24h digest (to Wed 21 Oct 2026 ~04:00 Zurich)

Compiled Wednesday 21 October 2026 ~04:00 Europe/Zurich (CEST, UTC+2). Daily 4am open-web issue.

Window: Tue 20 Oct ~04:00 Zurich → Wed 21 Oct ~04:00 Zurich.
Scope: science/engineering and business. Politics/culture-war excluded.
Sources: open public web only (SpaceX Q3 release on SEC EDGAR, StockAnalysis/Quartr call transcript, CNBC, NBC/Reuters, TechCrunch, The Verge, Bloomberg Law, BBC, Yahoo Finance, Motley Fool, Investing.com, TipRanks, Spaceflight Now, Space.com, Electrek, Not a Tesla App, FinancialContent, Elon Musk Archive mirror of public X posts). No X API / X MCP.

SpaceX blows past Q3, guides Q1 2027 to $33–37B, and Musk confirms Tesla and SpaceX merger committees

SpaceX’s second print as a public company was a big beat. Revenue was $14.58B against about $12.7B expected (LSEG), up 161% year on year. Net income was $4.4B and adjusted EBITDA $8.8B. The AI segment, now called Super Intelligence (SI), brought in $7.56B, which makes it the largest segment for the first time and a profitable one ($431M operating income). Capex came in at $19.59B, right on consensus. Then came the guidance: CFO Bret Johnsen said Q4 capex will be $35–40B and Q1 2027 revenue $33–37B, against about $22.2B expected (LSEG). On the call Musk said both boards have formed special committees of independent directors to look at a Tesla–SpaceX merger. He called it “a good idea” but said “not a certainty”, and pointed out that Tesla would need 50% of all shares to approve it. SPCX closed −0.11% at $147.79 and traded up about 8% after hours (CNBC, BBC). TSLA closed +1.54% at $373.71 and rose about 2% after hours, ahead of its own results tonight. Flight 15 will attempt to catch both the booster and the Ship. Musk said it is expected “next month”, and Shotwell said the FAA licence is expected to cover both tower catches.

SpaceX

1. Q3 2026 actuals: $14.58B revenue, $4.4B net income, SI the biggest segment — [24h / earnings]

Who / when: SpaceX earnings release (8-K / EDGAR) just after 3 p.m. CT (~22:00 Zurich Tue 20). Webcast from the Bastrop Starlink factory at 22:30 Zurich.
Substance:
Totals: revenue $14,575M (Q2 $7,814M; Q3 2025 $5,576M). Operating income $2,791M. Net income $4,404M, or $0.35 diluted GAAP EPS; IBES/LSEG adjusted EPS was $0.45 against $0.13 expected. Adjusted EBITDA $8,809M, up 306%.
Space: revenue $1,767M, op. loss −$7M, 35 launches (12 customer).
Connectivity: revenue $5,248M against $4.84B StreetAccount. Op. income $2,367M. 13.6M consumer subscribers (+1.6M in the quarter), ARPU $69 (up from $66), more than 3,800 aircraft. Gross margin 55%, including about a 3-point one-time tariff-refund benefit.
SI: revenue $7,560M against $6.47B StreetAccount. Op. income $431M, after $1.2B of Cursor-related stock compensation. SI adjusted EBITDA $4,953M. Nameplate compute 1.7 GW (was 1.4 GW). Cursor contributed $608M from its 14 Aug close.
Capex: $19,593M, of which SI was $16,444M.
Balance sheet and cash: cash $51.3B plus $38.6B of marketable securities. Debt and finance leases $38.8B (current plus long-term). Nine-month operating cash flow was $13.1B against $48.1B capex. Seeking Alpha’s Agar Capital puts nine-month FCF at about −$35B; SpaceX does not report a Q3-only FCF figure.
Backlog: $63B, of which about $42B is due within 12 months (Johnsen).
Why it matters: The quarter answered the “can AI pay for itself” question for now. SI made money at the segment level and showed a 65.5% adjusted EBITDA margin (Wedbush). Most of that SI revenue is still compute rental to Anthropic, Google and Reflection AI, which is a cloud-style business rather than revenue from SpaceX’s own models (Reuters).
Source: SpaceX Q3 release (SEC) · CNBC live · Reuters via NBC · Seeking Alpha

2. Guidance: Q4 capex $35–40B, Q1 2027 revenue $33–37B, $100B ARR at year-end “on track” — [24h / guidance, capital]

Who / when: CFO Bret Johnsen on the call, Tue 20 ~23:00 Zurich.
Substance:
Q4 capex “now… $35 billion to $40 billion”. That is roughly double Q3, driven by extra capacity coming online this quarter.
Q1 2027 revenue guided to $33–37B, citing “increased visibility into our compute contract pipeline over the next six months”. LSEG consensus was $22.2B (CNBC).
ARR: still on track for the stated $100B ARR by year-end, measured on December.
Compute: nameplate close to 2 GW by the end of October and 2.5–3 GW by year-end. Compute pricing is “slightly above the high end” of $30–50 per watt, with payback “less than a year”.
Margins: Q4 gross margin to look more like Q2.
Funding: he said SpaceX is investment grade and talking to “a number of available pools of debt capital”, and also to vendor and customer financing. He gave no specific raise.
Why it matters: The financing overhang Morgan Stanley flagged is now sharper. A $35–40B quarterly capex rate is a bigger draw than any 2026 consensus model assumed, but it comes with a revenue guide about 50% above consensus. How it gets funded (the Apollo-led debt package and others) is now the main open question.
Source: Call transcript (StockAnalysis/Quartr) · CNBC live

3. Musk: Tesla and SpaceX have formed special committees on a merger, “a good idea” but “not a certainty” — [24h / capital markets, governance]

Who / when: Musk, answering Deutsche Bank’s Edison Yu on the call, ~23:30 Zurich Tue 20 (Verge timestamp 21:30 UTC). He reposted a summary at 23:57 Zurich.
Substance:
Committees: both boards have formed special committees of independent directors “at the request of shareholders”. Musk is excluded as a “conflicted party” and called it “an odd kabuki dance”. The committees “have not reached agreement on the terms of any transaction… no guarantee that they will”.
Vote threshold: on the Tesla side a deal needs “a 50% vote of all shares… quite a high hurdle”.
Rationale: Musk cited cooperation on superintelligence software, chip design, Terafab, robotics, Starmind and connectivity.
Who could sit on them: TechCrunch lists the possible pools. Tesla has seven independent directors (Denholm, Ehrenpreis, Straubel, Gebbia, Hartung, Murdoch, Wilson-Thompson). SpaceX has six (Ehrenpreis, Glein, Harrison, Jurvetson, Nosek, Boutha).
Why it matters: Until yesterday the merger was investor speculation. It is now a formal board process at both companies. Tesla’s 3 Dec ballot has no merger item, and Tesla had issued no statement of its own at compile time, so any deal would need a separate vote.
Source: TechCrunch · The Verge · NBC News · Bloomberg Law · Musk repost (archive)

4. Musk’s compute and ARR claims: “less than 13 GW” by end-2027 would surprise him; $500B–$1T ARR by end-2027 — [24h / statements, SI]

Who / when: Musk, opening remarks and Q&A, Tue 20 22:30–23:45 Zurich. On X at 00:05 Zurich Wed: “As crazy as this sounds, $1T in SpaceX revenue in 2028 is no longer impossible.”
Substance:
Deployment pace: about 200,000–220,000 GB300s brought online per month for the rest of 2026, with 2027 “primarily Vera Rubin”.
2027 compute: “highly confident” of +10 GW in 2027, which would mean about 13 GW in total by the end of 2027. He described more than 20 GW as a “small possibility” and said SpaceX is aiming for 20 GW of incremental power and cooling.
Training cluster: a 440,000 Vera Rubin cluster targeted for the end of Q2 2027.
Turbines: SpaceX is casting its own turbine blades and vanes, which Musk says removes a 12–18-month supplier lag.
Compute allocation: roughly 50/50 internal vs rented, on short contracts at about spot rates, which he says are renewing at higher $/W.
ARR: SpaceX ARR of “$500 billion and $1 trillion” by end-2027. Musk joked that he didn’t “want to give Bret a heart attack”.
Why it matters: These are CEO targets, not guidance. The CFO’s guidance (Q1 2027 $33–37B) is the number to model. They do show the capex trajectory: 10 GW at roughly $30–50 per watt implies hundreds of billions of 2027 spending if it is built. That is an inference from the call figures, not a company number.
Source: Call transcript · Musk $1T post (archive) · CNBC live

5. Starship: Flight 15 “next month” with a dual catch; weekly flights in H1 2027 — [24h / Starship]

Who / when: Musk and Gwynne Shotwell on the call, Tue 20 ~22:35 Zurich.
Substance:
Flight 15: “we’re going to attempt to catch both the booster and the ship”. Musk put it at “more likely than not” to succeed and gave no date beyond “next month”. Shotwell said SpaceX expects the FAA licence for Flight 15 to include return to launch site and tower catch for both stages.
2027 cadence: Starship weekly at some point in H1 2027, then twice weekly or more in H2.
Reuse and refuelling: re-fly a Ship by early 2027, and attempt on-orbit refuelling early next year.
Mobile V2 and Starmind satellites start flying in 2027.
Starbase Louisiana: Musk’s internal goal is a first launch in 2028; Shotwell said 2029, and Musk called 2029 “highly likely”.
New contracts: three more NASA crew missions (through 2028), 18 NSSL missions, a Starship lunar-surface payload, and three Starfall missions on Starship.
Why it matters: This is the first company-voiced commitment to a Ship catch on Flight 15, which until now was in press reports only. It is consistent with CADENA’s provisional NET 8 Nov. There was still no public B22 static-fire confirmation at compile (NSF on 18 Oct had B22 awaiting transport to Pad 2).
Source: Call transcript · Space.com · NSF

6. Starlink: 13.6M consumer subscribers, 22M Starlink Mobile monthly users, ~12,000 planes contracted — [24h / Connectivity]

Who / when: Shotwell and Johnsen on the call. The release came at ~22:00 Zurich.
Substance:
Markets and subscribers: 174 markets (Vietnam and Uganda new). +1.6M consumer subscribers to 13.6M, ahead of the 12.19M analysts expected (Reuters). Enterprise and government revenue was $2.13B, up 91%.
Starlink Mobile: more than 22M monthly users, up 37% on the quarter.
Aviation: about 12,000 aircraft contracted, roughly a quarter of available aircraft. Aviation production has quadrupled this year and should double again by year-end.
Kits and V3: V5 kit production is ramping. A “few hundred” V3 satellites are needed before users feel the difference, which Musk expects by mid-2027.
Constellation: more than 11,100 satellites in orbit (SFN, from the filing; the count now includes de-orbit-approved satellites).
Musk’s ambitions: about 100,000 satellites and 100× today’s bandwidth within roughly three years, and “a clear path to several hundred billion dollars a year in connectivity revenue”. He promised a dedicated presentation on Starlink Mobile strategy, and posted after the call that cellular-industry veterans “don’t understand that SpaceX is developing entirely new pieces”.
Why it matters: ARPU rose for the first time in a while, helped by mix. Mobile usage is growing fast ahead of the 2027 V2 constellation and the early-2028 service start that Shotwell restated.
Source: SpaceX Q3 release (SEC) · Call transcript · Spaceflight Now

7. Terafab: first test chip around May–June 2027, volume production “around Q3 2028” — [24h / chips]

Who / when: Musk, answering Morgan Stanley’s Adam Jonas on the call. On X at 21:02 Zurich: “The value of Terafab’s output will place it in the top 20 of country GDP in a single building” (archive).
Substance: The first functional test chip (an SRAM-type line test) is targeted for about May–June 2027 at the Austin research fab. Volume production at Terafab Gibbons Creek is aimed at around Q3 2028, with “big error bars”; Q3 or Q4 2028 is “at least in the set of possible outcomes”. Musk thanked ASML and other suppliers for speeding up equipment deliveries. He did not directly confirm the EUV order Jonas asked about.
Why it matters: This is the first dated chip-production timeline, and Terafab is one of the shared Tesla–SpaceX assets behind the merger case. Treat it as aspirational.
Source: Call transcript · CNBC live

8. SPCX −0.11% to $147.79, then up ~8% after hours; analysts react — [24h / markets, analysts]

Who / when: Regular-session close 22:00 Zurich. After-hours moves were reported through 02:42 Zurich (CNBC update).
Substance:
Price: closed $147.79 (−0.11%) on 89.4M shares (Motley Fool). After hours it was first up about 2–5% and then up about 8% (CNBC, BBC). Pre-print options were pricing about an 8% move.
Wedbush (Ives): Outperform, $225 reiterated. He called it “a blowout third quarter… beating on every segment line”, and said “we would rather see this compute financed and deployed than deferred”.
Morgan Stanley (Jonas): Overweight, $300 held (pre-print note; financing flagged as the key risk).
Futurum (Daniel Newman): a Tesla merger “feel[s] imminent”.
Bear view: Agar Capital on Seeking Alpha rates it Strong Sell, central value about $114, at roughly 30× 2027 EBITDA.
Lock-up: up to 1.3B shares unlock Thu 22 Oct, about 28% of the main lock-up pool (Yahoo/Stocktwits). Another 328M follow on 24 Oct.
Why it matters: The beat and the guidance arrive two days before the largest lock-up release since the IPO. Thursday’s tape will show whether demand absorbs the supply.
Source: Motley Fool · CNBC live · BBC · TipRanks (Ives) · Investing.com (Wedbush) · Yahoo Finance

9. Bandwagon-5 flies with the second Starfall capsule; Starlink 15-15 scrubbed — [24h / launch]

Who / when: Bandwagon-5 lifted off at 7:04 a.m. ET (13:04 Zurich Tue 20) from SLC-40. B1085 landed on A Shortfall of Gravitas; SpaceX’s landing post was reposted by Musk at 18:57 Zurich. Starlink 15-15 from Vandenberg was scrubbed for an apparent technical issue (SFN, 01:06 Zurich). It is now set for Wed 21, 2:59 p.m. PDT (23:59 Zurich) on B1063’s 36th flight, landing on ASOG.
Substance: Bandwagon-5 carried 32 payloads. SpaceX’s own Starfall reentry capsule deployed last, at T+2:48:28. Shotwell on the call: “We just did a Starfall mission today.”
Why it matters: This is the second flight of SpaceX’s in-house microgravity-return product, which now has three commercial missions booked on Starship.
Source: SpaceNews · Space.com · SpaceX · Spaceflight Now (15-15)

Held / ahead: - Legal, Colossus: a federal court in Mississippi denied the DOJ’s bid to dismiss the NAACP’s Clean Air Act suit over the Colossus gas turbines (CNBC live, Tue). - Starshield / Ukraine: a WSJ report says Musk spoke to administration officials before imposing the speed cap. This was not independently verified, and the policy angle is excluded. - Dates: lock-ups Thu 22 and 24 Oct. FAA PEIS comments close 30 Nov. Flight 15 provisional NET 8 Nov.

Tesla

10. Q3 tonight: consensus ~$28.1B revenue, ~$0.45 EPS; merger committees now on the table — [preview / earnings]

Who / when: Electrek (Fred Lambert), Tue 20 22:13 Zurich. Results come after the close Wed 21, with the call at 23:30 Zurich.
Substance: Consensus is $28.06B revenue (Estimize $28.36B) and $0.45 non-GAAP EPS (Estimize $0.48), against $28.1B and $0.50 a year earlier. LSEG’s figures were $0.43 and $27.88B. Q3 deliveries were 486,532, production 464,391 and storage 13.7 GWh. Watch items are auto gross margin excluding credits (16.3% in Q2), operating expenses (+47% in Q2), FCF (−$1.09B in Q2; Reuters consensus about −$4.9B), and the 2026 capex guide of more than $25B. The Say Q&A closed at 22:00 Zurich Tue with 832 questions from about 2,860 holders representing 3.85M shares. The top questions are Cybercab beyond Austin, unsupervised FSD, Cybercab for purchase, robotaxi per-mile economics, a HW3 upgrade, and Optimus cost. New since yesterday: analysts will now certainly ask about the special committee Musk disclosed on the SpaceX call.
Why it matters: SpaceX’s print raised the bar for the Musk complex as a whole. Tesla goes into its print with the merger process confirmed and a core business that is roughly flat.
Source: Electrek · Reuters via Investing.com

11. TSLA +1.54% to $373.71, then about +2% after hours on the committee news — [24h / markets]

Who / when: Tue close 22:00 Zurich. The after-hours move was reported by the BBC.
Substance: Closed $373.71 (+$5.65, +1.54%) (FinancialContent), and was up about 2% after hours (BBC). CNBC says TSLA is down about 17% year to date, a market value of about $1.5T.
Source: FinancialContent · BBC

12. FSD v14.3.13 in 2026.38.5.10: Automatic Collision Evasion and Emergency Drive Away — [24h / FSD software]

Who / when: Tesla release notes for 2026.38.5.10, first seen Tue 20 (Not a Tesla App; early rollout, 1 tracked car).
Substance:
FSD v14.3.13 (HW4): carries over the 14.3.12 work (upgraded RL stage, vision encoder, MLIR compiler with about 20% faster reaction time, one model shared by ASS, FSD and Robotaxi).
“Upcoming”: reasoning extended to all behaviours, and pothole avoidance.
Automatic Collision Evasion (HW4): new. It engages FSD (Supervised) when the driver seems inattentive or FSD was unintentionally disengaged.
Other features: Emergency Drive Away (drive off while still plugged in, emergency only), Set Arrival Energy from the phone, and Supercharger site maps.
Grok: Tesla removed the “beta” label from Grok in the car.
Why it matters: Collision Evasion is the first FSD feature that steps in when the driver is not supervising. Expect regulators to look at it given the open NHTSA probes.
Source: Not a Tesla App

Held / ahead: earnings Wed 21 23:30 Zurich. AQ26002 answers due 30 Oct. Annual meeting Thu 3 Dec (no merger item). EU TCMV is the earliest possible FSD vote, in December.

xAI (SpaceXSI)

13. Grok Bot: 800,000 paid weekly users, enterprise wide release “within a few weeks”, big app update — [24h / product, metrics]

Who / when: Shotwell and Musk on the call (Tue 20 ~22:40 Zurich). Musk on X at 18:56 and 18:57 Zurich (“Grok @Bot upgrades”, “Big Grok @Bot update”).
Substance: Grok Bot had 800,000 paid weekly users “earlier this month”, more than double a month earlier (Shotwell). B2B is the fastest-growing category, and Grok enterprise customers above $1M run-rate, including Cursor, rose 50% on the quarter on a pro-forma basis. Musk called it “the fastest growth of a product that I’ve ever seen”. He said usage per user rises over time, Grok Bot “may end up being what uses most of our SI compute”, and wide enterprise deployment comes “within a few weeks”. Version 0.71.2 accepts files up to 250 MB and video up to 1 GB, adds X DMs, connectors, voice calls, email aliases and an Android Assistant (X Freeze / Matt Palmer threads reposted by Musk). Similarweb said Grok Bot DAU hit a new peak on 17 Oct.
Why it matters: This is the first company-disclosed Grok Bot paid-user figure, and it ties a consumer product to the SI revenue line, which today is mostly compute rental.
Source: Call transcript · Musk “upgrades” (archive) · Musk “Big update” (archive) · Similarweb via repost (archive)

X (engineering)

14. New SI ranking model and retrieval model on the timeline; Android app rewrite; nationality filter “Ok” — [24h / engineering]

Who / when: Musk on X, Tue 20 04:41 Zurich and 06:11 Zurich. Reply to a Swiss user at 16:58 Zurich.
Substance: Two recommendation changes: a “New AI/SI ranking model that predicts probable time spent much better” and a “Retrieval model trained to surface posts people read all the way through”. Musk claims user-active seconds are “hitting all-time highs almost every day”, with no figure given. “Rewriting the 𝕏 Android app was also a big deal.” Asked to improve the nationality filter because Swiss feeds get “overloaded with German content”, Musk answered “Ok”, with no timing.
Why it matters: These are ranking changes aimed at engagement quality, and the Swiss-feed point is relevant for you locally. No metrics were published.
Source: Musk ranking (archive) · Android rewrite (archive) · Nationality filter (archive)

Quiet / no material 24h move

Neuralink: no news. The 50,000-hour pretraining update (1 Oct) still stands.
Boring Company: no project move. Nashville MB3 is still due late October.
Excluded: Texas Senate and early-voting posts, the immigration and “Great Replacement” reposts, and the Isaacson interview remarks on wealth and taxes. The “suicidal empathy” repost was also excluded.

Dashboard moves this digest

SpaceX:
The Q3 row now holds actuals, replacing consensus.
New rows: Q4 capex $35–40B and Q1 2027 revenue guide $33–37B.
Merger special committees (both boards) added.
Flight 15: a dual catch is now company-stated (“next month”; the FAA licence is expected to cover it).
2027 targets: weekly Starship in H1 2027 and on-orbit refuelling early 2027.
Starlink: 13.6M subscribers, ARPU $69, 22M Mobile MAU.
Compute: 1.7 GW, then ~2 GW at end-October and 2.5–3 GW at year-end. Backlog $63B.
Terafab: test chip May–June 2027, volume around Q3 2028.
Launches: Bandwagon-5 SUCCESS; 15-15 moved to Wed 21 23:59 Zurich.
Market: Tue close $147.79, +~8% after hours.
Tesla: merger committee row added. Q3 consensus updated ($28.06B / $0.45). FSD v14.3.13 / 2026.38.5.10. Tue close $373.71.
xAI: 800k paid weekly Grok Bot users; SI segment profitable in Q3; 440k Vera Rubin cluster by end of Q2 2027.
X: timeline ranking and retrieval models; Android rewrite.
Neuralink / Boring: quiet.
FR: the chapter-5 “Starlink” jalon moved from 12M (30 Jun) to 13.6M (30 Sep) and was patched. Nothing else in FR was touched.