Why Tesla's Next 3 Weeks Could Change Everything: NHTSA, Roadster and Starship Calendar

The host sticks with last week's experiment — a ranked weekly roundup for busy Tesla holders — and puts the Thursday night federal signal at the top. The same regulator that launched a probe into the CyberCab a fortnight earlier now says it wants to replace hardware checklists with a performance test for driverless systems. The show frames everything else this week around that pairing: a letter that scores the car against existing rules and a speech that proposes moving the rules toward demonstrated driving ability. Viewers are left wondering how a self-certified design will fare while the benchmark it needs is still unwritten.

A 21-Question Order and the Pedal-Free Puzzle

Dated September 10, the order requires 21 detailed answers by September 30, attested under oath by a company officer and carrying daily penalties near $28,000. It is signed by Chief Counsel Peter Simshauser and sent to Tesla's litigation and regulatory lawyers, making it a formal demand rather than an informal inquiry. Its core asks where Tesla drew the line on which federal standards apply and why it avoided the exemption route, how it satisfied the braking provision that assumes a foot-operated control in a vehicle that has no pedal, and whether provisional driver inputs were installed for certification work and later removed in a way that implicates the ban on disabling required equipment. Electrek's September 15 coverage notes the agency stepping through the items that start to fail without a driver's seat — from telltales and gear indication to turn-signal canceling, mirrors and stability alerts — against a U.S. system where makers declare compliance themselves and are reviewed afterward.

The friction is between a plausible company argument and the regulator's current position. Tesla would argue that provisions written for a human at the wheel should be treated as not relevant for a fully automated trip, while the agency replies that relevance is not for the maker to decide until the code is rewritten, pointing out that its June draft to amend the brake provision remains a proposal and not final law. The most pointed ask concerns the brake clause: the text directs that stopping be triggered via a foot input, yet the CyberCab offers none, and the agency challenges how the declaration was made. A further set of questions recalls a 2022 agency view that selling a vehicle designed to be driven solely by software would normally require a Part 555 exemption; Zoox obtained such relief in July for eight items including braking and mirrors, limited to 2,500 units a year through 2028 after a formal application, whereas Tesla began paid rides in Austin on September 3 without filing. The final question asks plainly how the vehicle can be lawfully sold in that context.

Pittsburgh Remarks and the ASCEND Effort

The second development arrived late Thursday in Pittsburgh, where Administrator Jonathan Morrison described plans for a new federal performance benchmark for automated driving to support wider robotaxi rollout, as relayed by reporter Sawyer Merritt the same evening. Rather than dictating a specific control layout, the agency wants a test of whether the system can drive safely in practice. Morrison noted that much existing code was drafted assuming a person in the driver's seat, leaving future systems tied to past assumptions, and that a Wednesday agreement with SAE creating a consortium named ASCEND will pursue a competency model with measurable behaviors and reproducible evaluation methods. The remarks cited five draft updates to key vehicle standards that would allow novel designs without reducing safety, plus forthcoming guidance for mid-scale driverless operations, described as precision work with a scalpel rather than a broad cut and paired with a reminder that the agency remains the enforcer. The host preserves nuance: the competency concept surfaced in March, the ASCEND idea was introduced in July, and this week the partnership was signed and summarized in a single address. A federal benchmark of this type would sit above a patchwork of state permissions that would otherwise slow each new market, while Tesla's position that the CyberCab falls outside the 2,500-unit ceiling and that its April self-declaration remains operative is noted as unchanged.

Attention then turns to the fleet count in Texas. Monday's tracker snapshot listed 439 Tesla-registered robotaxi units in the state — 390 Model Ys and 49 purpose-built CyberCabs. That is up from roughly 186 a month earlier, an increase of 258 units in 30 days or about a 136 percent rise. The bulk of the gain came in one step on September 2, when the register jumped from 314 to 420 just before the CyberCab launch, a move highlighted at the time by analyst Musconomy. In the most recent week only seven units were added, suggesting the surge was concentrated around launch and the purpose-built tally has since held at 49. On real-world deployment, tracker operator Ethan McKenna estimates roughly 300 of the 439 are operating empty on public roads — a healthy utilization. Distance remains anchored to the one-million mile unsupervised figure disclosed on stage on September 3 by Ashok Elluswamy; the company cites weekly mileage growth above 10 percent, while the host's own math from the last month implies closer to 20 percent, pointing to further accumulation ahead.

Software Update and Three Dates Ahead

Software watch centers on the upcoming FSD revision labeled version 15. On September 11 a driver named Steve described an avoidance maneuver when a vehicle pulled from a parking area; Ashok replied with a brief safety note and indicated that earlier hazard anticipation, quicker response and markedly stronger collision avoidance are slated for the next major release. A week prior, in response to a request for continuous robotaxi service, Ashok suggested the upcoming technology integration for version 15 could make round-the-clock operation possible soon, which — as a reply to an owner's car — implies the update will also reach consumer vehicles. No customer rollout timing was offered, but the host still tags the exchange as encouraging.

The calendar's first entry is October 1 for the Roadster. Tesla posted a rear view under Launchpad Lightning stamped 10.01 with a go for launch caption; Elon reshared it noting the unveil date and adding that it is out of this world. Invites were sent the same day to reservation holders for an evening gathering at 8:30 Eastern in Waco, Texas. The city was not explained, yet Waco borders SpaceX's McGregor test area and earlier reporting had linked the site to a thruster demonstration. At this week's All-In gathering Elon said excitement is assured though success is not, and that a live audience is needed to confirm the exhibit is not synthetic; moderator Jason Calacanis reported seeing a preview and describing it as akin to assimilation. The host explains the relevance: the car represents the first joint hardware artifact — rocket-derived cold-gas thrusters applied to an automobile — which gained extra weight after Elon declined to rule out a corporate combination. Still, the show cautions that a demonstration is not a delivery, with volume output previously guided to 2027 or 2028 and this being the fifth announced reveal date this year.

The second entry is September 24 for the Nevada Semi plant. The Tesla Semi account had flagged a Nevada inauguration next month in August, and current invites brand it the semi rollout with tours, rides and a refreshed truck for fleet guests and media. The first unit from the high-volume line was completed April 29, so this is a formal opening rather than a first build. More substantive Semi news came Monday at Hanover's IAA commercial-vehicle expo, where program head Dan Priestley outlined capacity: 1,000 units per week, 50,000 per year from a single Nevada base serving both North America and Europe. Fleet totals already exceed 28 million kilometers in North America, with 2024 uptime at 98 percent versus 95 percent two years earlier — rising reliability alongside a larger fleet, the metric fleet buyers watch. Tesla Europe's September 11 spec sheet added that the truck is headed to Europe with up to 550 kilometers of range when fully loaded at 40 tons, consumption near 1 kilowatt-hour per kilometer, and an 800-kilowatt mega-charging option that restores 60 percent of range in half an hour; the 40-ton figure is the continental legal limit for a full rig.

Orbital Schedule and a Multi-Trillion Ladder

The third entry is September 28 for an orbital Starship flight tied to Starlink V3. Details circulated Tuesday afternoon by Sawyer indicate a mission carrying 26 next-generation satellites, about ten times the load of a current Falcon 9 Starlink launch, characterized internally as a revenue-generating demonstration. Elon noted the V3 deployment is slated to begin this month and, once complete, could provide more than a hundredfold increase over the existing roughly 11,000-satellite network's capacity. Analyst Sam Korus of Arc revisited his June estimate — roughly $19 million of connectivity income per terabit per second and about 61 terabits per second per fully loaded Starship — implying close to $1 billion per flight, while also observing that per-terabit pricing should naturally decline as supply expands. Tuesday evening Cathie Wood suggested each Starship flight could yield $1 billion, with Elon's ambition of 10,000 annual flights translating to $10 trillion in yearly Starship-related turnover, rendering a $1.75 trillion listing valuation attractive in retrospect; Elon replied that such math is not impossible. The host stresses separating layers: the hundredfold figure describes a completed constellation, the billion-per-flight math assumes a 61-satellite full manifest while this mission carries 26, and the 10,000-flight pace stems from an August post envisioning more than 30 Starship launches per day by 2030, against just 13 total flights to date and a six-day slip this week without published reason.

A further slip on the 28th would push the layers above it. The host treats the 26-satellite check-in as a binary, verifiable outcome for the day; the hundredfold capacity, billion-per-flight and ten-trillion aggregates are projections built atop that outcome. The ten-trillion number reflects Wood multiplying the billion-per-flight figure by ten thousand annual flights, while Elon's own 2030 SpaceX turnover estimate sits near one trillion dollars, so the reply is interpreted as a narrow acknowledgment that the arithmetic is not broken rather than a confirmation. The schedule slipped six days on Thursday without explanation, and the host flags any repeat delay as moving the entire ladder. The automotive backdrop tempers enthusiasm: a 52 percent share of a smaller U.S. electric market — where overall demand is down roughly 30 percent and Tesla's own domestic deliveries are off 16 percent while Model Y alone accounts for about a third of national battery-electric sales — leaves Q3 deliveries in two weeks as the decisive read on whether share gains amid contraction signal resilience or simply a slower retreat than Ford, General Motors, Honda and Volkswagen.

Funding the Ground Network and Talking Combination

On financing, CFO Brett Johnson told a Goldman conference on September 10 that a hosting agreement closed earlier this month will add about $1.1 billion per month from December 1, or roughly $13 billion on an annualized basis, lifting confidence toward a $100 billion year-end target. No customer was named; existing arrangements with Anthropic and Google are understood to be near $1 billion and $920 million monthly, making this a third anchor deal and prompting the host to speculate on Microsoft. The rationale for accumulating ground-based compute contracts is to fund the space-based layer: Elon wrote Sunday he is highly confident that Nvidia VR-class AI hardware will operate in orbit next year, advancing a date that in May's filing had been listed as 2028 at the earliest and has now been brought forward three times. At All-In, Gwynne Shotwell argued that orbital real estate is effectively unlimited, that radiators can reject heat directly to the cold of space eliminating chillers and water needs, and that launch mass remains the constraint Starship is meant to solve, presenting SpaceX as equally an AI infrastructure venture as a space enterprise.

The combination narrative resurfaced on stage at All-In on Monday, in footage released Tuesday, when Elon and Shotwell were asked why the two companies remain separate and Elon responded with a wry line about imagining potential moves given unusually deep cooperation across many fronts — delivered sarcastically but interpreted by the host as leaning toward eventual integration. Shotwell added that SpaceX and xAI are converging more quickly than anticipated yet remain distinct for now. No structure, document or timing was offered; possibilities range from a unified vehicle to an acquisition of Tesla by SpaceX at a premium, with implications for holders hinging on an exchange ratio and the share prices prevailing at announcement, neither yet known. What can be observed, the host argues, is behavior already resembling a single hardware organization: thruster development for the Roadster, silicon work and the TerraFab initiative. The first step is a joint research fab on the Giga Texas campus aimed at producing something useful by late 2027 rather than at scale, intended to learn the tooling before a full TerraFab in Grimes County; in Elon's own framing the choice is between building that factory or failing to scale, making the step unavoidable.

Silicon, Share and a New Insurance Read

On silicon, Korean press cited industry sources Tuesday that Samsung's Taylor site has begun pilot output of the AI5 device on a 2-nanometer flow, moved earlier from a November plan and ahead of plan, with yield validation targeted by year end and volume deliveries expected in 2027. The report is not an official company statement but aligns with Elon's public roadmap for sampling this year at both TSMC Arizona and Samsung Texas and ramping next year; AI5 is slated to sit beneath the next CyberCab generation, the forthcoming Optimus revision and the major FSD advance, so staying on plan is the desired signal. On demand, the Wall Street Journal on Saturday, drawing on Motor Intelligence, placed Tesla at 52 percent of U.S. battery-electric sales through August versus 43 percent a year ago, while Tesla's own U.S. volume is down 16 percent and the total U.S. battery-electric market is off about 30 percent, indicating share rose because rivals retrenched faster and Model Y alone represents roughly one in three electrics sold nationally — a share gain within a shrinking pool. On insurance, Zurich's Australian unit began classifying supervised FSD use as a lower-risk attribute this week, the first such move in Australia and the second globally after Lemonade in the United States; no discount was published but informal talk is near 5 percent compared with Lemonade's 50 percent, and the often-cited sevenfold reduction in incidents is drawn from Tesla's own safety publication which Zurich appears to have accepted. Zurich executive Alex Morgan noted that people err, tire and lose focus, Elon amplified the announcement and the host reads it as a template ahead of the European FSD vote on October 6.

Two brief items round out the week before closing implications. A next-generation Optimus was not shown and Elon made no mention of it; Chinese press reported a Tesla contingent arriving in Ningbo Tuesday to begin supplier audits Wednesday for mass-production readiness, with differing unit figures across reports, no 2026 target from Tesla and Elon's July characterization of an extremely slow start at first still standing, mirroring last week's 5,000-unit component order that signals suppliers to prepare and that the host deems plausible given cadence. On connectivity, KDDI in Japan announced Tuesday a commitment to a next-generation service promising high-throughput data and genuine voice coverage in unserved areas from 2027, with Starlink noting an ambition to enable space-based 5G-level speeds. On the other side, a T-Mobile finance chief argued at a conference that the service poses no wireless rivalry at all, contending that signal arriving at the handset would be too weak to penetrate a windshield and quipping that either he is the poorest allocator of capital or the project will not materialize. Two well-informed parties are taking opposite bets while T-Mobile remains Starlink's exclusive U.S. partner through year end, so 2027 will provide the answer.