Launch
US$4.1B
Proven cadence · feeds orbital economics
Strengthened · as of Signal № 013
One owner from power through compute to application — and whether that chain compounds faster than the market can replicate. H15 is a cross-vendor thesis; we map it through the integrator furthest along the stack. Today that is SpaceX. This page now reads in chapters: the foundation map from Signal 006 is left as it was drawn. Chapter 2 records what closed.
Vertical integration of power, data centres and frontier compute becomes structurally consequential by end of 2027.
Added mid-cycle after the S-1 disclosure. H15 is cross-cutting by design — it touches H1, H3, H5, H6, and H14 simultaneously. Weekly status reads live in the Signal; this page holds the longer map we iterate in chapters, so traction is visible against the original drawing.
Chapter 1 · Signal № 006 · left unchanged
H15 applies to any owner trying to own power → compute → application. Helicopter view: power generation is the industry bind; compute access is the valve that feeds frontier capability. The map below is SpaceX-specific — the furthest-along live case, not the only possible outcome. Read top-down: filing-disclosed base first, roadshow fiction at the horizon.
Foundation map
What the S-1 discloses → what the deck prices for 2030
S-1 · 2025
US$18.7B consolidated revenue
US$4.1B
Proven cadence · feeds orbital economics
US$11.4B
Profitable growth engine · real margins
Building now
Infrastructure power co-located with compute
Dedicated generation installed with Colossus build-outs
Terrestrial compute · leased today · owned build-out ramping
AI segment
The bet the IPO is really selling
US$3.2B · 2025
Models · Grok · stack AI revenue today
US$322B · by 2030
Roadshow projection · not our forecast
US$60B · binding deal
Enterprise AI coding · the accelerant the organic run-rate could not bridge on the IPO clock
Roadshow sequence
Next on the deck plausible · not asserted
Colossus 3+ feed · shift from landlord leases to owned xAI + Cursor training
Tesla combination · market speculation · Megapack · grid-scale generation
2030 horizon
Deck fiction economics contested
~3× terrestrial cost · Barron's framing · insurance talks are pathfinding, not proof
Solid = filing or under construction. Green = binding at last update. Dashed = roadshow projection or speculation. The hundred-fold xAI path only works if each projected layer converts.
Likely sequence: Cursor close (Q3) → Colossus feed turns captive → Tesla combination if the listing window holds.
Watch: Cursor close, segment reporting, Colossus feed rate, and whether fed compute compounds frontier capability before the 2030 clock runs out.
Chapter 2 · Signal № 013 · what converted
Chapter 1 left Cursor as a binding deal still waiting to close. That watch landed. Read this map against the drawing above: solid engines that already pay, the $60 billion layer that converted, product now shipping, and the horizon that is still a projection. Diversification is the assurance. The AI is still the ambitious part.
Traction map
What already pays → what closed → what is still a claim
Paying now
Three engines, not one the diversified business
Rockets at cadence · the original cash engine from the S-1
Connectivity that already makes money · the profitable growth engine
Colossus leased to labs that need chips · a paying compute line today
Converted
The application layer is owned
US$60B · completed
Chapter 1 called this a binding deal. This week it closed. The coding tool is inside the stack, not sitting on the option list.
Shipping now
Product on the clock not the same as revenue
Shipped the same window as the close. The bot is being assigned work. Benchmarks vs a frontier rival at a lower price are claimed, not verified here.
SpaceX and xAI operating under one name. Nvidia holds a $21B SpaceX stake through xAI — chip vendor as shareholder in the buyer.
Still a projection
The horizon has not printed ambitious, dated, not proven
By September · Musk projection · the dated claim to mark. Not a filing.
US$322B · by 2030
Unchanged from Chapter 1 · not our forecast
Anthropic is the most aspirational revenue story of the last eighteen months. Earlier IPO talk sat near $1 trillion. This week’s papers sit at $2 trillion and $2.8 trillion — talk, not a struck price. Both can take off. A lab asking the market to pay for a horizon; SpaceX asking the market to pay for a horizon that sits on rockets, a network, and rented chips. The difference is not who wins. It is what you still own if the horizon takes longer than the slide.
Green = converted or shipping in this chapter. Solid = paying now. Dashed = still a claim. Cursor close strengthens H15. The thesis does not resolve until application-layer revenue prints in segment reporting.
Landed: Cursor close (the Chapter 1 watch).
Watch next: Cursor usage as SpaceX AI revenue; Grok Bot out of waitlist; whether Musk’s September claim shows up as a line, not another model launch.
H15 only resolves if several other theses hold or break in compatible ways:
The stack updates when binding moves land — not every week. Chapter 1 is Signal № 006 (Cursor as a binding deal). Chapter 2 is Signal № 013 (the close). Earlier context: Signal 004 (Goldman roadshow sketch), Signal 005 (IPO lock-up and Cursor option), Signal 012 (SpaceX as the public-market floor).