Roko's Basilisk
Posts
The Engine Behind The Empire

The Engine Behind The Empire

Plus: NVIDIA's agent PCs, Microsoft's AI dev mode, a fresh AI rule.

Roko's Basilisk
June 03, 2026

Here’s what’s on our plate today:

🧪 How Starlink is funding Musk's biggest ambitions.
📰 NVIDIA's PC power play, Windows becomes an AI dev platform, and a new twist in AI regulation.
🛠️ Three tools worth trying: Dust, Krea, Lavender.
🗳️ Poll: What are you really buying in the SpaceX IPO?

Let’s dive in. No floaties needed…

Launch fast. Design beautifully. Build your company's website on Framer.

Framer helps teams design, build, and launch their marketing sites lightning fast. With the ability to publish hundreds of CMS pages in a single click, operate at a global scale with seamless localization, and even host unified content across multiple domains, teams have never been able to ship faster.

Trusted by companies like Miro, Bilt, and Perplexity

_{*This is sponsored content}

The Laboratory

TL;DR

One business, three bets: Starlink generated $11.4B in revenue and $4.4B in operating income in 2025. The AI and space segments together led to a $4.9B consolidated net loss for SpaceX.
Growth without margin: Average revenue per subscriber dropped from $99 to $66 as residential markets replaced high-paying aviation and maritime customers. Subscriber counts climb; unit economics quietly deteriorate.
The AI reframe: Acquiring xAI lets SpaceX trade at 94 times trailing revenue, a multiple that requires orbital AI compute to materialize by 2028, before Starship has delivered a single commercial payload.
Governance by design: Musk controls 85% of voting power, and revised Nasdaq index rules could auto-enroll index fund holders before they read the prospectus.
The real stake: Investors are financing a multi-planet empire through a broadband business whose cash flows increasingly determine what any of those ambitions can afford.

How Starlink is funding Musk’s biggest ambitions

Somewhere between 30,000 and 35,000 feet above the Atlantic, a commercial aircraft is probably streaming this article to a passenger who has no idea their in-flight internet is being routed through a constellation of low-orbiting satellites operated by a rocket company in Texas. That routing happens via Starlink, the satellite broadband network that SpaceX has spent the better part of a decade building into the world’s largest constellation by satellite count.

However, the story of SpaceX may, in the coming days, become more about the financial structures that helped achieve the biggest IPO in history than about the technical expertise that enables airborne passengers to access the internet.

SpaceX’s IPO filing is packed with ambitious visions of asteroid mining, lunar infrastructure, and Mars settlement, raising questions about how much of its valuation rests on existing businesses versus future possibilities. Photo Credit: Getty.

When SpaceX filed its S-1 registration statement with the Securities and Exchange Commission, the coverage predictably leaned toward superlatives: a $1.75T target valuation, a June 12 Nasdaq listing under the ticker SPCX, and a raise of up to $75B that would nearly triple Saudi Aramco’s 2019 record. The accompanying prospectus referenced asteroid mining, lunar mass drivers (electromagnetic catapults designed to launch payloads off the Moon’s surface without rockets), and a Mars colony. In places, it reads more like a manifesto than a financial filing.

Underneath all of it is a satellite internet business with 10.3M subscribers and a shrinking revenue-per-customer problem that the filing discloses but does not dwell on.

For years, SpaceX’s public image has been shaped by rockets, launch videos, and Mars timelines. The IPO filing changes the perspective slightly, with public filings forcing companies to reveal which parts of the business actually generate money, which consume it, and how the entire structure holds together financially.

What the S-1 actually says

As of 2026, SpaceX operates in three segments. These are connectivity (Starlink), space (rockets and launch services), and AI (xAI, Grok, and X, the platform formerly known as Twitter, which SpaceX acquired in February 2026). This diverse portfolio, however, masks the fact that the financial profiles are radically different.

Connectivity generated $11.4B in revenue in 2025, with $4.4B in operating income. Space generated $4.1B in revenue but ran at an operating loss, with roughly $3B spent on Starship development alone. The AI segment generated $3.2B in revenue and incurred more than $6B in losses in 2025, consuming capital at a pace that turned a company with positive segment-level earnings into one with a $4.9B consolidated net loss for the year. In the first quarter of 2026, the net loss was $4.28B, nearly matching the full prior-year loss in a single quarter.

Which means that just one business is profitable, while the other two are not. For every dollar of Starlink operating profit, the rest of the company is committing roughly two dollars in losses and capital expenditure. It is a balance point that depends entirely on Starlink continuing to grow fast enough to absorb the burn. And that dependence would be less concerning if Starlink’s economics were becoming stronger as it scaled. The filing, instead, shows a business expanding rapidly while quietly earning less per customer.

Starlink’s average revenue per subscriber reportedly fell from $99 per month in 2023 to $66 by early 2026. The reason for this shift is early growth driven by high-paying customers in the aviation and maritime sectors. However, new growth is increasingly coming from lower-priced residential markets. So even as subscribers have been doubling, revenue per subscriber has been falling. Meanwhile, total revenue continues to climb because the first trend is currently outpacing the second, but that relationship is not guaranteed to hold.

The revenue model would historically have forced public markets to value the company conservatively, but SpaceX is instead attempting one of the richest technology valuations in history. To understand why this is happening, one would need to examine the company’s AI narrative more closely.

The narrative layer

When SpaceX acquired xAI, the acquisition did something specific to its story: it gave the company a reason to be valued like an AI infrastructure business rather than a satellite internet provider. The most ambitious version involves running artificial intelligence workloads directly from orbit, which current AI infrastructure companies do not do.

At roughly 94 times trailing revenue, SpaceX’s valuation exceeds that of nearly every major public technology company. That valuation multiple assumes a future in which SpaceX’s orbital AI computing vision evolves into a meaningful business by the end of the decade. The concept centers on satellites equipped with onboard processors, powered by space-based solar energy, and capable of performing AI inference directly in orbit rather than relying solely on terrestrial data centers.

The S-1 states that deployment could begin as early as 2028, but Starship, the rocket that would make such deployment possible at scale, has not yet successfully delivered a single commercial payload.

Even though the bull case is plausible, given grid-constrained terrestrial data centers and SpaceX’s unique position to own both the rockets and the constellation, the opportunity is substantial only if the underlying assumptions hold. The difficulty with this is that the valuation depends on several large technological and financial bets succeeding simultaneously rather than independently.

It also rests on the assumption that the capital being poured into the AI segment is being deployed productively rather than recirculated within Musk’s wider business empire. Assessing that risk is increasingly difficult because the company’s finances are becoming more closely entangled with other Musk-controlled ventures.

The closed loop

That last concern has a paper trail revealed in the S-1’s related-party transactions section. The filing revealed extensive transactions between Musk-controlled entities. SpaceX disclosed spending $697M on Tesla Megapack batteries and $131M on Cybertrucks for xAI data centers in 2024 and 2025. A SpaceX subsidiary also signed GPU lease agreements totaling more than $20B with a fund controlled by Antonio Gracias, a longtime Musk associate and future SpaceX board member. While these arrangements are disclosed, they are not independently reviewed for fairness, and oversight ultimately rests with a board appointed by Musk.

Then there is the issue of shareholder influence. Musk controls 85% of SpaceX’s voting power through Class B shares, which carry 10 votes each, while public investors will receive Class A shares with a single vote. SpaceX has also classified itself as a controlled company, allowing exemptions from certain Nasdaq independence requirements. That structure extends to executive compensation, including a performance-share grant that could be worth hundreds of billions of dollars if SpaceX reaches a $7.5T valuation and establishes a permanent Mars colony.

Under normal circumstances, investors could weigh such governance concerns before buying the stock. But many may never make that choice. Nasdaq’s revised index rules, which took effect on May 1, 2026, allow newly listed mega-cap companies to join the Nasdaq-100 after just 15 trading days. If SpaceX qualifies, millions of investors holding index funds will automatically become shareholders, without having to evaluate the prospectus, governance structure, or related-party transactions.

None of this diminishes the strength of the underlying business. If anything, the filing underscores how strategically valuable Starlink has become.

What remains open

Starlink is a genuinely exceptional business with an infrastructure moat comprising satellites deployed across 164 countries and years of operational experience. It is real and durable in ways that balance sheets do not fully capture, and the company that owns it deserves serious valuation.

The question the S-1 ultimately raises is whether public investors are being given any meaningful ability to protect the company’s most valuable asset, or whether they are simply being asked to finance a broader constellation of ambitions over which they will have little practical influence.

The current valuation estimates assume that those ambitions will succeed. And if they do not, the governance structure limits shareholder oversight.

The mechanics of index funds add another layer to this story by making people invest in this rather complicated structure without ever actively choosing to do so.

For the ones opting to invest, the assumption may be that they are buying exposure to rockets, Mars, or artificial intelligence. But the filing suggests something different. They are buying exposure to a broadband network whose cash flows increasingly determine what all of those ambitions can afford to become.

Wednesday Poll

🗳️ SpaceX is chasing a $1.75T IPO valued like an AI company. What's the real bet?

Be the future of AI

Hunting for AI roles eats hours. Careers pages, group chats, internal referrals, a dozen job boards.

The Athyna AI Job Board does the watching for you. We track openings at frontier AI companies, match them to your profile, and notify you when a role hits a 75% matching index.

Set up a profile and let the matches come to you.