FSD Johnny Mac

This is an extremely bullish sign that people are getting value from their bots For $5k a year you can have two or three competent mid-tier employees… I think grok bots represent $30 an hour employees on average (from $15-25 clerical and $50-100 an hour developer blended it all to $30) And this is as bad as it will ever be!

Our capex efficiency is greatly underestimated

Our capex efficiency is greatly underestimated

A Sequoia partner says SpaceX will bring in $731 billion in 2028. That's more than Walmart or Amazon sold last year, from a company that made $18.7 billion in 2025. It sounds made up until you check where his numbers came from. Almost every big one is straight off SpaceX's own earnings call, and a few are actually lower than what the company told investors. Shaun Maguire's model has SpaceX doing $36.3 billion in the first quarter of 2027. On the October call, SpaceX's CFO guided to somewhere between $33 and $37 billion. On computing power, SpaceX says it's aiming for about 13 gigawatts by the end of 2027, with a stretch plan above 20. Shaun only gave them 10.2, so he's assuming they build about 3 fewer gigawatts than they say they will. Pricing works the same way. Back in August, Elon said a watt of AI compute is worth $30 to $50 a year in revenue, and Shaun picked $40, right down the middle. Multiply those two together and you've got most of the model. Get to 21 gigawatts through 2028, charge $40 a watt, and that's $569 billion in compute revenue in one year. Starlink and everything else fill in the rest. Wall Street is nowhere close. When SpaceX went public in June, Goldman and Morgan Stanley both had 2028 at roughly $160 billion. Morgan Stanley's September model assumed 4.9 gigawatts by the end of 2027, then knocked 50% off the AI business for execution risk. So a $571 billion disagreement between a Sequoia partner and the banks that took SpaceX public comes down to one question, whether you believe SpaceX can build what it says it's building. If I were going to poke at one number, it'd be price per watt. SpaceX did $7.56 billion of AI revenue last quarter on 1.7 gigawatts, which works out to about $18 a watt over a year. Shaun's model needs that to more than double while capacity grows twelvefold, and that's the line I'd watch every quarter. What gets me is how small the rockets end up in this picture. In the model, the space business is 2.4% of 2028 revenue, Starlink is about 12%, and AI compute is around 85%. And the plan still has Starship flying 250 times that year, from a rocket that's flown fewer than 15 times so far. If Shaun's model is wrong, SpaceX's own guidance is wrong first.

Elon Musk literally told us how insane growth $SPCX is not only next year but by NEXT QUARTER(S). If you missed this, here are the VERY important parts of the SpaceX's Q3 earnings call. Pay very close attention and try to keep your brain from melting.

Macroharder and Minihard both have 220k GB300s with 800 gig nics configured for training. Gun Range will have the same. Spam (used to house Hormel products) will have ~200k GB300s, but configured for inference, so much simpler networking and less than half the fiber. I walked the whole Colossus 2 site around midnight and again this morning. Good progress by the SpaceX team and contractors. Gun Range is tracking to be operational by end of November and Spam in December.

Launching super intelligence satellites into deep space with a mass driver on the Moon can achieve >>petawatt/year capability, which would far exceed the best case value of Earth’s economy

Congrats to the @SpaceX team on over 700 Falcon 9 flights!