SpaceX beat Street revenue expectations by 13% in the June quarter, which was expected given it was the company’s first quarter as a public company, which often has nice upside built into the guidance. The company gave high level guidance that landed revenue for the September and December quarters about 15% higher than the Street.

The substance of the guide, which calls for the company to exit CY26 with a $100B revenue run rate, suggests December quarter revenue of about $22B (about 17% higher than where the Street was going into the quarter) and up from the $7.8B reported in June.

While the company did not give specifics around CY27, the Street is currently at about $100B in revenue, which would be up an impressive 122% y/y. I believe that number may still be too low if the business continues to grow m/m. My logic is the company’s guide implies about $8.5B in monthly revenue in December. If you assume 5% m/m growth off that December base, CY27 revenue would be about $135B, compared to the Street’s $100B. I want to stress, I’m new to modeling the company’s financials, and there are unknowns around seasonality in the launch business. But that business is small, about 8% of revenue this year, and will remain small next year. On the other hand, the core AI business is growing fast. Take the Cursor business for example, which will account for about 15% of revenue this year, and is growing by my math at 20% m/m through the end of this year. The hosting business is also growing m/m, at about a 30% clip. My point is that most of SpaceX’s businesses appear to be growing rapidly m/m, and that is likely to continue into CY27. My $135B is about 35% above the Street’s $100B.

Despite this good news, shares sold off by 14% the following day on the capex guide that suggested capex for the back half of the year will be about 65% higher than where the Street was. Most investors pushed that level of spending into CY27 given the company said they will be “aggressively” investing in capex going forward. That implies that investments for next year will be up about 120% compared to the Street’s previous expectations of up 91%.

My take: Investors are missing the point. Of all of big tech, SpaceX should be the one most aggressively spending given their vision around vertical AI, which includes having in house everything from power to chips, models, and distribution, and is the most aggressive in the world. I would view light investments in capex, growing next year by 50%, as a red flag, and see the quickened pace as a positive.

Taking a step back, the most important takeaway from the quarter is the high level message: SpaceX is on a mission to build the world’s first vertical AI company. While Elon did not use those words, that vision was the substance of the IPO roadshow, and remains front and center based on both the progress each segment is making along with the heavy investments the company will be making.

The key areas of focus highlighted on the earnings call in each segment in pursuit of building a vertical AI company include:

Launch: The launch business currently brings about 2,500 tons of payload to space per year through Falcon, representing about 80% of global payload brought to space. Elon expects the next generation launch vehicle, Starship, will have the launch frequency and payload to eventually take that to 1m to 10m tons per year. Even if the company comes in at 500,000 tons a year in a decade, it would be a huge win all but ensuring SpaceX remains the primary access point to space.

Starlink: Version 3 satellites launch later in August and will bring faster speeds, which means more opportunities for price tiering. Version 3 is one step on a bigger mission. Elon believes Starlink will become the world’s primary internet access point, and said it is not out of the question that Starlink carries a majority of the world’s internet in under 10 years, limited to countries where SpaceX is permitted to operate.

Grok: By the end of this year, there will likely be 3 major updates to Grok, concluding with Grok 5 which is expected to be out late in December and trained on all of SpaceX’s engineering data. Taking a step back, Grok has been climbing the AI leaderboards in recent months, and generally lands around 5th place, compared to 10th earlier in the year. This progress suggests something is working when it comes to their training methodology. I expect by early next year, Grok will be a top three model, particularly strong in engineering, a field that yields the highest prices for advanced tokens.

Disclaimer