Tesla self-certified the Cybercab into AQ26002 — Zoox already wrote the playbook Tesla didn't use

Tesla began commercial Cybercab service in Austin on Sept. 3, 2026, and on the same day notified NHTSA that it had self-certified the vehicle as compliant with all applicable Federal Motor Vehicle Safety Standards. NHTSA responded by opening Audit Query AQ26002 within hours. The OVSC Resume, posted by the Office of Defects Investigation, is blunt: NHTSA will \"examine the process and technical data on which Tesla relied when certifying the Cybercab\" and \"the extent to which Tesla's certification depended on determinations that certain FMVSS are inapplicable to the Cybercab.\"

That phrasing matters because it surfaces the legal mechanism Tesla used to escape the 2,500-vehicle annual cap under Part 555. In an April 2026 interview, Tesla VP of Vehicle Engineering Lars Moravy said the Cybercab \"is not subject to NHTSA's 2,500-vehicle annual production cap for autonomous vehicles\" because Tesla designed it to comply with all existing FMVSS — like a Camry or F-150 — using self-certification, not a petition. NHTSA's AQ26002 now asks whether that \"compliance\" claim is real, or whether Tesla simply declared eight standards inapplicable and walked through the door Zoox opened legitimately.

Five weeks before AQ26002, NHTSA granted Amazon-owned Zoox the first-ever Part 555 commercial exemption for a purpose-built robotaxi, published as Federal Register docket 2026-15485. Zoox's exemption package is the closest comparable Tesla has — and the eight FMVSS Zoox successfully argued are inapplicable to a no-controls robotaxi are the same eight Tesla now has to defend. The Zoox exemption, capped at 2,500 vehicles annually for two years, covers FMVSS Nos. 103 (windshield defrost/defog), 104 (wipers/washers), 108 (lamps and reflective devices), 111 (rear visibility), 135 (light-vehicle brake systems), 201 (interior impact), 205 (glazing) and 208 (occupant crash protection).

The structural difference is procedural. Zoox filed a public petition under 49 CFR 555.6(d) in August 2025, let NHTSA run notice-and-comment, and accepted the 2,500-vehicle annual cap as the price of admission. Tesla skipped the petition, told NHTSA the standards simply did not apply, and self-certified up to ~1,000 Cybercabs. AQ26002 is the regulator asking, in effect: \"If FMVSS 135 (brakes) and 108 (lamps/mirrors) are inapplicable to a robotaxi, why did Zoox petition for the exemption — and why won't you?\"

Tesla confirmed continuous Cybercab production at Giga Texas on its Q1 FY2026 earnings call in April 2026, and in July 2026 disclosed a Cybercab production capacity above 125,000 units per year. Q2 FY2026 capex tells the same story: capex-to-revenue jumped to 20.5% in the quarter ended June 30, 2026, up from 11.1% in Q1 and 8.0% a year earlier — the highest capex intensity Tesla has booked since the Model 3 ramp. CapEx reached $5.79B against revenue of $28.24B, per the 10-Q filed July 23, 2026.

That spend is now underwritten by a fixed supply-chain footprint. Samsung Electro-Mechanics and LG Innotek won all of the Cybercab camera-module business, with each vehicle carrying eight cameras; the initial order is reported at 20,000–30,000 Cybercabs, or 160,000–240,000 camera modules. The compute side is anchored by Tesla's $16.5 billion AI5/AI6 supply pact with Samsung Electronics — 2nm production at Samsung's Texas fab — and shared production at TSMC. LG Innotek also feeds the broader Tesla vision stack, and Cybercab units roll on the in-house 4680 cell, supplied by Samsung SDI and LG Energy Solution for Tesla's other 4680-equipped vehicles.

Cybercab's Sept. 3 launch lands against a robotaxi incumbent that is already a year ahead. Alphabet-owned Waymo operated roughly 4,000 vehicles across 14 U.S. metros as of early September 2026, logging about 500,000 paid public rides and 4 million autonomous miles per week. Tesla's entire robotaxi footprint in Texas — 420 registered autonomous vehicles as of Sept. 2, 2026 — is about one-tenth Waymo's U.S. fleet, and only 45 of those are Cybercabs. The Austin launch was framed as a \"Day 1\" moment, but on volume metrics it is closer to a pilot.

That gap is the reason Tesla chose self-certification over a Part 555 petition. A petition would have capped Cybercab at 2,500 vehicles per year for two years — the same ceiling Zoox accepted. To hit Musk's \"exponential ramp by year-end\" framing and reach the millions-of-robotaxis narrative, Tesla needed the cap to disappear. AQ26002 now threatens the math in both directions: if NHTSA forces Tesla to retroactively petition, Cybercab inherits the 2,500-unit ceiling; if NHTSA accepts the inapplicability argument, Zoox's head start on the petition path becomes the competitive moat instead.

Short-term (days to quarters). The immediate catalysts are procedural: NHTSA typically completes Audit Queries in 30–120 days, and the AQ26002 record already references \"related issues\" beyond the inapplicability test. A negative finding could force a Part 555 petition, which carries a multi-month public comment period and a likely 2,500-unit annual cap. A positive finding validates the self-certification strategy and lifts the regulatory ceiling. Watch for any NHTSA Federal Register notice or Tesla 8-K disclosure of a petition filing; the next Tesla 10-Q (Q3 FY2026) is the first hard surface for management to disclose AQ26002 in risk factors.

Long-term (one to three years). The structural question is whether the U.S. settles on a petition pathway for no-controls AVs — as NHTSA signaled in its June 2025 streamlining of Part 555 — or accepts industry self-certification of inapplicability for purpose-built robotaxis. Zoox has now locked in the petition path; Tesla is litigating the self-certification path. The June 2025 NHTSA proposal to drop the federal brake-pedal requirement for AV-only vehicles, if finalized, would soften the immediate FMVSS friction but not unwind AQ26002. Watch for the final rule and for any follow-on Cybercab recall or stop-sale that AQ26002 may surface.