robotaxi service carrying public passengers is not expected to launch this year; initial testing will involve hundreds of employee-driven vehicles
The pace of improvement from here is uncertain since the slope of V12-specific gains over time has not yet been observed
Some observers expect robotaxis could cause wholesale abandonment of private vehicle ownership, but the speakers view this as unlikely to be the base case, with private ownership persisting as a large share of the market long-term
The hosts note they do not actually know Tesla's internal roadmap or the mix of training techniques being used, and caution that much of their robotaxi/Optimus timeline reasoning is speculative
BullishTesla FSD v12.3 - Hits and MissesMar 17, 2024
Dave Lee is impressed by FSD v12.3's dramatic improvement in natural, human-like driving behavior and reduced interventions versus v11, seeing this rapid progress on the end-to-end neural net approach as encouraging for Tesla's path toward robotaxi capability. He stops short of making any stock price or investment call, focusing purely on product/technology observations.
Key arguments
FSD v12.3 handled multiple tricky real-world scenarios well: a confusing temporary two-way parking lot detour from both directions, parallel parking, yielding to pedestrians, and navigating around a bus and a roundabout
The system feels much more natural and human-like because it's trained end-to-end on video data rather than relying on V11's rigid heuristic-based rules
Interventions, including minor pedal presses, have gone down significantly in his limited testing compared to V11
The jump from early end-to-end development (around December 2022) to this level of capability in roughly a year is described as impressive progress
He believes more compute and more training data will keep improving the system, and highlights the need for 'emergent behaviors' (unexpected good decisions) as a sign of progress toward robotaxi-level driving
Risks acknowledged
Feeling safer is not the same as being safer; only Tesla's internal intervention data can truly validate safety improvement
He experienced a clear fail/disengagement when FSD got stuck too close behind a car
Observed hesitation at a left turn and confusion over a time/day-restricted no-turn-on-red sign
His assessment is based on only a couple hours of personal driving, not large-scale data
Highway driving is suspected to still run on the older V11 stack, not the new V12 approach
BullishTesla FSD v12.3 - First Impressions!Mar 16, 2024
Dave Lee is impressed that FSD v12.3's end-to-end neural network approach already drives more naturally and makes smarter decisions than v11, despite Tesla still being early in scaling data and compute. He sees this rapid improvement as evidence Tesla can keep advancing the system quickly, though he stops short of calling it robotaxi-ready.
Key arguments
v12.3's end-to-end neural net approach produces noticeably more human-like acceleration, braking, and turning than v11's heuristics-based system
v12.3 made smarter decisions in several situations where v11 historically struggled, including a dangerous recurring lane-merge issue in Austin that v11 never fixed
The speed of improvement from v11 to a comparable or better system in about a year suggests Tesla can scale the approach quickly with more data and compute
Tesla's planned buildout of training compute (referenced ~100 exaflops) is framed as key to further rapid gains in FSD capability
Risks acknowledged
v12.3 made some clear mistakes: failed to complete a U-turn, an unsignaled lane change, and GPS/parking-lot confusion
Current small-sample personal testing cannot establish true safety statistics or robotaxi readiness
High-criticality interventions must become extremely rare, which has not yet been demonstrated with real fleet data
The test conditions (low-traffic Sunday afternoon) were not the most challenging scenario for the system
Tesla's Q4 2023 report shows strong long-term potential in Optimus, FSD, and energy storage, but a near-term growth and margin 'limbo' as Model 3/Y demand matures while the next-gen low-cost vehicle slips to H2 2025. The host views this as a manageable transition rather than a fundamental problem, contingent on how management balances volume growth against margin stability.
Key arguments
Optimus development execution is graded very highly, with Elon suggesting possible unit sales as early as next year
FSD v12's end-to-end/neural net approach could meaningfully help sales in the second half of 2024 if it impresses
Energy storage business continues to grow and should contribute more significantly to financials over time, graded A-minus
Next-gen low-cost vehicle delayed to H2 2025 creates an extended limbo period with limited unit growth catalysts for 2024-2025
Model 3/Y have largely saturated demand at current price points, risking further price cuts and margin pressure if volumes are pushed higher
Cybertruck ramp will be slow (management guidance), likely around 50,000 units in 2024 versus a longer-term potential of 250,000/year
Cost of goods sold has fallen over $3,000 per vehicle year-over-year, which could help margins if the trend continues
Tesla has about $29 billion cash and $4.4 billion free cash flow, while increasing capex to over $10 billion for AI/compute investment
Point-of-sale federal tax credit and improved FSD are potential wildcards that could help 2024 demand
Risks acknowledged
Some people think there won't be meaningful humanoid robot demand for another 10 years
Some investors want Tesla to be a better car company rather than an AI/robotics company, viewing rising capex as a negative for cash flow
Falling gross margins concern investors not just in the present but regarding whether and how much margins can improve going forward
BullishTesla FSD v12: The Quest for General Driving IntelligenceJan 23, 2024
The host presents a technical explanation of why Tesla's shift from narrow driving intelligence to general driving intelligence in FSD v12 is significant, expressing optimism about the technology's potential to generalize across novel driving scenarios. He explicitly avoids making timing predictions for unsupervised FSD or robotaxi deployment.
Key arguments
Narrow driving intelligence requires solving each edge case individually with specific algorithms, which never fully resolves the overall problem (described as an endless game of whack-a-mole)
General driving intelligence uses a massive neural network trained on millions of driving videos to learn holistic driving patterns, enabling it to handle novel edge cases it has never seen before
FSD v12's end-to-end neural network approach removes code-based rules and could be the foundation for a true general driving intelligence
If successful, this transition could unlock a new paradigm in autonomous driving capability
Risks acknowledged
The host explicitly states he is not predicting exactly when FSD will become unsupervised or when robotaxis will roll out
The transition's success is conditional ('if Tesla's able to successfully navigate this transition')
BullishTesla FSD v12 Rollout - FIRST THOUGHTSJan 22, 2024
Dave Lee views Tesla's FSD v12 end-to-end neural net rewrite as an unprecedented and highly ambitious technical accomplishment, expressing optimism that compute and data improvements through the year will lead to sizable gains in driving ability. He frames the current limitations as expected growing pains of a necessary, high-risk architectural shift.
Key arguments
Tesla replaced over 300,000 lines of rule-based C code with an end-to-end neural net for planning, a first-of-its-kind shift at this scale
Early FSD 12 testing results were promising enough for Tesla to go all-in on the end-to-end approach
Tesla increased compute investment specifically to support training the end-to-end neural net
The host expects sizable improvements in FSD 12's abilities over the next several months as larger compute clusters come online
Giving a non-employee access to FSD 12 signals Tesla's internal confidence in the current iteration
Risks acknowledged
FSD 12 still stops with 'red hands' and gives up control when confidence drops below a threshold
The jump from rule-based to purely neural-net planning is described as 'almost an impossible task'
It remains unclear how long it will take to reach unsupervised FSD and robotaxi capability
The host believes Tesla's Optimus robot is further along on hardware than people realize and that its AI 'brain' capability will see a big leap in 2024, since humanoid robots can be deployed usefully without needing near-perfect performance (unlike FSD). He frames this as unlocking a massive new market opportunity for Tesla, though he stops short of giving a stock price target or explicit buy/sell call.
Key arguments
Tesla's Optimus hardware is roughly 70-80% of the way to a first truly useful version
Tesla's FSD work has built a real-world foundation AI model that can transfer to humanoid robots
Humanoid robots don't need to be safety-critical-perfect to be useful, unlike self-driving cars, so they can be deployed and improved faster (similar to ChatGPT's iterative improvement via mass usage)
A large and growing list of household/service tasks (cleaning, pet care, elder care, errands, etc.) could create a massive addressable market once baseline competency is reached
The ability of robots to learn new tasks via neural nets (rather than being hard-programmed) gives Optimus near-endless room for improvement and rising value over time
Risks acknowledged
The robot's hardware might be clunky and slow early on
The AI may not be able to pick up every complex task just by observation
Tesla's AI 'brain' for Optimus is only about 10-20% along currently
BullishBeating Wall Street at Their Own GameJan 19, 2024
Dave argues that Tesla's humanoid robot (Optimus) and full self-driving technology are revolutionary products that Wall Street has not yet fully priced because their financial impact hasn't materialized, similar to how the market was slow on the iPhone and Model 3-to-Model Y transition, and he presents this gap as an interesting opportunity for investors who believe these technologies will eventually generate billions in profit.
Key arguments
Wall Street tends to wait for a new product's financial statement impact (Revenue, margins, profits) before fully valuing it, as happened with the iPhone and Tesla's Model 3-to-Model Y transition
Optimus and full self-driving are new technology categories with enormous potential that haven't yet shown up on Tesla's P&L
Early skepticism toward revolutionary companies like Google and Facebook was eventually proven wrong, suggesting similar skepticism toward Tesla's new bets could be misplaced
Risks acknowledged
Hype around a company can outpace its actual worth, meaning the absence of financial statement proof doesn't necessarily mean a stock is undervalued
Many new technologies and so-called revolutionary products never end up panning out, making this inherently a high-risk, speculative approach
The strategy only works if the investor can accurately judge the technology's future, which is described as not an easy task
BullishThe Robots Are Coming... A Billion of ThemJan 18, 2024
Dave Lee believes Tesla is the best-positioned company to lead the humanoid robot revolution because of its existing FSD computer/neural net foundation, top hardware engineering talent, abundant internal use cases (factories, Mega packs), and Elon Musk's vision and execution ability. He sees the potential upside for Tesla as immense if the company executes well on Optimus.
Key arguments
Tesla has been building the 'brain' for robots for years via the FSD computer and real-world AI foundation model
Tesla has some of the best hardware engineers across motors, inverters, and electronics
Tesla has abundant internal use cases to test humanoid robots, such as factory work and unloading trucks
Elon Musk combines foresight, strategy, and execution ambition, exemplified by the Optimus 'thread a needle' goal
Optimus is being designed for mass production rather than as a research project
A simple illustrative model suggests meaningful market cap upside per million robots in service if Tesla becomes a leader
The board can likely reach an incentive agreement with Elon on voting power to keep him leading AI/robotics efforts
Risks acknowledged
Competition for humanoid robots will be intense, likely including nearly every major tech company
Modeling financials off robots is speculative since the product isn't being sold yet
There remain open questions about whether Elon will get the voting power he wants at Tesla
Historical robot prototypes (e.g., Boston Dynamics) have not translated into widespread real-world adoption
The host believes Tesla's Optimus robot already has best-in-class hardware, particularly dexterous hands, and expects rapid progress in autonomy this year, potentially leading to in-factory use by year-end and external sales starting in 2025. He sees this as pointing to enormous long-term value creation for Tesla, though he stops short of a specific stock price or return forecast.
Key arguments
Tesla's Optimus hardware, especially the hands, is already capable of extremely dexterous tasks like threading a needle and folding clothes.
If the hardware can fold clothes, it's likely capable of learning many other tasks via teleoperation data, simulation, and video imitation.
He expects huge jumps in Optimus capability this year because the hardware is ready and only the 'brain' (neural nets) needs to catch up.
He speculates Tesla could use Optimus in its own factories by end of year and potentially start external sales in 2025.
Unlike FSD, Optimus's lower-stakes mundane tasks mean the bar for safety/reliability is lower, so progress could come faster than some expect.
He frames the robot as a general-purpose value-creation machine with 'millions' of potential use cases, implying massive long-term demand.
Risks acknowledged
Some people are biased by how long Tesla's FSD has taken and may be unfairly pessimistic about Optimus's timeline by comparison.
The robot was teleoperated, not autonomous, in the demonstrated laundry-folding video, so true autonomous capability is not yet proven.
He acknowledges his own timeline predictions might be wrong or overly optimistic.
BullishWhy Tesla LEADS the Robot RevolutionNov 9, 2023
Tesla's custom FSD computer, real-world AI foundation model, model-shrinking expertise, and vehicle data pipeline give it a multi-year lead in building the 'brains' needed for humanoid robots. The host argues this positions Tesla to manufacture robots at massive scale with cost and performance advantages, making it the likely leader in the robotics market.
Key arguments
The company with the most cheap, low-energy local inference will win robotics, and Tesla's custom FSD computer already delivers this.
Tesla had to design its own compute hardware because nothing off-the-shelf meets the power/cost/energy requirements.
Tesla is far along on a 'real world foundation model' that understands and navigates the physical world, which is much harder than LLMs.
Tesla has unique experience shrinking and quantizing large models to run locally in cars, a non-trivial capability needed for robots.
Tesla's fleet gives it a continuous data pipeline to keep improving its real-world model, which competitors lack.
Combining hardware, foundation model, data, and compute lets Tesla manufacture robots at scale with lower costs and better performance.
Risks acknowledged
Tesla won't be the only player in the robot market; it will be one of the biggest markets ever with room for others.
Other companies may eventually catch up by waiting for off-the-shelf hardware (e.g., from Nvidia) or building their own stack.
NeutralTesla Stock: How Sell-side and Buy-side Analysts Influence Your InvestmentOct 2, 2023
Tesla's Q3 delivery number of ~435K vehicles came in below sell-side analyst consensus of ~455K, but closely matched the more focused Tesla-specific tracker's estimate of ~441K, suggesting the 'miss' narrative driven by sell-side analysts is overstated. The stock barely moved on the news, which the speaker attributes to institutional (buy-side) investors having already anticipated the planned production dip from factory upgrade shutdowns and instead focusing on longer-term revenue, margin, and macroeconomic trends.
Key arguments
Delivery numbers matter because they are the earliest input into quarterly revenue, gross margin, and operating profit calculations
Sell-side analysts covering many automakers may be less focused and less accurate on Tesla-specific delivery estimates than dedicated Tesla trackers
A Tesla-focused tracker's estimate of 441,000 was much closer to the actual 435,000 delivered than the ~455,000 sell-side consensus
Tesla had already warned in the prior earnings call that production would be down this quarter due to planned factory upgrade shutdowns
Buy-side analysts, who actually manage institutional money and move stock prices, are evaluated on investment performance rather than generating trading commissions, making their (non-public) expectations arguably more representative of true market expectations
Institutional investors, not retail investors or sell-side analyst headlines, are the ones capable of moving a large-cap stock like Tesla significantly over time
Institutional investor outlook is driven more by future revenue, margin trends, and macro factors (like Fed interest rate decisions) than a single quarter's delivery number
Risks acknowledged
Some people are calling the Q3 delivery number a 'big miss' relative to the ~455,000 sell-side consensus estimate
BullishIs this Tesla’s ChatGPT Moment? Inside Look at FSD v12 w/ James DoumaSep 14, 2023
The hosts view Tesla's FSD v12 end-to-end neural network as a potential 'ChatGPT moment' that could dramatically improve driving smoothness and human-likeness, opening a faster capability trajectory toward commercially viable robotaxi service. They discuss probability estimates for technical readiness but frame this as a capability/technology forecast rather than a stock price or investment return call.
Key arguments
V12 replaced ~300,000 lines of heuristic C++ code with a trained neural network end-to-end, enabling back-propagation through the entire control path.
The demo showed markedly smoother, more human-like driving behavior after only a few months of work on the new approach, which impressed both speakers.
Dave Lee states his own probability estimates: ~25% chance FSD is technically capable of unsupervised, safer-than-human driving by end of this year, 50% by end of 2024, and the rest further out — essentially unchanged by the v12 reveal.
James Douma estimates the robotaxi rollout could still take longer due to the need to build confidence and work through a new architecture, but ultimately expects faster capability gains over a 12-18 month horizon.
They argue Tesla's fleet scale, data engine, and compute infrastructure (e.g., 100 exaFLOPS target) give it an advantage that is hard for competitors like OpenAI to replicate quickly for a comparable self-driving stack.
Risks acknowledged
Removing heuristic guardrails could reintroduce old regressions (e.g., the demo's missed traffic light), and it's unclear how quickly or easily these can be fixed without heuristic code.
It's unknown internally whether current intervention rates are already better than v11, worse, or comparable, which clouds the robotaxi timeline.
Regulatory approval, not just technical readiness, may be the actual bottleneck for wider release of a non-beta FSD.
The end-to-end 'pictures to controls' training signal is described as comparatively weak/sample-inefficient versus supervised methods, suggesting large amounts of data and compute are still needed to close remaining gaps.
BullishTesla, GPT-4, FSD - AI is a BIG DEAL w/ James DoumaApr 13, 2023
The hosts view Tesla's investor day manufacturing redesign (modular assembly, dry-electrode batteries, Gen 3 factory) and FSD v11's improved path-planning architecture as major positive developments. They believe Tesla's FSD/robotaxi potential could ultimately dwarf the value of the vehicles themselves.
Key arguments
Modular, front/back/side assembly approach is a radical, well-thought-out departure from decades of conventional auto manufacturing
Dry-electrode battery line is running continuously with minimal human involvement and a dramatically smaller factory footprint
FSD v11's new path-planning (plotting multiple maneuvers and predicting other actors' paths) is a fundamental, significant improvement over prior heuristic-based behavior
Hardware 3 may ultimately prove sufficient for robotaxi-level FSD, since Hardware 4 was made non-retrofittable partly because Tesla is confident Hardware 3 can get there
Tesla has the compute, data, fleet, and manufacturing scale advantages that make it hard for competitors to catch up in the next five years
Risks acknowledged
Hardware 4 not being retrofittable could be read as a negative signal externally, even though it may actually reflect confidence in Hardware 3
FSD v11 still exhibits ugly, occasionally inexplicable failures (e.g., unnecessary turns) and is at 'probably the most dangerous time' due to guardrails being removed
Price cuts versus advertising trade-off could pressure near-term margins
Long-term robotaxi market is likely to be commoditized with multiple regional/global competitors rather than full Tesla dominance
