Project Kepler: The Case for Combining Tesla & SpaceXProject KeplerThe Case for Combining Tesla & SpaceX

A merger earns its keep in two ways. It moves faster, and it sells more. We see both, and we see them compounding.

Speed is revenue. In compute, revenue follows energized gigawatts. After our base assumptions for usable share and billing, each gigawatt of nameplate capacity supports about $5.5 billion of annual revenue. Musk has said he is highly confident of 10 GW of new compute in 2027. At base, every quarter gained or lost on that capacity moves about $14 billion of revenue. That is revenue, not profit or merger value, and it counts only if a quarter is actually saved.

A board that can move a wafer, a megawatt or a launch slot in one meeting can keep that quarter’s build on pace.

We add no separate speed valuation. No measured schedule gain or separate speed valuation is added. A two-owner agreement may match the proposed decision speed.

Our model values the cost of two owners at about $6.2 billion in its base case. That figure already includes about $2.2 billion of faster revenue, $4.2 billion of avoided capital spending and $1.2 billion of financing benefit, net of $1.4 billion of integration cost. It assigns no value to orbital compute, solar consolidation, Starlink in vehicles or Megapack supply planning. A positive base case is a reason to examine the deal closely. The figure is the model’s base case, not a measured speed saving.

With capital-spending savings set to zero, the same model is about $2.0 billion. It excludes overhead savings and assigns no value to strategic options. The low, base and high model values are −$1.6 billion, $6.2 billion and $56.0 billion. Those figures include modeled timing and capital savings, but no value for orbital compute or the other engines in Section 4. The joint removal of modeled ownership advantages can leave value near zero.

We propose a simple deal. Tesla survives, and SpaceX holders receive Tesla shares at the market ratio with no premium for either side. At the locked Oct 20, 2026 prices, a 10% premium transfers about $80 billion from Tesla holders to SpaceX holders before any operating gain. A 25% premium transfers about $186 billion. Those amounts are far larger than our model’s $6.2 billion base estimate of the cost of two owners, so we propose no premium. One chief executive already runs both companies, but Tesla holders and SpaceX holders still own different assets and voting rights.

The machine is a factory for intelligence, and it ends in orbit. The binding constraints on artificial intelligence are watts, wafers and launch. Ideas are abundant. Energized capacity is not.

Tesla and SpaceX are already building the pieces side by side. An orbital data center is a chip, a solar array, a power system and a radiator riding a rocket. The scarce inputs are wafers, power and time.

They are planning Terafab together. Musk has said roughly three quarters of its AI compute is planned for AI spacecraft and one quarter for Optimus. Those public planning estimates are a picture of the build, and one owner can plan the factory and its chips as one build. Cars have no stated share in that planning estimate.

The owner who pays for the factory can also decide how its next wafer is used. A joint venture can share the list. One owner can rank it on a single budget. Each board has formed a special committee to study a combination. Their review can compare one budget with the best binding two-owner plan.

If that two-owner plan publishes a cost split and handles a major mid-build change just as quickly, it would answer the case. That is the report’s own proposed test, not a company target.

We believe the committees’ review should include six protections for outside shareholders. They give the committees a real review, and they give outside holders a separate say on any deal.

The Tesla-led model gives Musk roughly 27% to 37% of the vote, above the roughly 25% he has said he wants, and short of a majority. The Independent Holder Vote gives outside holders their own say beside that.