Archived Oct 2, 2026. No longer updated. See the current board.

The Musk Orchestra

Value-routing map across SpaceX ↔ Tesla (and the JV seam). Shared anchor up top; everything below is Grok’s first-principles cook — Fact / Judgment / Unpublished labeled. Autonomy is a separate live board.

Grok · As of Oct 2, 2026

Since yesterday

Oct 2, 2026

Transporter-18 launched Oct 1 (~11:32 a.m. PT, Vandenberg); Google Project Suncatcher M1 reached orbit with confirmed contact and operating as expected (SpaceX / Google blog) — moves launch-day → post-deploy / in-orbit hardware-test framing on the competitor-validation thesis; still one-server-class MVP (rad/thermal data weeks ahead; AI compute in orbit not demonstrated).
Musk: AI5 RAM cut to 72GB LP5 / AI6 to 144GB LP6 to secure Optimus production volume and cut cost (X, Oct 1) — memory supply as Optimus-rate / SILICON constraint; bandwidth held constant; do not invent wafer or unit counts.

Exec brief

Skim here · open tabs for depth

Lift· Silicon· Labor· Focus Focus = scarce founder bandwidth

Thesis: The orchestra is one scarce-constraint system — LIFT · SILICON · LABOR · FOCUS — routing value across two balance sheets. Price the binding constraint, not the slide deck. Starship cadence and Terafab dual-mission are the common throats; Optimus only matters at internal rate; Starmind is lift and thermal/power, not lift alone. Focus (founder calendar) is a first-class input, not a soft culture footnote.

What matters now: Google Suncatcher as orbital-compute competitor-validation (not disruption) — MVP now in orbit with contact confirmed (Transporter-18 / Planet Labs + Google TPUs, Oct 1) — still a one-server-class hardware test (rad/thermal weeks ahead; AI compute in orbit not demonstrated); elevate Terafab as the chip-cost answer to Nvidia margin / Google TPU; Starship as cheap iteration (F14 first orbit + Starlink V3 done — Gate A now cadence/reuse); Semi Sparks volume claim (Sep 24–25) as Gate G (sustained rate/fleet hours, not ribbon-cut); Optimus unsupervised hours (AI5/AI6 RAM cut framed as volume/cost necessity); calendar collisions across Starship / Semi / Optimus / Terafab / xAI.
Watch this month: Suncatcher post-deploy telemetry / rad/HBM (~2 krad HBM ground issues) and weeks of in-orbit TPU stress data (orbit + contact logged Oct 1); whether Google books Starship (bull) vs Blue/ULA/RL (bear); terrestrial Google↔SpaceX ~$920M/mo renew vs cancel (~90d); 2027 formation flying as the real ODC test; Tesla Semi Sparks volume-production claim (Sep 24–25; deliveries “start now”) as Gate G / production-fleet signal; Starship post-F14 cadence (first orbit + 26 Starlink V3 done Sep 28; next belief flip = boring reuse / catch / iteration toward F15+) as Gate A / lift; Roadster Waco Oct 1 + Q3 P&D ~Oct 2 (calendar); Starlink 15-25 Sep 30 / Crew-13 NET Oct 1; JV language that ring-fences D3 vs AI6; Terafab public-records/trademark path; SPCX float/lock-up (~Dec 9) vs TSLA liquidity; Optimus boring factory utilization — not demos (AI5/AI6 memory-supply watch).
What falsifies: Suncatcher scales past one-server demo into formation/clusters without needing Starship economics or SpaceX lasers ⇒ competitor, not customer+shareholder validation. Starship reliable + Starmind still stuck ⇒ thermal/power was the real bet. Optimus demos without internal shift replacement ⇒ labor optionality is narrative. Shared Terafab slogans with no capacity walls ⇒ future tug-of-war. Semi volume claim without follow-through rate/fleet hours ⇒ Semi stays narrative.
Do not invent: Flight counts, wafer starts, or TAM — prefer unpublished governance facts and belief thresholds.
Liquidity asymmetry: Both TSLA and SPCX are listed; the asymmetry is liquidity, float, and disclosure cadence between two public multiples — not private-space optionality by proxy. Day-90 / affiliate lock-up eligibility is a float watch (not forced selling). Watch capital actions, related-party flows, and Terafab disclosure boundaries that clarify whose book absorbs which story.
Key-person / Focus: Bus factor ≈ 1 is the org chart. FOCUS is scarce founder bandwidth — when two “existential” programs collide, improvisation is the binding constraint.

The orchestra

Skim: Map only — SPCX lift/compute, TSLA labor/product, Terafab at the seam. Autonomy = separate live board. Later tabs are mechanisms and belief flips.

Shared starting map — the value-routing idea. Space side and Earth/product side; programs that move capital, chips, labor, and lift across the boundary. Constraint spine: LIFT · SILICON · LABOR · FOCUS. This box is the grounding frame; later tabs are Grok’s build-out.

Fact program names & structure Judgment Grok thesis Unpublished governance / ops not public

SPCX Space side

Starship — lifts everything (mass to orbit is the root capability). LIFT
Starmind — orbital data centers; compute that wants vacuum, power, and cooling in space.
Vehicles / Energy intentionally off this page.

Terafab JV seam Tesla + SpaceX + xAI straddling the boundary.
Rad-hard D3 and AI6-class split. SILICON

Autonomy / Robotaxi Adjacent product — not expanded here.
Autonomy Race · live board

TSLA Earth / product side

Optimus — humanoid at rate (labor + factory throughput story). LABOR
Semi — Sparks factory → volume production + customer fleet hours (Gate G). Not a vehicles footnote.
Autonomy / Robotaxi — adjacent; treated as demand for compute/fleet data, not duplicated. → Autonomy Race
Other Core Vehicles & Energy kept off-page on purpose.

Frame · not a score Liquidity & float differ wildly across the orchestra. Both names are public listed multiples — category error to treat market-cap snapshots as a head-to-head score when float and disclosure rhythm diverge. Prefer capital actions and disclosure boundaries over reconstructed ops ledgers.

How to read this board: The orchestra is the map. Tabs below invent Dependencies, Gates, Risk, Log, and Barriers from first principles — what has to be true for value to route. Prefer unpublished judgment over fake ops numbers. Labels: Fact Judgment Unpublished · Autonomy is separate · autonomy-race.