Tesla Robotaxi Is Expanding Everywhere

Tesla is aggressively scaling its Cybercab and robotaxi program, as evidenced by a record 102 Cybercabs spotted at Giga Texas and growing fleets at Houston and Dallas hubs, signaling a strategic shift from centralized testing to broad field deployment. This expanding output—still in early Q2 ramp-up—reflects growing confidence in both the hardware and the combined FSD software solution, with an emphasis on pushing production vehicles into real-world urban environments. Multiple signals point to imminent FSD approval in China, including aggressive hiring across nine cities and a rebrand to "Tesla Assisted Driving," while recent NHTSA data shows only one Tesla at-fault crash, reinforcing the narrative of FSD as a "de facto Robotaxi." With FSD version 14 expected by year-end, the conversation with manufacturing expert Jeff Lutz frames Tesla’s unconventional production strategy as a deliberate move to scale robotaxis ahead of traditional rollout logic, nearing a "tipping point" where test vehicles and customer-ready units are increasingly mixed.

• ▶ 0:00 A record 102 Cybercabs were spotted at Giga Texas and at Robotaxi hubs across the country (31 in Houston, 50 in Dallas), showing an expanding field presence.
• ▶ 0:25 Multiple signals suggest Tesla FSD approval in China is imminent: aggressive recruitment across nine cities, rumors of approval by Friday, June 19, and a rename to "Tesla Assisted Driving."
• ▶ 0:40 New NHTSA data shows only one Tesla crash (not at fault) while Forbes calls FSD a "de facto Robotaxi"; meanwhile Rivian and Li Auto claim they can match Tesla FSD soon.
• ▶ 1:23 Tesla's FSD version 14 is expected by the end of the year.
• ▶ 1:28 Jeff Lutz is introduced as a guest with supply chain and manufacturing expertise from Motorola, Google, and Lenovo.
• ▶ 1:28 His background makes him well-suited to comment on Tesla's production strategy and scaling Cybercab output.
• ▶ 1:39 The host formally welcomes Jeff Lutz, signaling the conversation will center on Tesla's manufacturing-related developments, particularly Cybercab and robotaxi efforts.
• ▶ 1:41 Jeff sets the agenda, noting "lots to go through" and "a lot of movement actually occurring on a number of fronts," describing the situation as "pretty encouraging."
• ▶ 1:49 Jeff says they will "step through" the news and "interpret what's really happening," framing the discussion as an analytical look beyond surface-level reports.
• ▶ 2:02 A new record of 102 Cybercabs was spotted at Giga Texas, though this is only a small section of the outbound lot, suggesting the overall production operation is far larger.
• ▶ 2:57 The entire ramp is happening in Q2, with production starting near the end of April; being less than two months in, output is transitioning from "dozens to hundreds to thousands."
• ▶ 3:35 Tesla is increasing throughput after improving station yields; new techniques like reaction injection molding add early difficulty, but the rising output indicates Tesla is working through these challenges well.
• ▶ 4:16 Tesla's vehicle production numbers are the focus of analysis, with attention on rising output.
• ▶ 4:19 The speaker questions why Tesla would be putting more vehicles into output.
• ▶ 4:20 The core strategic puzzle is framed: why would Tesla deliberately increase vehicle output?
• ▶ 4:20 Introduces a central contradiction: why would Tesla be increasing output if it has production uncertainty?
• ▶ 4:34 Lays out mass-production logic: when a new product has hardware or software uncertainty, manufacturers slow or stop production and avoid moving units far from the source.
• ▶ 5:25 Contrasts this with Tesla's actual behavior, which deviates from the expected mass-production playbook.
• ▶ 5:25 Tesla's strategy is explicitly not a traditional distribution model, setting it apart from conventional product rollout approaches.
• ▶ 5:28 The core concept is "moving product out to the field," deploying vehicles directly into real-world urban environments rather than keeping them centralized or controlled.
• ▶ 5:30 Upcoming shots of individual cities are meant to illustrate this field deployment in action, showing concrete examples across different urban locations.
• ▶ 5:38 Tesla is simultaneously shipping Cybercab test vehicles and production vehicles into cities.
• ▶ 5:45 Expansion across many cities at once requires both types of vehicles, indicating a broader deployment effort.
• ▶ 5:56 The ratio and sizing of vehicles being moved suggest Tesla is transitioning from pure testing toward a production framework and production flow.
• ▶ 6:06 The current shift is purely in the testing flow, not an end to testing; Tesla will still need vehicles in various cities for validation and data collection.
• ▶ 6:14 The key metric to watch is the ratio of production vehicles to test vehicles, which is beginning to change.
• ▶ 6:20 This ratio is "on its way to flipping," suggesting a transition toward more production deployment, though more evidence is needed to confirm the trend.
• ▶ 6:26 More evidence is needed: Tesla should show more Cybercabs produced without the steering wheel, though a quick field retrofit could likely remove it if necessary.
• ▶ 6:38 Summary of confidence: Tesla is "very confident" about Cybercab hardware, and increasingly confident that the combined hardware-software FSD solution is becoming a reality.
• ▶ 6:51 Growing confidence is shown by Tesla moving vehicles further from the production source, indicating trust to deploy the product into broader real-world environments.
• ▶ 7:19 Houston Robotaxi Hub fleet grew to 31 Cybercabs by June 13, up from 20 a few weeks earlier, with many vehicles still retaining steering wheels while some lots show fully badged, ready-to-go units.
• ▶ 8:10 Tesla is nearing a “tipping point” where test and production vehicles are mixed, evidenced by finished details like front plates, wheel caps, decals, and high-gloss appearance—costly additions unlikely on pure test mules.
• ▶ 10:27 Dallas stands out with 19 badged, ready-to-go vehicles, while sightings expanded to Seattle, Mesa (Arizona), and Las Vegas preparing robotaxi infrastructure; China FSD approval is still expected by Q3 based on accelerated hiring across nine cities.
• ▶ 13:51 Tesla has shifted focus to the beginning and end of the ride, with new end-of-ride options and “chauffeur”/“robo taxi”-style messaging that reinforce computer control.
• ▶ 15:02 Robotaxis get new software releases first, accelerating the test-code-release loop and giving Tesla a real-world testing advantage, though supervised validation still remains necessary.
• ▶ 16:02 Tesla changed the name of FSD in China for regulatory/marketing reasons, evolving from “FS intelligent assisted driving” to “intelligent assisted driving,” then to “Tesla assisted driving.”
• ▶ 16:18 NHTSA's monthly data release shows Tesla has 18 total reported robotaxi incidents, but only 7 are Tesla at-fault, and all are minor (e.g., curb rashes); the newest crash was a rear-end collision while stopped at a red light, not Tesla's fault.
• ▶ 17:20 Even with the latest incident, Tesla's 3-month rolling all-crash rate remains better than the average human baseline, and there has been no major at-fault accident caused by a computer-driven Tesla robotaxi since the program started.
• ▶ 19:31 Direct comparisons between Waymo, Zoox, and Tesla are misleading because miles-driven data isn't public; normalized safety data is required for meaningful evaluation, and robotaxis overall are achieving "superhuman driving" performance.
• ▶ 20:31 Real-world clips show Tesla FSD taking evasive action before the human driver even perceives the danger, demonstrating predictive defensive driving that can avoid head-on collisions.
• ▶ 21:34 FSD's reaction time is measured at 0.13 seconds from partial visual cues (a visible foot), a speed the hosts call "the difference between serious injury and death" and a reason to scale the technology.
• ▶ 22:45 Media coverage has shifted: Forbes calls FSD a "de facto robotaxi" and says version 14.3 feels like Level 4/driverless ADS, while a Toyota bZ4X comparison shows its system is merely lane centering—not equivalent to FSD.
• ▶ 23:44 Tesla's media narrative is shifting from negative clickbait to positive coverage, and the prediction is that the total volume of critical articles will drop because there is less to attack.
• ▶ 24:43 Tesla should now advertise and "grab the baton" since FSD's engineering is essentially done and the technology is strong enough to promote broadly.
• ▶ 25:44 Competitor ADAS systems—Toyota's basic lane centering, GM's Super Cruise, Ford's Blue Cruise—remain far short of FSD, while an Uber autonomous vehicle incident at ▶ 26:29 highlights that even supervised robotaxis can still crash unexpectedly.
• ▶ 27:33 FSD proactively moves over on two-lane highways when cars poke out from neighborhoods, positioning itself to react if they pull out.
• ▶ 27:58 Tesla smoothed out FSD's earlier overly cautious brake-stabbing, but the speaker still prefers that conservative behavior over the collision shown.
• ▶ 28:18 More sensors (LiDAR, cameras, radar) don't solve the problem — the vehicle's core issue was intelligence, not sensor count, so don't over-read the report.
• ▶ 29:02 Rivian's data flywheels now enable major capability jumps, starting with a supervised point-to-point system later this year on Gen 2 vehicles and R2.
• ▶ 29:42 Rivian is partnering with Uber for robotaxi distribution rather than building its own ride-hail network, focusing on tech while leveraging Uber's scale.
• ▶ 30:12 Rivian plans to deploy 50,000 R2 robotaxis with paid rides targeted for 2028, preceded by safety-driver operations.
• ▶ 30:25 Speakers want Rivian to succeed but doubt CEO RJ Scaringe's timeline for achieving unsupervised autonomy.
• ▶ 30:36 Skepticism centers on Rivian claiming to match Tesla without Tesla's massive advantages: 100,000+ H100 GPUs, bit-exact training inference, millions of FSD users, and 10B+ supervised miles.
• ▶ 30:23 Jeeoff raises an open question: Rivian's promise of "Tesla-like supervised FSD" is vague—does it mean Tesla's v12 from 2023, a current version, or just marketing?
• ▶ 31:44 There is no shared understanding of how well autonomous systems perform; we don't know levels of efficacy, and no boilerplate framework exists for evaluating “how good is this system.”
• ▶ 31:55 Federal autonomous vehicle regulations should include measurement standards defining how to measure these systems and their “goodness.”
• ▶ 32:06 With many OEMs and no agreed-upon efficacy metrics, the regulatory process will either become very long or produce confusing data—so the path forward is to narrow down the metrics.
• ▶ 32:18 The key missed interview question would ask Rivian how they are advancing AI at an accelerated pace without heavy investment in compute.
• ▶ 32:33 The more pressing concern is how Rivian can deliver a real-world physical AI solution without major compute investment — that would be genuinely interesting to understand.
• ▶ 32:45 Without at least basic clarity on the underlying approach, Rivian's claims become less credible and harder to believe.
• ▶ 33:13 Real, verifiable metrics for comparing autonomous systems are essentially a black box, even though a proper measurement standard should exist.
• ▶ 33:36 SAE autonomy levels measure automation, not goodness; a high-level system can still be limited or impractical, as with Mercedes' narrow Level 3 certification.
• ▶ 33:54 A genuine autonomous product is defined by whether it "can work anywhere," not by regulatory levels or restricted certifications.
• ▶ 34:03 Tesla’s global, region-by-region approval process is a key bottleneck for FSD/Robotaxi deployment.
• ▶ 34:10 Tesla is intentionally rolling out slowly to avoid incidents, but this cautious pace lacks a clear accelerated path.
• ▶ 34:31 There is a credibility gap: the speaker would buy if Tesla actually delivers, but the unclear approval roadmap undermines conviction.

• ▶ 34:34 Both speakers are enthusiastic about buying a Rivian self-driving product, calling Rivian "a great car."

• ▶ 34:39 Rivian's self-driving must be buyable and physically testable by customers, just like Tesla's FSD has been from its earliest versions.

• ▶ 34:50 There is currently no other U.S. car company with a self-driving system you can buy, so a buyable, testable Rivian system would be exactly what they want.

• ▶ 35:08 In the U.S., advanced systems like Waymo and robotaxi services aren’t actually buyable, and Toyota, Ford, and GM only offer lane-assist ADAS, not point-to-point self-driving.
• ▶ 35:25 Rivian matters because delivering a true point-to-point self-driving system would make it a notable buyable option in the U.S. market.
• ▶ 35:27 In China, BYD offers buyable, advanced ADAS but with major complaints and issues; still, its camera/lidar data collection makes it a company to watch, while other players like Huawei have had serious safety problems.
• ▶ 36:02 Chinese consumers can already buy cars with real driving capabilities today, and Li Auto's CEO claims its system will match Tesla FSD by Q4.
• ▶ 36:20 These Chinese systems use dual Mach M100 chips with 260 TOPS, improving imitation learning by 50%, and companies have their own chip designs and possibly supercomputers.
• ▶ 36:39 Xpeng is likely the leader in China, followed by Huawei and Li Auto, yet all are still about two years behind — which challenges Rivian's "this year" timeline.
• ▶ 36:58 Regulatory approval outside a company's home market is the real signal of autonomy progress, removing home-field advantage.
• ▶ 37:41 Tesla's approvals in Belgium and multiple European countries, and across continents, point to a real generalized AI solution rather than one tailored to specific locations.
• ▶ 38:18 Tesla's product is immediately testable—anyone can walk into a store and request a test drive—showing a fairly decent gap over competitors in a generalized, scalable solution.
• ▶ 39:18 A major SpaceX IPO on Friday initially dominated the news cycle.
• ▶ 39:23 By the weekend, attention had shifted back to Tesla robotaxi and FSD-related news.
• ▶ 39:28 FSD is "starting to percolate again," signaling renewed momentum and more developments ahead.
• ▶ 39:33 FSD will "actually sell Teslas" in a meaningful way during the second half of this year, not just as a minor factor.
• ▶ 39:47 The purchase motivator shifts from typical car comparisons to wanting the car specifically because it has FSD.
• ▶ 40:00 The trend will be visible through FSD subscription numbers and overall sales data.
• ▶ 40:09 Robotaxi rollout has been disappointing: nearly a year after the June 22 launch, service is limited to Austin plus a handful of cars in Dallas and Houston.
• ▶ 40:27 Despite limited deployment, the safety data is proving out the technology's readiness and validating what was shown previously.
• ▶ 40:38 Visible accumulation of Cybercabs indicates production, delivery, or stockpiling in noticeable quantities.
• ▶ 40:43 Rollout expected "sometime this year if not next year," suggesting cautious optimism rather than a firm deadline.
• ▶ 40:43 Host reaffirms the rollout will happen "this year if not next year."
• ▶ 40:46 Host thanks Jeff for being on the show.
• ▶ 40:49 Host plugs Jeff's X account (@thejefflots) and signs off.
• ▶ 40:52 The speaker thanks the audience and references "the Jeff Lots" before closing.
• ▶ 40:54 The speaker announces the creation of a website, calling it “the most comprehensive resource for the Tesla investor.”
• ▶ 41:00 Viewers/listeners are directed to visit the site at herdomm.com.

• ▶ 0:00 Cybercab Production and RoboTaxi News (0:00 - 6:55) - - Weekend news, Jeff Lutz intro, and Cybercab production ramp at Giga Texas.
• ▶ 6:55 Houston Hub, Cybercab Fleets, and FSD 14.3.4 (6:55 - 13:51) - - Production signals, China FSD rumors, and early driving impressions of FSD 14.3.4.
• ▶ 13:51 FSD Refinements and China Naming Change (13:51 - 16:18) - - UI/parking improvements, release cadence, community reaction, and China FSD naming.
• ▶ 16:18 NHTSA Safety Data and Crash Analysis (16:18 - 20:31) - - Robotaxi crash data, at-fault metrics, and normalization challenges.
• ▶ 20:31 Real-World FSD Defensive Driving and Media Shift (20:31 - 23:47) - - Near-miss saves, reaction-time comparisons, Forbes coverage, and Toyota ADAS contrast.
• ▶ 23:47 Media Strategy and Competitor Incidents (23:47 - 27:33) - - Positive coverage predictions, advice to advertise, and Rivian/Uber robotaxi incidents.
• ▶ 27:33 FSD Defensive Behavior, Sensor Limits, and Rivian's Roadmap (27:33 - 41:04) - - Over-cautious braking, sensor caveats, and skepticism about Rivian's autonomy timeline.