SpaceX IPO: the map Wall Street got wrong

GM. This is Milk Road AI, your daily briefing on the biggest stories shaping the future of artificial intelligence, technology, and investing.

SpaceX goes public Friday at a $1.75T valuation and before you decide whether that's insane or inevitable, you need to understand what you're actually buying.

Speaking of infrastructure, today’s partner is Arc by Circle.

Arc is an open Layer-1 blockchain built for real-world financial applications. Public testnet is live, App Kits are here, and mainnet is getting closer. Start building on Arc.

THE NEXT TRILLION-DOLLAR INDUSTRY IS HERE

In 1507, a German mapmaker named Martin Waldseemüller did something that had never been done before.

He drew the western hemisphere, but not perfectly.

The coastlines were wrong, the distances were off, and he placed an entire ocean where he thought there was none.

But he was the first person to name the new continent, calling it America after Amerigo Vespucci, the explorer who confirmed its existence.

Waldseemüller got the details wrong, but he got the shape of the future exactly right.

Five hundred years later, investors are staring at another map that may have the numbers wrong but could still be pointing in the right direction.

SpaceX is going public this Friday at a $1.75T valuation, the largest IPO in market history.

Source: Giphy

Goldman Sachs's bull case requires $474B in revenue by 2030, a 25x increase in five years that has never happened for a company this size.

What SpaceX actually is

Before you can have an opinion on the valuation, you need to understand what you're actually buying, because if you think this is a rocket company, you've already made the mistake.

The S-1 breaks into three segments.

Starlink, the satellite internet business, generated $11.4B in 2025, accounting for 61% of total revenue, with a 38.6% operating margin.

That margin beats AT&T, Verizon, and every traditional telecom that spent the last 30 years building copper wires into the ground.

The launch business, the actual rockets, the thing everyone calls SpaceX, contributed $4B and is currently losing money, because they are burning capital to scale Starship into something that will eventually make every other launch vehicle on earth look like a horse-drawn carriage.

The AI segment, which is xAI and Grok following the acquisition of Elon Musk's AI company in February 2026, brought in $3.2B and lost $6.4B because they are building the infrastructure for a business that does not yet have a name in most financial models.

SpaceX generated $18.7B in revenue in 2025, growing 33% year-over-year, while investing aggressively with a $4.9B net loss and $20.7B in capital expenditures to fund future growth.

On paper, that looks like a company on fire.

In context, it looks like Amazon in 2012, when Bezos was burning billions on AWS infrastructure that Wall Street kept calling reckless, right up until AWS became the most profitable cloud business in history.

The crown jewel Wall Street keeps undervaluing

Starlink has surpassed 12M subscribers across 160+ countries, growing into the world's first truly global satellite internet network.

Source: @TeslaZoa

The bears keep pointing at one number, ARPU, or average revenue per subscriber, has fallen from $99 per month in 2023 to around $64 today, which is an 18% decline in two years.

At $100 per month, Starlink can serve maybe 200M people globally, the segment of the world wealthy enough to pay American broadband prices.

At $10 per month, the total addressable market is 5B people.

SpaceX is trading margin compression now for market dominance later, exactly the way Uber traded profitability for geographic density in 2015, and exactly the way that trade looks obvious in retrospect.

Goldman Sachs projects Starlink revenue alone will hit $144B by 2030.

That is more than Comcast, AT&T, and T-Mobile's current total revenue from a business that didn't exist six years ago.

And unlike those telecoms, which spent decades building physical infrastructure into the ground that now depreciates every year, Starlink's constellation is largely a sunk cost.

Once the satellites are up, every new subscriber is near-pure margin.

In Q1 2026, Starlink generated $1.19B in operating income on $3.26B in revenue, a 36.5% margin on a business that is still in its growth phase.

There's also the wildcard most analysts are modeling at zero, Direct to Cell.

Starlink's mobile service had 16M unique connections as of March 2026, adding 18,000 new users every single day.

SpaceX is targeting hundreds of millions of devices.

At $5 to $10 per month in carrier revenue sharing, 200M D2C connections would add $12 to $24B in annual revenue by 2030.

Almost pure margin, because the constellation is already in orbit.

The revelation that changed the room

Just four days before the Nasdaq debut, Elon Musk and SpaceX engineer Ian Dahl sat down with institutional investors and said something that shifted how this company gets valued.

SpaceX plans to deploy up to 1M satellites that function as computing nodes in orbit, powered by solar energy and cooled naturally by the vacuum of space.

Source: Tesliens

If successful, it could eliminate many of the constraints facing terrestrial data centers, including power grid bottlenecks, water requirements, and years of permitting delays.

And this is the part that matters: most of the required technology already exists.

The solar arrays are already flying on Starlink V3 satellites, while laser inter-satellite links are already operating across the constellation.

Thermal management may actually be easier in orbit, where the vacuum of space provides a natural cooling environment.

The compute satellites themselves could also be simpler than broadband satellites since they would not require phased-array antennas for internet connectivity.

The biggest unanswered question is not the technology but rather the launch economics.

Everything hinges on Starship reaching the launch cadence and cost targets needed to deploy infrastructure at scale.

Beyond that, much of the orbital compute thesis is simply existing technologies being combined in a new way.

And then there are the contracts that make this vision far easier to take seriously.

Google is paying $920M per month for AI compute capacity while Anthropic is paying $1.25B per month to lease 325,000 Nvidia GPUs at SpaceX's Colossus II facility in Memphis, through May 2029.

Together, those two contracts represent more than $25B in annual revenue, more than SpaceX's entire 2025 top line before a single orbital AI satellite has launched.

Goldman Sachs's bull case has the orbital AI segment reaching $322B in revenue by 2030.

That number sounded delusional in January, but Google and Anthropic are currently writing the checks that make it less delusional every month.

THE NEXT TRILLION-DOLLAR INDUSTRY IS HERE

Now for the part we'd be doing you a disservice to skip.

93x revenue is the most expensive large-cap IPO multiple in market history, and the people saying so are not wrong.

Saudi Aramco, the only comparable IPO by size, had $360B in revenue and was profitable.

SpaceX has $18.7B and a $4.9B net loss.

Goldman's base-case scenario requires revenue to grow 25x in five years, and that has never happened for a company at this scale.

The math that gets you to $474B by 2030 requires the orbital AI segment to essentially become the largest software business ever built, from a standing start, in 48 months.

The orbital data center plan, 1M satellites, has an implied capital cost of roughly $2T.

That is the entire IPO valuation, spent on satellites, before earning a dollar of AI revenue at scale.

The thermodynamics of cooling hyperscale compute in orbit remains a legitimate unsolved physics problem.

SpaceX carries $29.1B in long-term debt and burns roughly $13.8B in free cash flow annually, with Goldman not expecting positive free cash flow until 2031.

But here is what the bear case assumes away, and why it might be the wrong assumption.

The global AI buildout has one constraint nobody has solved, and it’s power.

Every hyperscaler from Amazon to Google to Microsoft is racing to build AI infrastructure, but power availability has become the industry's biggest bottleneck.

Permitting can take years, water resources are increasingly constrained, and utility interconnection queues stretch nearly a decade in some regions.

If SpaceX can deploy orbital compute at competitive economics, they don't improve that situation but rather route around it entirely.

That is the bypass to the central chokepoint of the entire next decade of AI infrastructure.

The companies that solve the power problem will capture disproportionate value from the AI buildout, and SpaceX is the only entity on earth attempting to solve it from orbit.

Starship is the variable that makes or breaks every model.

When it reaches commercial operations at $20M per flight with 200 metric tons of payload, it will be 200 times cheaper per kilogram than NASA's Space Launch System and 70 times cheaper than Falcon Heavy.

That doesn't improve the economics of space. It creates an economy where none existed.

By June 2026, SpaceX will have completed 12 integrated Starship test flights with consistent orbital insertion and first-stage catch.

My honest verdict

SpaceX is reportedly more than 4x oversubscribed ahead of its IPO, which means investors should expect extreme volatility once shares begin trading.

Stocks with this level of demand often experience a powerful initial run-up followed by sharp pullbacks as early investors take profits and the market searches for a fair valuation.

If you're planning to buy, proceed with caution.

Chasing momentum on opening day has historically been a difficult strategy, especially for highly anticipated IPOs.

That said, the long-term opportunity remains compelling.

SpaceX is not a stock you buy because of next quarter's earnings report but it’s rather a bet on Starship, Starlink, orbital infrastructure, and the possibility that space becomes the next major economic frontier.

There will likely be periods of euphoria and periods of disappointment.

The path from here to a multi-trillion-dollar valuation will almost certainly be volatile but for investors with a 10–20 year time horizon, SpaceX could ultimately prove to be one of the most important companies of the century.

If you want to see the exact moves we're making around SpaceX, including how we're approaching the IPO, potential entry points, position sizing, and our highest-conviction opportunities across AI and technology, come join us inside Milk Road PRO.

Alright, that's it for this edition of Milk Road AI. We want to hear from you.