SpaceX is telling IPO investors it is an enterprise AI company. The numbers tell a different story.

Reuters dropped an exclusive today based on SpaceX's IPO filing, and I have been sitting with it for a few hours trying to figure out whether I am looking at the most ambitious corporate repositioning in history or the most aggressive piece of IPO storytelling I have ever read. Probably both.

Here is the line that stopped me cold, pulled straight from the filing: "We believe we have identified the largest actionable total addressable market in human history."

That is SpaceX. The rocket company. Telling public market investors that the prize it is chasing is bigger than anything anyone has ever chased, and that most of it is not in space at all.

What SpaceX is actually pitching

The filing puts the total addressable market at $28.5 trillion. More than 90 percent of that number, roughly $26.5 trillion, is tied to AI. And $22.7 trillion of the AI piece is specifically enterprise AI. Selling agents and copilots and automation platforms to corporate buyers.

So when SpaceX walks into IPO roadshows this summer, targeting roughly a $1.75 trillion valuation and about $75 billion raised, which Reuters notes would make it the largest IPO in history, the story is not going to lead with reusable rockets. It is going to lead with Grok Enterprise, forward deployed engineers, and an AI TAM that dwarfs every other line in the filing.

I want to be fair here. Big TAM slides are a normal part of IPO theater. Every company at this scale inflates the market they could theoretically serve. But the ratio matters. When more than 90 percent of your pitched opportunity sits in a business you bought two months ago and that is still losing billions, the narrative is doing a lot of heavy lifting that the income statement is not.

The financial picture underneath the pitch

Let me lay out the 2025 numbers as reported, because they matter.

Total revenue: $18.7 billion. Starlink produced $11.4 billion of that, and generated $4.4 billion in operating profit. That is the part of SpaceX that looks like a real enterprise business. Recurring revenue, massive installed base, cash generative.

The AI unit is the opposite. Operating loss of $6.4 billion in 2025, up from $1.6 billion the year before. That is a loss line that roughly quadrupled in twelve months. SpaceX as a whole lost $4.9 billion for the year even with Starlink carrying the load.

Capex is where the strategic shift screams at you. Total 2025 capex was $20.7 billion. AI alone accounted for $12.7 billion of that, which the filing notes is more than space and connectivity combined. Read that sentence again. A rocket and satellite internet company spent more on AI in one year than it spent on rockets and satellite internet.

This is not a company that is going to diversify into AI. This is a company that has already made the call and is using the IPO to finance what comes next.

The xAI acquisition is load bearing

SpaceX acquired xAI in February. That is where Grok, Grok Enterprise, and whatever comes after them now live. The filing says SpaceX plans to build on those products, stand up a specialized enterprise salesforce, and ship forward deployed engineers to help customers adopt AI.

Forward deployed engineers is a Palantir playbook, and it works when the software alone is not enough to get enterprises over the integration hump. That tells you how SpaceX is thinking about this. They are not planning to win the enterprise AI race by having the best model. They are planning to win it by putting humans on the ground inside customer accounts and closing the messy last mile.

The filing also mentions Macrohard, an agentic or autonomous platform in development with Tesla. The details are thin in the source material, so I am not going to speculate about what Macrohard actually is. But the name alone tells you how Musk wants you to read the strategy. He is not positioning this against Anthropic or OpenAI. He is positioning it against Microsoft.

Which brings up the uncomfortable question. Reuters points out that enterprise AI is currently dominated by Anthropic and OpenAI. Those two companies have years of enterprise traction, huge sales organizations, and deep integrations into the Fortune 500. SpaceX is showing up late, through an acquisition that is still bleeding cash, with a pitch that the market is effectively infinite. That is a tough entry.

The chip problem is its own category of risk

The Yahoo syndicated version of the Reuters reporting surfaces a second thread that I think matters as much as the enterprise AI pitch. SpaceX warned investors in the filing about heavy AI spending and possible chip supply constraints. And the filing lists manufacturing its own GPUs as a substantial capex item.

That connects to a project called Terafab, an advanced AI chip manufacturing complex in Austin being pursued by SpaceX, xAI, and Tesla. Musk has said publicly that Terafab targets chips for cars, humanoid robots, and space-based data centers. The filing quotes are carefully hedged in a way IPO filings usually are when the risk is real. "We do not have long-term contracts with many of our direct chip suppliers." And this one, which I think is the honest sentence in the whole filing: "We expect to continue sourcing a significant portion of our compute hardware from third-party suppliers, and there can be no assurance that we will be able to achieve our objectives with respect to TERAFAB within the expected timeframes, or at all."

In plain English, SpaceX is telling investors that Terafab may or may not work, that they cannot guarantee it will hit its timeline, and that they will keep buying chips from other people while they try. It is also unclear, as Reuters notes, whether GPU is being used literally or as shorthand for AI processors broadly.

Chip manufacturing is brutally hard. Nvidia designs and TSMC fabricates for a reason. The list of companies that have tried to vertically integrate advanced chip manufacturing and succeeded is short. The list that have tried and blown money is long.

What I actually think is going on

Here is my read, and I want to be clear this is analysis and not reporting.

SpaceX had to file eventually. Once you file, your story has to justify your valuation. A $1.75 trillion valuation cannot be justified by rockets and Starlink alone, even with Starlink doing $11.4 billion in revenue and $4.4 billion in operating profit. Reuters notes that a source familiar with the company's finances said a sober valuation of the visible businesses would not be in the same ballpark as what the market may set. I believe that source.

So the filing has to reach for something bigger. Enterprise AI is the only market on the planet large enough, vague enough, and hot enough to carry the weight of that valuation. It does not matter that xAI is losing $6.4 billion a year. It matters that the TAM slide says $22.7 trillion, that the filing quotes itself using the phrase "largest actionable total addressable market in human history," and that the retail investor base shows up to buy the biggest IPO ever.

The harder question is whether this is a story that survives contact with the first earnings report after the IPO.

A few things I will be watching

Starlink's margins. If the one business that actually prints cash starts to soften because AI capex is pulling resources, the whole pitch gets wobbly fast.

The xAI loss curve. $1.6 billion in 2024 to $6.4 billion in 2025 is the wrong direction. If 2026 comes in at $10 billion or worse, IPO investors are going to start asking whether enterprise AI TAM claims pay any bills.

Terafab milestones. Any real signal on yield, tape-out, or customer commitments would change the risk picture. Silence for twelve months after the IPO would too, in the other direction.

Enterprise wins against Anthropic and OpenAI. Not pilot programs. Signed, eight-figure, multi-year contracts. That is the only thing that turns the $22.7 trillion slide into something you can model.

I am not rooting for this to fail. SpaceX has earned a lot of benefit of the doubt in its actual rocket business. But the IPO filing is asking investors to believe in a version of this company that does not really exist yet, funded by capital raised on a TAM claim that is doing almost all the work. That is a lot to carry into a public offering.