Archived board for Oct 11, 2026 (before the redesign). See today's board.
New board, live. Daily scan running. Real items carry a date and source. Sample tags are illustrative only.
Musk Orchestra goals & background
Goals & background
Why each program matters, how it got here, and what has to go right. Open this when a daily item touches something you have not thought about in a while. Status as of Oct 9, 2026 (morning scan).
Market values as shown on Ryan's diagram (design session, Oct 2, 2026). They are not refreshed by this page.
Orchestra-level gates (apply to the whole map). Focus: does a written priority exist when two big programs collide? Liquidity: what is TSLA's money, what is SPCX's? Dated items: SPCX Day-90 lock-up tranche eligible Sep 9, affiliate block Sep 10, later releases (checked against the prospectus Oct 7): up to 328.4 million shares on Oct 9 and another 328.4 million on Oct 24, each about 2.5% of shares outstanding, then the 180-day date on Dec 8 (not Dec 9) for the rest of that group, up to 328.4 million or 797.6 million shares depending on the earnings-linked release. A TipRanks claim of 1.3 billion shares after Q3 earnings does not match the prospectus. (SpaceX 424B4, Jun 11; Oct 7 daily) On Sep 14 to 15, Musk declined to rule out a TSLA and SPCX structural action, with no plan announced. (Musk Orchestra board, Sep 10 to 24.) Sep 29, 2026 (filed Oct 5): Tesla signed $30B of new unsecured credit: a $20B three-year delayed-draw term loan, an $8B five-year revolver, and a $2B 364-day revolver, replacing a $5B revolver. Nothing drawn; Tesla does not plan to draw in 2026; it must keep at least $5B of liquidity; the term loan commitment steps down to $10B after one year and ends after 18 months if unused. Use of proceeds: general corporate purposes. Context for TSLA's money, no gate moved. (Tesla 8-K, Sep 29; Oct 5 daily) Oct 6 to 7, 2026: SpaceX is in preliminary talks to borrow about $40B to buy Nvidia GPUs: about $10B of bank loans and $30B of investment-grade debt, Apollo expected to lead, Pimco among lenders in talks, close expected 2027, GPUs likely the collateral. SpaceX reported nearly $40B of debt at the June quarter and its existing bonds trade near junk levels. Not confirmed by SpaceX, Apollo, or Nvidia. SPCX closed $167.60 Oct 7, down about 2.5%. Context for SPCX's money, no gate moved. (FT via Quartz; CNBC, Oct 7; Oct 8 daily and deep dive) Oct 8, 2026: SpaceX agreed to buy Grain Management's nationwide 800 MHz spectrum (up to 14 MHz paired) for Starlink Mobile, price undisclosed, FCC approval needed; AT&T, Verizon, and T-Mobile fell about 7% to 8% after hours. SPCX closed $160.57 Oct 8, down about 4.2%. First lock-up release Oct 9. Context, no gate moved. (Grain release; Via Satellite; MarketWatch, Oct 8; Oct 9 discovery sweep) Oct 9, 2026 (filed Oct 11): The Wall Street Journal reports the Grain price at about $8 billion in cash; not disclosed by either company. Context for SPCX's money. (Fierce Network; Light Reading, Oct 9)
Starship
Starshipunder SPCX · "lifts everything"
Status, Oct 11: New this scan: Musk said Oct 10 at the X Takeover event that Flight 15 is "hopefully in 4-6 weeks," probably a little before Thanksgiving, and will try to catch the ship, which would be the first full recovery of an orbital-class rocket; he hopes Flight 16 reflies that booster and ship. Flight 14's engine issue was engine computer software, he said. Fits the Nov 8 planning date at the early end; no SpaceX date. Context, no gate moved. Oct 10: New this scan: after its engines were removed, Ship 42 rolled back to Massey's for a third cryogenic proof test (reason not given); NASASpaceflight says early-November readiness still fits the Nov 8 planning date, and Booster 22 could static fire as soon as late next week. Ship 40 (Flight 13) returned to Texas after 75 days at sea. Context, no gate moved. Oct 9: New this scan: new CADENA planning slides list Flight 15 for no earlier than Nov 8 (was about Oct 19) and the first Starship Starlink mission, Starlink 30-1 likely from Kennedy, for no earlier than Dec 10 (was Oct 30). Planning dates only, no SpaceX announcement. Small Hurts on cadence. Oct 8: Quiet on hardware: no Flight 15 static fire yet; Booster 22 and Ship 42 are getting engines installed and could be flight-ready by the end of October (NASASpaceflight via Liftoff Watch). Ship 40 from Flight 13 arrived back at Brownsville Oct 7. Demand side: SpaceX's Nicolls said Oct 6 that Starmind will be "the largest consumer of orbital capacity for Starship," with first Starmind launches next year (Context). Oct 7: Quiet day; no Flight 15 static fire reported yet. Flight 14 reached orbit for the first time on Sep 28. Flight 15 (Booster 22, Ship 42) is still provisionally about Oct 19 with no official SpaceX date, and Flight 16 from Kennedy pad 39A is penciled in for no earlier than Oct 30. Falcon ops keep humming. Not yet routine on Starship itself.
Why it matters
Starship is the root capability. Every orbital plan, including Starmind, needs lots of cheap mass to orbit. If it flies often enough that losing a payload is no big deal, experiments like orbital racks can improve quickly. If every flight is a national event, they cannot.
History and progress
Dec 10, 2026 · NET, planning slidesStarship Starlink 30-1, likely from Kennedy in Florida. Moved from Oct 30 in the earlier slides. (LabPadre on CADENA, via Basenor, Oct 8)
Nov 8, 2026 · NET, planning slidesFlight 15 from Starbase, Booster 22 and Ship 42. Moved from about Oct 19. No official SpaceX date. (LabPadre on CADENA, via Basenor, Oct 8)
Oct 10, 2026Musk at the X Takeover event: Flight 15 "hopefully in 4-6 weeks," a little before Thanksgiving, with an attempt to catch the ship, the first full recovery of an orbital-class rocket if it works; Flight 16 hoped to refly that booster and ship. Flight 14 engine issue blamed on engine computer software; heat shield "robust." No SpaceX date. (TeslaNorth; interview transcript, Oct 10)
Oct 9, 2026Ship 42, its six engines removed, rolled back to Massey's for a third cryogenic proof test; reason not given. Booster 22 still in Mega Bay 1, static fire possible as soon as late next week. Ship 40 from Flight 13 returned to Texas after 75 days at sea. (NASASpaceflight, Oct 9)
Oct 8, 2026New CADENA planning slides push Flight 15 from about Oct 19 to no earlier than Nov 8 and Starlink 30-1 from Oct 30 to no earlier than Dec 10. Planning dates, subject to change. (LabPadre on X, via Basenor, Oct 8)
Oct 19, 2026 · superseded targetFlight 15 (now NET Nov 8, see above). As of Oct 8, Booster 22 and Ship 42 are in the Mega Bays for engine installation and could be flight-ready by the end of October; no static fire yet. (NASASpaceflight via Liftoff Watch, Oct 8) Ship catch retargeted toward this flight after Flight 14 had no recovery. (Oct 1 digest; Sep 28 board)
Oct 6, 2026Nicolls at the International Astronautical Congress: "We believe that Starmind will be the largest consumer of orbital capacity for Starship." Demand claim, not cadence. Does not move Gate A. (PCMag, Oct 6)
Oct 2 to 3, 2026SpaceX posted the first public Starlink V3 deployment photo from Flight 14. Confirms the Sep 28 orbital payload release in imagery. Restated 26 V3 sats and high bandwidth claims. Does not move Gate A. (SpaceX on X; Starlust, Oct 2 to 3)
Oct 2, 2026SpaceX flew three Falcons, four landings, and a Dragon docking in under 13 hours. Falcon cadence, not Starship. Does not move Gate A. (SpaceX; TeslaNorth, Oct 2)
Sep 28, 2026Flight 14 lifted off from Starbase Pad 2 around 8:48 AM EDT. The ship reached orbit for the first time, deployed 26 Starlink V3 satellites, then was brought down early after about 1 orbit (plan was about 6) for a controlled Pacific splashdown north of Hawaii. One ship engine shut down early on ascent. The booster splashed down in the Gulf with no tower catch. (Musk Orchestra board log, Sep 28)
Sep 26, 2026FAA issued the modified launch and reentry authorization for Flight 14. (Board log, Sep 28)
Sep 24, 2026Wet dress rehearsal completed on Booster 21 and Ship 41. (Board log, Sep 25)
Sep 17, 2026Flight 14 slipped about 6 days, from no earlier than Sep 22 to Sep 28. (Board log, Sep 17)
Sep 16, 2026SpaceX named Flight 14 as the first planned Earth-orbit attempt, with 26 Starlink V3. (Board log)
History before Sep 2026 is not on the existing boards. Backfill needed.
Current gates (the how)
Partly open
Gate A · Cheap iteration in orbit
Can Starship make flying experimental hardware boring? First orbit and a real payload are checked off. The gate closes when failed payloads are acceptable and cadence is steady, not after one great flight. (Board, Oct 1)
What would change the outlook
Better: A quick Flight 15, reuse, a tower catch, payloads treated like software releases. Worse: Long freezes after every anomaly. If Starship works and Starmind still stalls, the real bottleneck was power and heat, not lift. (Board, Sep 24)
Status, Oct 11: Quiet day; commentary only. Oct 10: Quiet day; commentary only. Oct 9: Quiet day; commentary only. Oct 8: New this scan: at the International Astronautical Congress Oct 6, SpaceX's Michael Nicolls said the first Starmind satellites launch next year, released a "very notional" video of the one-million-satellite network, and gave craft details (up to about 4,000 kg, solar panels up to about 246 feet, about five-year life, retired to graveyard orbits around the sun). FCC review still pending; astronomers objecting. Fits the Q4 2027 target. Nothing new on cooling or radiation, so Gate B stays Shut (Context). Indirect: SpaceX's reported $40B Nvidia chip financing (Oct 6 to 7) is for ground data centers, not Starmind. Oct 7: No new Starmind news; rename recaps only. Indirect: Oct 7 Terafab control news affects the rad-hard chip path (see Terafab). Timing remains: Musk said Aug 25 the space-optimized Nvidia Vera Rubin NVL72 system launches to orbit in Q4 2027 with "significant scale in 2028." Latest power claim is over 250 kW per craft (Oct 1 to 2), up from 150 kW peak on SpaceX's earlier AI1 spec sheet. Google's Suncatcher is still collecting radiation and heat data. SpaceXAI, which runs this work, is being renamed SpaceXSI (Oct 4, Context). Gate B stays Shut.
Why it matters
The bet is that AI computing wants to live where the sun is constant and there is room. Power, heat, radiation, and links between satellites all have to work together. This is the program that turns Starship's lift into a business beyond launches.
History and progress
2027 · plannedGoogle plans two more Suncatcher satellites with laser links to test them working together. (NPR, Oct 1)
Q4 2027 launch, 2028 scale · SpaceX targetSpaceX's own orbital compute. Checked Oct 5: Musk posted Aug 25 that SpaceX and Nvidia designed a space-optimized Vera Rubin NVL72 system "for launch to orbit in Q4 next year, with significant scale in 2028." That squares the Sep 24 board (Q4 2027) with NPR's "as early as 2028" (scale). (Musk on X via Light Reading, Aug 25)
Oct 6, 2026Michael Nicolls at the International Astronautical Congress in Turkey: SpaceX plans to launch the first Starmind satellites next year; "We believe that Starmind will be the largest consumer of orbital capacity for Starship." SpaceX released its first video of the proposed one-million-satellite network, which he called "a very notional concept" that will "change a lot." About half would fly sun-synchronous orbits. Each craft up to about 4,000 kg with solar panels up to about 246 feet, design life about five years, retired to heliocentric graveyard orbits. FCC application still under review. Context, no gate moved. (PCMag, Oct 6)
Oct 4, 2026Musk said SpaceXAI will be renamed SpaceXSI ("super intelligence"), following a Sep 29 executive order on federal terminology. No timeline; legal versus branding change not stated. The division runs SpaceX's orbital compute effort along with Grok, X, Colossus, and Cursor. Name only, no gate moved. (Musk on X; Teslarati, Oct 4)
Oct 1 to 2, 2026Musk said each Star Mind spacecraft gets over 250 kW of solar, about 20% more than the ISS. He also said SpaceX is cautiously optimistic it can run its Nvidia VR72-based system near 250 kW average (peak about 10% higher), crediting co-design with Nvidia, and that it would apply to ground systems too. First concrete power number for SpaceX's own orbital compute craft. Cooling and radiation still unproven. AI1 ~75 m wingspan from @XFreeze is analysis, not SpaceX-confirmed. (Musk on X; TeslaNorth / Benzinga, Oct 2)
Sep 28, 2026Replying to an X thread that put Wall Street's cost for one gigawatt of orbital AI compute at about $160B to $180B (Bank of America about $165B, Wood Mackenzie about $170B, as attributed in the thread), Musk wrote "Will be way better than that by 2028." A claim, not a cost; nobody has run a commercial data center in orbit. (The Inference, Sep 28; noted Oct 5)
Oct 1, 2026Google's prototype launched on a SpaceX Falcon 9 at 1:32 PM CT. Google and Planet confirmed contact that evening. Google says it is operating as expected and will collect radiation and heat data over the coming weeks. (Google blog; Planet release)
Sep 24, 2026Board logged Suncatcher as competitor validation, not disruption: one satellite, four TPUs, about 1 kW. It noted Starcloud already has an H100 chip in orbit and that SpaceX talks gigawatt scale. Also: a terrestrial Google and SpaceX deal of about $920M a month, cancellable on about 90 days. (Board log, Sep 24)
Nov 2025Startup Starcloud launched a spacecraft carrying an Nvidia H100 chip and demonstrated Google's Gemini from space. (NPR, Oct 1, "last November")
Aug 25, 2026SpaceX and Nvidia announced a space-optimized Vera Rubin NVL72 design for first-generation Starmind satellites. SpaceX's AI1 spec rendering: about 20 m tall, about 70 m wingspan, about 150 kW peak power and about 120 kW average compute, around 3 ms latency. SpaceX talks of "up to" 1 million orbital data centers. Musk said in June the AI1 reference design supports Nvidia GB300 and Vera Rubin, with a TPU reference design too. (Light Reading, Aug 25)
Earlier Starmind milestones (before June 2026) are not yet on the board. Backfill needed.
Current gates (the how)
Shut
Gate B · Closed-loop orbital compute
Power, heat rejection, radiation tolerance, and networking must all work together. The test is a small rack running useful work for months without heroic care. Musk's 250 kW power claim fills in one half. Cooling and radiation are still unproven. Suncatcher is a first look, not proof. (Board, Oct 1; Musk via TeslaNorth, Oct 2)
Partly open
Gate A · Cheap iteration in orbit (shared with Starship)
Starmind only improves quickly if launches are cheap enough to try and scrap hardware. See Starship.
Elevated
Gate D · Terafab chips (shared with Terafab)
Rad-hard D3 silicon comes from the JV. See Terafab.
What would change the outlook
Better: Months of useful compute with radiator weight and power that leave room to spare, published as engineering, not renders. Google books Starship. Worse: Suncatcher grows into clusters without needing SpaceX's rockets or lasers, making Google a competitor, not a customer. Radiators prove too heavy. (Board, Sep 24; NPR, Oct 1)
Oct 1 to 2 · Star Mind 250 kW solar / VR72 (Helps) (prior daily)
Oct 1 · Google and Planet confirm contact (prior daily)
Oct 1 · Suncatcher launches on Falcon 9 (prior daily)
Oct 2 · Deep dive: Suncatcher (prior daily; superseded by Oct 3 Terafab dive)
Optimus
Optimusunder TSLA · humanoid at rate
Status, Oct 11: Quiet day. Musk said Oct 10 the robot hand is "shockingly difficult" and production will start slow because no humanoid supply chain exists yet. Restates known bottlenecks; no gate moved. Oct 10: Quiet day for the robot. Musk said Oct 10 that Digital Optimus, the Tesla and xAI software agent, can play about halfway through Diablo by watching the screen; that is the software project, not the humanoid. No gate moved. Oct 9: Quiet day. A Tesla patent application published Oct 8 describes flexible touch sensors for robot fingertips and palms; it does not name Optimus (Humanoids Daily). No gate moved. Oct 8: Quiet day; recaps of the 1,000-a-week year-end goal and hand assembly bottleneck. Chosun Biz says Tesla is hiring dedicated hand manufacturing engineers (secondary). Oct 7: Quiet day; memory-cut recaps only. Indirect: Musk says Tesla and SpaceX will run Terafab themselves (Oct 7), the planned source of AI6-class chips. Oct 6: a Chinese outlet says the Texas Optimus plant's steel frame is about 40% up (secondary, unverified). Reported production is several hundred robots a week at Fremont, but they cannot yet do general work reliably. Tesla has not announced outside customers. Latest real move is still the AI5/AI6 memory cuts for Optimus volume (Oct 1 to 2). Gate C still Shut (no new utilization proof). Indirect read-across from robotaxi night vision work in the Oct 5 deep dive. The Oct 6 deep dive notes hand-plugging robotaxis as a natural internal Optimus job to watch for.
Why it matters
Optimus is the labor bet: a robot that can do dull and dangerous work cheaper than a person. Demos do not count. What counts is robots doing real shifts inside Tesla's own factories, because that is where downtime, safety, and maintenance show up honestly.
History and progress
New The items below come from the Oct 2 build's quick search, not the old boards. None of this was on the Oct 1 boards, which is the kind of gap the rebuild should close. Production figures are reported by The Information, not confirmed by Tesla.
Summer 2027 · reported targetMass production at a dedicated Optimus plant at Giga Texas (about 5.2 million sq ft), long-term goal of 10 million a year. Oct 5 report says drone footage by Joe Tegtmeyer shows the steel frame about 40% complete. Chinese outlet PCauto via Geisou; not confirmed by Tesla. (Geisou, Oct 5)
End of 2026 · reported goalA continuous automated line capable of more than 1,000 robots a week. Longer-term goal of about 20,000 a week. (The Information, Sep 25, via Electrek)
Oct 8, 2026Tesla patent application US20260310299A1 published: flexible, curved touch sensors for robot fingertips, palms, and humanoid skin. Filed Sep 11, 2025. Does not name Optimus or show the design is in use. (Humanoids Daily; Electric-Vehicles.com, Oct 8)
Oct 2, 2026Musk nudged AI5 memory up to 96 GB LP5 so Tesla would not be the only buyer of the min-density LP5 part. (Musk on X; TeslaNorth, Oct 2)
Oct 1, 2026Musk said AI5 cut to 72 GB LP5 (half) and AI6 to 144 GB LP6 (down one-third from 216 GB) as the only way to get Optimus volume and lower cost; bandwidth held constant; claimed negligible performance hit. Supply-chain engineering for Gate C, not proof of unsupervised factory work. Ties lightly to Gate D / Terafab chip path. Missed on the Oct 3 quiet line. (Musk on X; TeslaNorth, Oct 1)
Sep 25, 2026The Information reports Tesla built several hundred Optimus V3 robots a week at Fremont in August, about 10 times Q2. Most are used internally for testing, training, and data collection. Hands and forearms still need workers to assemble 100 or more small parts by hand, and the AI cannot reliably handle a wide range of tasks. (Electrek, Verge, Humanoids Daily, Sep 25)
As of Sep 25, 2026The V3 reveal Musk promised for mid-2026 had not happened yet. (Electrek, Sep 25)
Q2 2026A few dozen robots a week in small-batch testing. (The Information, via Electrek)
Jan 2026Musk said no Optimus robots were doing useful work at Tesla yet. (Electrek, Sep 25 summary)
Jan 2025Musk said Tesla would build about 10,000 Optimus robots in 2025 and that several thousand would be doing useful work by year-end. (Electrek, Sep 25 summary)
Current gates (the how)
Shut
Gate C · Optimus at rate, doing ugly work
The test is sustained, unsupervised hours on dull and dirty tasks inside Tesla or SpaceX facilities. Until the factory trusts the robot over a contractor, outside "humanoid market" talk is narrative. (Board, Oct 1)
Elevated
Gate D · Terafab chips (shared with Terafab)
AI6-class silicon is the planned brain. See Terafab.
What would change the outlook
Better: Quiet replacement of real shifts, with maintenance and safety costs lower than hiring. A first outside customer. The V3 reveal. Worse: Demos without any internal utilization, and a production ramp stuck on hand assembly.
Daily items filed here
Oct 10 · Quiet; Musk calls the hand "shockingly difficult," slow start (today)
Oct 8 · Quiet; hand assembly recaps, hand engineers being hired (secondary) (prior daily)
Oct 7 · Quiet; Terafab control read-across in the deep dive (prior daily)
Oct 6 · Texas Optimus plant steel about 40% up (Flag, secondary) (prior daily)
Oct 1 to 2 · AI5/AI6 memory cuts for Optimus volume (Helps) (prior daily)
Autonomy · Robotaxi
Autonomyunder TSLA · old board archived Oct 1
Status, Oct 11: New this scan: Tesla's Robotaxi account said Oct 10 it has "300+ Cybercabs serving Austin, an 8x increase in 30 days," Tesla's first official Cybercab fleet number, and Ashok Elluswamy wrote "Dallas soon," the first named next Cybercab city (no date). No rides, utilization, or monitor counts given. Small Helps on Gate D (disclosure) and Gate B (paid service at scale). Oct 10: New this scan: in Europe, Tesla renamed FSD (Supervised) "Tesla Assisted Driving" after Germany's transport ministry called the old name misleading; Germany will back EU-wide approval if the software stays within 10% of the speed limit. EU vote December at the earliest; eight EU countries approved, Germany not yet. Supervised product, Context for Gate C. Tesla also posted "Floodgates are opening" over four Cybercabs in Texas traffic (no date, city, or fleet size; paid Cybercab rides still Austin only). Context. Oct 9: New this scan: NHTSA set Oct 30 as the deadline for Tesla's sworn Cybercab answers (Tesla asked for Nov 20; got 30 days from Sep 30), per a public NHTSA memo. Fills in the "new deadline not published" gap; Context on Gate C. Texas Cybercab registrations jumped 150 overnight on Oct 8, from 169 to 319; Model Y robotaxis 420; fleet 739 (Context, supply not rides; paid Cybercab rides still Austin only). Oct 8: New this scan: a drone pilot estimates 190 to 200 Cybercabs at Tesla's Houston lot, up from about 105 on Sep 29 and about 50 in mid-September, with cabs testing on public roads. Paid Cybercab rides are still Austin only and Tesla has not announced Houston (Context, supply not rides). Oct 7: Quiet day. Hours still 6 AM to 11 PM; the Oct 6 software release notes do not mention FSD v15. A volunteer tracker counted 2.2 Cybercabs per 1,000 Austin vehicles on Sunday versus about 3 for Waymo (CyberCab.News, Oct 6, Context). A report that Musk is weighing lidar traces to Electrek commentary, not Musk. Oct 6 scan: a Cybercab got stuck looping a pickup lot for about 7 minutes on Oct 5 and needed a remote stop and a 7-minute software restart (small Hurts, Gate A); a Tesla employee was filmed hand-plugging ten empty robotaxis at a Supercharger (Context, Gate B); California SB 1246, signed Sep 30, regulates remote drivers and first-responder blocking from July 2028 (Context, Gate C). FSD v15 not in public builds as of Oct 5. Carried over: Austin robotaxi hours stretched to 6 AM to 11 PM (announced Oct 2), and Musk named night detection of hard-to-see pets as the main thing left before later hours; 24/7 is tied to FSD v15, reported as early as this month but not Tesla-confirmed. Small Helps on Gate A. Limited paid Cybercab rides in Austin since Sep 9. CNBC Oct 3 reports mixed month-one reviews (long waits, wrong pickups/drop-offs, door/trunk issues) and confirms about 169 Texas Cybercabs authorized. Waymo scale gap is concrete (about 1,154 Texas AVs, more than 500k paid rides/week). NHTSA gave Tesla an extension past the Sep 30 AQ26002 deadline (agency spokesperson via CNBC); new deadline not published. Gate A/B under slight pressure from usage quality; Gate C stays Live with extension Context, not clearance.
Why it matters
The race is who delivers a ride that riders prefer, insurers allow, and regulators clear, at a cost that makes cash before patience runs out. Tesla's bet is cameras only and a cheaper car, about $30K to $40K per the board's framing. Waymo's is a costlier stack in dense cities. The Oct 1 detail is preserved on the Archived Autonomy Race board (Oct 1) ↗.
History and progress
October 2026 · reported targetFSD v15 on the robotaxi fleet, which Tesla AI lead Ashok Elluswamy has tied to 24/7 service. Not a Tesla App reports it as early as this month and describes v15 as roughly 10x bigger (about 1B to about 10B parameters). Not confirmed by Tesla. (Not a Tesla App, Oct 4)
Oct 30, 2026 · deadlineTesla's sworn answers to NHTSA's Cybercab Special Order (AQ26002) are due. Tesla asked Sep 18 for eight weeks, to Nov 20; NHTSA granted 30 days on Sep 22. Key question: how a car with no foot brake meets FMVSS 135, which calls for a foot control. Service not paused; no violation found. (NHTSA memo dated Sep 29, via TeslaNorth, Oct 8)
Oct 10, 2026Tesla's Robotaxi account: "300+ Cybercabs serving Austin, an 8x increase in 30 days," the first official Cybercab fleet number. Ashok Elluswamy: "Dallas soon," no date or fleet size. Dallas has driverless Model Y rides since April. No ride, utilization, or monitor numbers. (Robotaxi and Elluswamy on X; Teslarati, Oct 11; Electric-Vehicles.com, Oct 10)
Oct 9, 2026Tesla renamed FSD (Supervised) "Tesla Assisted Driving" in Europe after Germany's transport ministry called the old name misleading. Germany will back EU-wide approval with a maximum 10% over the speed limit; EU vote no earlier than December. Tesla's X account posted "Floodgates are opening" over four Cybercabs in Texas traffic, with no date or city. (TechCrunch; BMV; Tesla on X; TeslaNorth, Oct 9)
Oct 8, 2026Texas Cybercab registrations jumped by a record 150 in one batch, from 169 to 319. Model Y robotaxis 420; Texas robotaxi fleet 739, about 43% Cybercabs. Supply, not paid rides. (Texas DMV via Sawyer Merritt, Robotaxi Radar, FSD Database; CyberCab.News, Oct 9)
Oct 4 to 6, 2026Drone footage of Tesla's Houston robotaxi lot: an estimated 190 to 200 Cybercabs, about 15 charging, more inside the Tesla Hempstead Center, cabs moving in and out for testing. About 50 in mid-September, about 105 on Sep 29. Houston has driverless Model Y rides; Cybercab rides remain Austin only. Estimate, not a Tesla number. (S.E. Robinson Jr. on X; CyberCab.News, Oct 6; Drive Tesla Canada, Oct 7)
Oct 5, 2026A rider's Cybercab looped in an Austin pickup lot for about 7 minutes; Tesla support stopped it remotely and took about 7 more minutes to restart FSD. No crash. Rider called it an edge case. (Rider video on X; TeslaNorth, Oct 5)
Oct 5, 2026Newest public Tesla software build (2026.27.6.15) still carries FSD v14, so v15 has not reached owners. (Tessie stats, Oct 5)
Oct 3, 2026One Cybercab and nine Model Y robotaxis arrived empty at an Austin Supercharger and parked themselves; a Tesla employee plugged each in by hand. Wireless charging is patented, not deployed. (Observer on X; Teslarati, Oct 5)
Sep 30, 2026California SB 1246 signed: US-based, US-licensed remote drivers only; local incident technicians; alerts to cities in fleet-wide failures; fines if a robotaxi blocks first responders over 30 minutes. Effective July 2028. TechCrunch: Tesla is the only company that has said publicly its remote operators can take direct control. (TechCrunch, Oct 1; noted Oct 6)
Oct 2 to 3, 2026Robotaxi account: Austin service now until 11 PM (window 6 AM to 11 PM, up from 6 AM to 10 PM set Aug 26). Musk: the main thing left is not running over pets that are hard to see at night, "grey kittens on grey tarmac in the dark." Oct 3: "We are being extremely careful with autonomous safety." Still one hour shorter than the June 22, 2025 launch window (6 AM to midnight, with an employee aboard). (Tesla Robotaxi and Musk on X; TeslaNorth, Oct 3 to 4; Electric-Vehicles.com, Oct 3)
Oct 3, 2026CNBC (Lora Kolodny): Cybercab's first month in Austin drew mixed passenger reviews (long waits early, wrong pickups/drop-offs, butterfly-door and trunk issues). About 169 Cybercabs authorized in Texas (almost 4x from about 45 at unveil). Waymo contrast: about 1,154 AVs authorized in Texas as of Friday, more than 500,000 paid rides a week, about 270 million fully autonomous commercial miles, over 4,000 driverless vehicles in US ops. Tesla also has about 420 Model Y robotaxis in Texas (not yet retail unsupervised). No significant Cybercab collisions reported in Austin. California: still no unsupervised public-road robotaxi permits. Expansion talk: San Antonio, Dallas, Nevada, Florida. (CNBC, Oct 3)
Oct 3, 2026NHTSA spokesperson confirmed via CNBC that Tesla obtained an extension on AQ26002 Special Order past the Sep 30 deadline. New deadline not published. Do not treat as clearance. Secondary blogs claiming a full Tesla response already filed: unverified. (CNBC, Oct 3)
Oct 2, 2026Texas Cybercab registrations climbed to about 158, then about 169 the same day (board had about 126 as of Sep 28). Supply, not paid rides. (Texas DMV via Sawyer Merritt / wholemars; Basenor roundup, Oct 2)
Sep 30, 2026NHTSA's original deadline for sworn Tesla answers on Cybercab certification. Board had no public outcome as of Oct 1 to Oct 3; Oct 4 scan adds CNBC extension Context. (Autonomy board, Oct 1; CNBC, Oct 3)
Sep 30, 2026Austin Fire Department finished Tesla-led Cybercab emergency training. (FOX 7 via Autonomy board)
Sep 28, 2026Public reporting says the Austin Cybercab fleet roughly doubled, with Texas registrations of about 126, up from about 69 a few days earlier. This is supply, not usage. (Autonomy board log)
Sep 23, 2026Waymo began driverless freeway rides in Austin through Uber. Tesla's Austin service stays on surface streets. (CBS Austin via Autonomy board)
Sep 16, 2026NHTSA escalated its inquiry (AQ26002) into a formal Special Order on how the Cybercab, with no wheel, pedals, or mirrors, was certified to safety standards. (Autonomy board log)
Sep 9, 2026Tesla began limited commercial Cybercab rides in Austin. NHTSA opened AQ26002 the same day. Waymo opened Denver, San Diego, and Tampa. (Autonomy board log)
Current gates (the how)
Shut
Gate A · Unsupervised that sticks
Useful hours per day, in real weather and cities, without monitors creeping back. Month-one mixed reviews (waits, wrong drop-offs) are slight pressure. (CNBC, Oct 3) New scoreboard as of Oct 5: service hours past 11 PM, then midnight, then 24/7. Tesla has named night detection of low-contrast animals as the blocker. One extra hour is a small Helps. (Musk on X, Oct 2 to 3) Oct 6: a stuck Cybercab needing a remote stop and restart is a small Hurts; the new thing to watch is humans per car. (TeslaNorth, Oct 5)
Open question
Gate B · Cost edge converts to money
Cheap cars only matter if they earn. The Austin surface-street limit, Waymo's freeway rides, and Waymo's scale gap are pressure here. (CNBC, Oct 3) Labor per car matters too: robotaxis are still charged by hand. (Teslarati, Oct 5)
Live
Gate C · Hard-city and regulator parity
The NHTSA Special Order is the live test. Sep 30 clock extended per CNBC; not resolved. California still closed for unsupervised public-road robotaxi, and its new SB 1246 sets remote-driver and first-responder rules from July 2028. (TechCrunch, Oct 1)
Live
Gate D · Disclosure that survives sunlight
Can Tesla publish rides, fleet, and crash numbers that hold up to scrutiny? (Autonomy board, Oct 1; Gates E and F cover Waymo densifying and China as a control.)
What would change the outlook
Better: Paid stranger rides at meaningful volume in low-cost cars, highway or faster routes, a clean NHTSA result. Worse: Cars parked in lots, regulators treating the stack as permanent probation, one serious crash cluster.
Oct 7 · Quiet; Cybercab road share 2.2 per 1,000 (Context) (prior daily)
Oct 5 · Cybercab stuck in pickup lot, remote stop and restart (Hurts) (prior daily)
Sep 30 · California SB 1246 remote-driver and first-responder law (Context) (prior daily)
Oct 6 · Deep dive: the people behind the driverless car (prior daily)
Oct 2 to 3 · Austin robotaxi hours to 11 PM; Musk names the night-pet blocker (Helps) (prior daily)
Oct 5 · Deep dive: one more hour of night and the grey kitten test (prior daily)
Oct 3 · Cybercab first month mixed reviews (Hurts) (prior daily)
Oct 3 · NHTSA extension on AQ26002 (Context) (prior daily)
Oct 4 · Deep dive: Cybercab first month + NHTSA (prior daily)
Oct 2 · Texas registrations ~158 then ~169 (Context) (prior daily)
Oct 1 · NHTSA deadline passed; Oct 4 adds extension Context (prior daily; still live)
Sep 30 · Austin FD Cybercab training (prior daily)
Terafab
TerafabJV · Tesla + SpaceX + xAI
Status, Oct 11: New this scan: Musk said Oct 10 Terafab will put logic, memory, packaging, and masks in one building, start on standard vendor equipment, then build Tesla's own, and aim for about ten chip designs a year. Approach, not a deal or date. Gate D unchanged. Oct 10: Quiet day; follow-up coverage of Musk's Oct 7 "we will build and run the fab." Gate D unchanged. Oct 9: New this scan: ENR read the Texas JETI applications: up to $119B across four phases through 2036 (later phases optional), first phase 2026 to 2028, commercial operations expected in 2028. Tesla is hiring its own construction managers near the site. Moves the $119B figure from unverified press claim to public filings. Context, Gate D unchanged. Oct 8: Follow-up coverage only. Tom's Hardware says Intel is expected to license its 14A process to Terafab, but whether Intel helps ramp production is unclear. Quartz cites a Grimes County public notice of $55B to begin construction, versus the $16.8B first phase on this board; the notice has not been checked here. Indirect: SpaceX is reportedly seeking about $40B of debt to buy Nvidia GPUs (Oct 6 to 7), so the orchestra keeps buying outside chips until Terafab produces. Gate D unchanged. Oct 7: New this scan: Musk said Oct 7 that Tesla and SpaceX will build and run Terafab, "Let there be ZERO doubt," with TSMC at most subleasing part of it; Intel CEO Lip-Bu Tan told Bloomberg Intel stays involved. Rules out the TSMC-operated scenario Culpan reported Oct 2. Context on Gate D: control answered, execution harder. Oct 6: Oct 5 recaps (Motley Fool) only restate the TSMC talks and note Intel has been the only named chip partner. Earlier: Oct 4 recaps only restate the $16.8B first phase and TSMC talks. Tesla's new $30B credit lines (8-K, Sep 29) are dry powder at the orchestra level, not a Terafab commitment. Partner discussions with TSMC confirmed Oct 3 (Musk: "Just discussions, but something may come of it"). No deal, capacity, or timeline published. Tom's Hardware Oct 3 is secondary elaboration of Culpan's Plan A/B and Intel 14A risk, already covered by the Oct 3 deep dive. Two legal fights remain open: public records and the name. Gate D stays Elevated · shut. No new Helps today.
Why it matters
It is the chip factory that sits across the value boundary. It is meant to supply rad-hard D3 chips for Starmind and AI6-class chips for Optimus. Since Suncatcher, the board also treats it as the answer to buying AI chips at Nvidia's margins or from Google's TPU program.
History and progress
Oct 10, 2026Musk at the X Takeover event: logic, memory, packaging, and masks in one building at Giga Texas; start with standard vendor equipment, run it faster, then build Tesla's own; aim for about ten chip designs a year versus one major design every two years. No partner, capacity, or date. (NextBigFuture, Oct 10)
Sep 29, 2026Tesla signed $30B of new credit facilities, undrawn, no 2026 draw planned, general corporate purposes. Not tied to Terafab in the filing. Filed at the orchestra level; noted here because Terafab is the biggest capital ask on the map. (Tesla 8-K, Sep 29; noted Oct 5)
Oct 7, 2026Musk on X, replying to a post that TSMC would most likely own and run the fab: "No, we will build and run the fab. Let there be ZERO doubt about that." "Maybe TSMC subleases part of the Terafab if they want, but nothing more than that." Same day, Intel CEO Lip-Bu Tan told Bloomberg Intel will remain involved. TSMC and Intel did not comment to Reuters. No terms, capacity, or node published. (Reuters via CNA; The Next Web, Oct 7)
Oct 8, 2026ENR, reading the Texas JETI incentive applications: up to $119B across four manufacturing phases through 2036, later phases dependent on decisions to proceed; first phase 2026 to 2028 with commercial operations expected in 2028. Tesla is hiring construction project managers and superintendents in Anderson, near the site. (Engineering News-Record, Oct 8)
Oct 15, 2026 · scheduledTSMC Q3 results; Terafab talks a likely topic. (Data Center Index citing TrendForce, Oct 6)
Oct 2 to 3, 2026Tim Culpan reported late Oct 2 that TSMC is exploring ways to work with Terafab, most likely scenario he heard: TSMC owning and operating a new Texas factory with Terafab as an anchor. On Oct 3, Musk replied on X confirming discussions: "Just discussions, but something may come of it." No deal announced. Culpan also framed Intel collaboration as not making progress (his reporting). Korean press claims about groundbreaking / $119B / Korean equipment: unverified secondary. (Culpan / Culpium, Oct 2 to 3; Musk on X via Culpan)
Sep 24, 2026Board elevated Terafab after Suncatcher: it is now also the chip-cost answer, not only a two-mission fab. (Board, Sep 24)
Sep 16 to 17, 2026Tesla and SpaceX (plus SpaceXAI) sued TERA-print LLC in federal court in Austin, asking a judge to confirm they can use the "Terafab" name after a cease-and-desist. (Board log, Sep 17)
Sep 10, 2026SpaceX sued in Travis County to block release of Terafab tax-abatement records held by Grimes County. (Board log, Sep 10)
History before Sep 2026 is not on the existing boards. Backfill needed.
Current gates (the how)
Elevated · shut
Gate D · Dual mission without civil war
Can rad-hard and AI-throughput chips share one fab without one becoming the neglected child? Evidence would be separate capacity commitments and a clear rule for who loses when they collide. Partner talks with TSMC are a necessary early signal, not capacity. Slogans do not count. Oct 7: Musk says Tesla and SpaceX will run the fab themselves, so the priority call stays inside the orchestra; the open question is now execution by a first-time operator. (Board, Oct 1; Oct 3 scan; Oct 7 scan)
What would change the outlook
Better: A signed term sheet, ring-fenced capacity, and separate success measures. Worse: Years of "we'll do both" with no walls, talks that stall, or legal fights that expose the terms.
Oct 7 · Musk: "we will build and run the fab"; Intel stays (Context) (prior daily)
Oct 7 · Deep dive: who runs the fab (prior daily)
Oct 2 to 3 · Musk confirms TSMC and Terafab talks (Helps) (prior daily; no new Helps Oct 4)
Oct 3 · Deep dive: TSMC and Terafab (prior daily)
Tesla core business
Gate G · TSLAunder TSLA · Semi, vehicles, Roadster · a gate, not a goal
Status, Oct 11: Weekend, no new closes (TSLA last $382.70 Oct 9). Pilot and Tesla opened their first two truck-stop Semi Megacharger sites, Ellabell, Georgia (Oct 8, six stalls) and Tulare, California (Oct 9), with Fort Stockton, Texas, and North Las Vegas next. Still no Semi unit counts. Context. Oct 10: TSLA closed $382.70 Oct 9, up about 2.1%. Europe rename to "Tesla Assisted Driving" with German backing for EU approval is a possible software-revenue lever (vote December at the earliest). Context. Oct 9: Quiet day. TSLA closed $375.00 Oct 8, down about 0.7%. NHTSA Cybercab answers due Oct 30, nine days after the Q3 call. Oct 8: Decided Oct 4: Semi, vehicles, and the Roadster stay a Tesla gate on this board, not a fifth goal, until real news warrants its own page. Q3 deliveries beat (Oct 2), Semi volume production claimed at Sparks (Sep 24 to 25), Roadster reveal Oct 15. Oct 6: IMC Logistics is adding 50 Semis, with delivery around late Sep or early Oct (small Helps). Oct 7: Powershare Home Backup now on new Model 3 and Model Y with Powerwall 3 (Context). Oct 8: first East Coast Semi Megacharger opened in Conley, Georgia, eight stalls at up to 1.2 MW (Context).
Why it matters
The car, truck, and energy business is the cash and credibility engine that pays for Optimus, robotaxi, and Terafab. If it wobbles, every big bet gets more expensive. If it holds, the bets get time.
History and progress
Oct 21, 2026 · scheduledQ3 earnings call after the close; Q&A webcast 4:30 PM CT. (Tesla IR)
Oct 15, 2026 · scheduledRoadster reveal, moved from Oct 1 because forecast weather made the outdoor event hard to hold. (Benzinga, Oct 4; Oct 1 digest)
Oct 8 to 9, 2026Pilot and Tesla opened their first truck-stop Semi Megacharger sites: Pilot in Ellabell, Georgia (Oct 8, six stalls) and Flying J in Tulare, California (Oct 9). Fort Stockton, Texas, and North Las Vegas next; nine more under construction; four to eight stalls each at up to 1.2 MW. No Semi unit counts. (Pilot and Tesla via Business Wire, Oct 8; Electrek, Oct 9)
Oct 7 to 8, 2026Tesla opened its first East Coast Semi Megacharger in Conley, Georgia, outside Atlanta: eight stalls at up to 1.2 MW each, about 60% charge in roughly 30 minutes per Basenor. Earlier working sites were almost all in California and Nevada. No Semi unit counts. (Sawyer Merritt on X via Basenor, Oct 7 to 8)
Oct 6, 2026Powershare Home Backup turned on for new Model 3 and Model Y ordered Oct 1 or later, paired with Powerwall 3: over two days of backup for a 30 kWh-a-day home, more than three with a Cybertruck. Shipped with software 2026.38.3. Ford and GM have offered this for years. Feature, not a revenue number. (Electrek; TeslaNorth, Oct 6)
Late Sep to early Oct 2026IMC Logistics (drayage, more than 2,700 trucks, 50+ US locations) is adding 50 Tesla Semis. Announced in September; its marketing VP said Sep 29 that delivery would come that week or the next. One of the first customer handovers since Sparks opened. No Tesla unit counts. (Teslarati, Oct 5)
Oct 4, 2026Ryan decided Semi, vehicles, and Roadster stay a TSLA gate (Gate G), not a fifth goal. Open Discovery sweep ask closed Oct 5.
Oct 2, 2026 · baseline for the daily scanTesla's Q3 report: 486,532 vehicles delivered (beat the roughly 462,000 expected by about 25,000), 464,391 produced, and 13.7 GWh of energy storage (below the roughly 15.9 GWh expected, second-best quarter ever). Deliveries are down about 2.1% from Q3 2025 and up about 1.3% from Q2. Other models (Cybertruck, Semi, S and X) were 8,295, down 48%. (Tesla IR / 8-K, Electrek, Oct 2)
Oct 1, 2026Reuters: Tesla's September registrations rose across Europe, extending a recovery. (Reuters, Oct 1; headline only, figures not checked by this board)
Sep 24 to 25, 2026Sparks factory inauguration; coverage says high-volume Semi production started and deliveries "start now." No unit counts published. (Musk Orchestra board, Sep 24 to 26)
Aug 18, 2026Einride agreed to add 500 Tesla Semis, the largest Semi deal to date. Einride's CEO later said about 75 run in 2026 and the rest by end of 2027 (date of that remark not confirmed by this board). (TechCrunch, Aug 18; EV Wire)
Current gates (the how)
Holding
Gate G · Core business funds the bets
Do deliveries, Semi, and energy keep throwing off enough cash and credibility to fund Optimus, robotaxi, and Terafab without strain? Q3 deliveries beat; storage missed; low-volume models fell 48%. IMC is adding 50 Semis (Oct 6 scan). Car-to-home backup now ties Model 3 and Y to Powerwall 3 (Oct 7 scan, Context). First East Coast Megacharger opens (Oct 8 scan, Context). Next tests: Roadster reveal Oct 15 and Q3 margins Oct 21. (Oct 2 to 8)
What would promote this to its own goal
Real news that changes the thesis on its own: Semi unit counts at scale, a Roadster that moves demand, or a core-business miss big enough to threaten funding for the other programs.