Not Starship. Not Starlink. This Is the Part of SpaceX's Business That Could Lift It Into the $3 Trillion Club.

When people picture Space Exploration Technologies (NASDAQ: SPCX), the first things that likely come to their minds are its reusable Falcon rockets. What investors may not fully grasp is that SpaceX is building a three-headed machine across launch, satellite internet, and artificial intelligence (AI) computing.

Currently, SpaceX boasts a market capitalization of just under $2 trillion. The question I am wondering is whether the AI pillar can stretch that value toward $3 trillion without the broader story falling apart.

The short answer to that question is yes, it's possible. But it's not guaranteed. SpaceX's path to a $3 trillion market cap would have to run through cash flow that thus far has come from just one side of the business, but that now has a second engine quietly warming up.

Image source: Getty Images.

SpaceX today: What is paying the bills, and what is still burning cash?

The space business was SpaceX's original foundation. This segment specializes in launching rockets -- the Falcon 9 and Falcon Heavy -- that carry NASA crews and cargo, as well as private company payloads.

It's also pouring money into the development of its larger Starship launch vehicle. Through the first six months of the year, the space business generated $1.6 billion in revenue from 78 launches. This business is still losing money at the operating line, largely because Starship development is eating billions of dollars in research and development costs.

While Falcon flights are a mature, high-cadence product, most of the flights are carrying SpaceX's own payloads -- hauling more Starlink satellites into orbit -- rather than serving external customers.

Starlink is the only SpaceX segment that is consistently profitable at scale. Through June 30, it generated $7.5 billion in revenue, or roughly 60% of the company's total. This segment also produced $2.8 billion in operating income and $4.7 billion in adjusted earnings before interest, taxes, depreciation, and amortization (EBITDA).

As of the end of the second quarter, Starlink had 12 million subscribers -- twice as many as a year prior. While consumer broadband matters, Starlink is making a conscious effort to push into enterprise plans across the aviation, maritime, and government sectors.

Then there is AI. In 2025, this was the smallest of the company's three businesses, with about $3.2 billion in revenue, and had a sizable operating loss due to hefty capital expenditures. By the second quarter of 2026, the AI business had jumped to $3.4 billion in sales and flipped to positive adjusted EBITDA. This shift occurred because SpaceX stopped treating its giant compute cluster, known as Colossus, as a private lab for Grok and began renting out some of its cloud capacity.

What Starlink and Starship could look like in the long run

Starlink's goal is no longer just to provide home internet connectivity in rural and remote areas. Rather, Starlink is becoming a global connectivity utility as subscriber counts continue to grow and enterprise and government contracts scale up. Starlink's next-generation V3 satellites are designed to put more capacity in orbit per launch, creating favorable economics because the network is already operating and the incremental cost of adding more customers is relatively low.

Meanwhile, with Starship, SpaceX is trying to once again meaningfully cut the cost of launching mass into orbit.

Based on its payload capacity, one Starship flight could theoretically deploy a large constellation of V3 satellites at the same time. In addition, Starship is the vehicle SpaceX wants to use to deploy a constellation of data center satellites. It's important for investors to realize that these are test programs right now. Until Starship rockets can be launched at a rapid cadence and reused repeatedly on a routine commercial basis, this entire opportunity remains an expensive science project.

How AI can change SpaceX's profitability and valuation

SpaceX's AI business is mostly infrastructure. The company built its Colossus supercomputer to train its own models, but then discovered it could sell the spare capacity. These are the material deals it has inked so far:

• Anthropic is paying $1.25 billion a month for capacity from Colossus 1.
• Google Cloud committed to $920 million a month starting this fall for access to a cluster of Nvidia GPUs.
• Reflection AI signed an agreement for $150 million a month in a contract that is worth up to $6.3 billion over its total life to lease access to Nvidia GB300 chips at the Colossus 2 data center.
• A newer hosting agreement from an unnamed customer is slated to bring in about $1.11 billion a month beginning in December.
• During SpaceX's second-quarter earnings call, the company disclosed a six-month AI cloud deal valued at $6.7 billion. While the customer name was not named, William Blair analyst Louie DiPalma has said his firm thinks that the deal was with the Department of Defense.

CFO Bret Johnsen has said that SpaceX is on track to reach a $100 billion annual run rate by the end of this year. Elon Musk expressed even more optimism, saying in August that its AI revenue could pass the rest of the company's sales as early as this month.

This is the pivot to be on the lookout for. For years, Starlink's earnings have funded SpaceX's other ambitions. But if its AI compute leasing business continues to scale at incremental margins, the big picture stops looking like one profitable segment supporting two expensive ones. While Starlink remains a steady cash engine, AI could become a second, potentially larger one. Starship, if it works, can make both of those businesses cheaper to scale -- creating a lucrative virtuous cycle for the SpaceX ecosystem.

SpaceX currently has a market cap of around $2 trillion. This means a move to $3 trillion would require a 50% step-up. If the company exits this year with an annual run rate of $100 billion and grows this base into the mid-hundreds of billions as more GPU clusters come online and Starship proves it can reduce payload costs, a forward price-to-sales (P/S) ratio of between 15 and 20 on a $150 billion (or more) business gets you into the $3 trillion zone at the high end. Once earnings start showing up more materially, say $60 billion to $80 billion in net income, a price-to-earnings (P/E) multiple in the 35 to 40 range would land the market cap in the same neighborhood.

Admittedly, these are premium multiples. But they are not unreasonable for a company that is compounding infrastructure and building a one-of-a-kind launch system. The realistic case is not that the AI cloud business will magically make SpaceX's problems disappear. It is that demand for AI compute could help turn SpaceX from a profitable satellite company subsidizing rockets into a more diversified cash-flow machine that owns the cheapest path to orbit. That is how a $2 trillion name starts looking like a $3 trillion one without asking investors to buy into hopes or an unrealistic hype narrative.

Adam Spatacco has positions in Nvidia. The Motley Fool has positions in and recommends Nvidia. The Motley Fool has a disclosure policy.