SpaceX's $731 billion bull case is a Sequoia shareholder's model, and it now prices a Tesla merger

The most-quoted SpaceX▲+3.42% — SpaceX, up 3.42 percent today number this week was published by a man who tells you not to trust it. A model by Shaun Maguire↗, a partner at the venture capital firm Sequoia, takes SpaceX to $731.1 billion of revenue in 2028, and it carries its own health warning: "Sequoia Capital has a large position in SpaceX so I'm extremely biased," he wrote, calling the numbers his 25th-to-50th-percentile case, the lower-to-middle band of his own scenarios ¹.

Elon Musk endorsed the direction over the weekend, replying "I think we can do better" ¹. Read the model as what it is: a large holder's estimate, published while the share price it flatters feeds the exchange ratio of a live Tesla merger ².

The model is the moderate document in the stack

It puts 2.5 gigawatts of ground-based compute in place by the end of this year ¹. The 2028 total leans on $569 billion of ground-based AI computing revenue ¹. Capex peaks at $369 billion that year, after two years of cash burn, with cash flow turning positive only in 2028 ¹³.

Musk had passed the model before it was written. On the October 20 earnings call↗ he put annual recurring revenue, one month or quarter of revenue multiplied out to a year, at $500 billion to $1 trillion by the end of 2027 ⁴⁵, and posted that night that $1 trillion of 2028 revenue was "no longer impossible" ⁵. His own targets already run past the model, so the shareholder's forecast is the cautious end of the official story.

What SpaceX has booked is real, and much smaller

Third-quarter revenue was $14.6 billion, more than double the year before ⁴. The Super Intelligence compute segment brought in $7.6 billion, passing the connectivity segment that houses Starlink ⁴, and backlog stands at $63 billion ⁴. At current pricing, new compute pays for itself in under a year, management said, with prices tracking just above the top of the $30-to-$50-per-watt band quoted last quarter ⁴. Maguire assumes roughly $40 billion of annual revenue per gigawatt, which sits inside that band ⁷.

The contested link is how fast the gigawatts go in

Nameplate compute rose 0.3 GW last quarter, to 1.7 GW↗ ⁸. Musk says he would be surprised by less than 13 GW of compute by the end of 2027 ⁴. New Street Research, sketching thousands of small Starlink-linked compute sites to sidestep grid queues, still calls pace the real constraint and doubts 10 GW can land in 2027 ¹.

In revenue terms, SpaceX has guided the first quarter of 2027 to $33 billion to $37 billion ⁴. Maguire's model books $273.8 billion for that year ¹; on our arithmetic, the three quarters after March would have to run at more than twice the pace just guided ¹⁴. If the build is real, the bill is concentrated: analyst Kyle Reidhead counts about $470 billion of capex over the next five quarters, with Nvidia▲+1.00% — Nvidia, up 1.00 percent today capturing 83 to 89 percent ³, on top of the $40 billion of debt SpaceX has already raised to buy GPUs ³.

The forecast now prices a merger

SpaceX closed Friday at $134.72, just under the $135 price of its June IPO ²³. On Bank of America▲+0.15% — Bank of America, up 0.15 percent today analyst Alexander Perry's share counts, TheStreet works an all-stock swap at that price out to about 2.97 SpaceX shares per Tesla▲+2.23% — Tesla, up 2.23 percent today share, leaving Tesla holders close to 46 percent of the combined company with no premium ².

Perry's own no-premium case, struck at October 20 prices, gave them 42.4 percent ²: the market's move has already handed Tesla part of the premium he expects its committee to negotiate ². A charter vote in which abstentions count against the deal, and Chinese regulators, still stand in front of it ².

The orbital promise sits beyond any checkable quarter

Musk's weekend line that terrestrial compute will be "peanuts" next to orbital compute↗ pushes the promise past 2028, when he says capacity added in orbit starts to surpass what is built on the ground ³. That is also where cooling, radiation and fast chip cycles are unsolved, and no quarterly report can check the claim ⁷.

The demand side is already contracted, in that backlog ⁴, and whoever actually deploys the gigawatts is doing the work that turns a forecast into revenue. The checkpoints are on the calendar: December's $100 billion annualized-revenue goal ⁴, the first-quarter report against guidance, and the gigawatts added each quarter ⁴. Until those print, the only line in the model that is not a forecast is the disclosure ¹. Price that first.