Musk's Vertical Stack and the Coming Age of the Billion-Dollar Monthly Compute Bill

SpaceX's IPO filing isn't just a capital markets event — it is the logical culmination of a decade in which the world's wealthiest industrialist built an empire structured around zero-sum positioning against his own former partners.

By Monexus Staff WriterUS5-minute read20 May 2026☆ Save ↗ Share ⎙ Print

SpaceX filed its IPO paperwork on 20 May 2026, and the markets have responded with the appropriate awe. A company that was once a punchline in aerospace circles — a billionaire's vanity project built around the premise that rockets could be landed, reused, and rendered cheap — is now on track for the largest public offering in history. The arithmetic is straightforward: a private valuation north of $300 billion, a near-monopoly on US launch capacity, and a captive government customer base that includes NASA, the Department of Defense, and a growing list of commercial satellite constellations. Elon Musk, already the world's wealthiest individual by a comfortable margin, stands to become considerably wealthier still.

But the IPO filing deserves to be read alongside two other moves that landed in the same news cycle. Anthropic disclosed that it is paying xAI — Musk's AI company — $1.25 billion per month for compute access. Meanwhile, OpenAI confirmed that it is barreling toward a public listing, potentially by September 2026, days after a lawsuit that Musk brought against the organization collapsed. The three events are not coincidental. They describe a single strategic architecture: Musk building a vertical stack from silicon to orbit, while his former collaborators at OpenAI prepare to go public, and his AI upstart Anthropic simultaneously rents compute from his infrastructure.

The Vertical Stack as Power Move

There is a difference between a conglomerate and a vertical stack. A conglomerate accumulates businesses for financial reasons; a vertical stack accumulates them because each layer reinforces the others. Musk's current arrangement is the latter. SpaceX manufactures rockets and operates Starlink, the satellite broadband constellation. xAI builds foundation models and rents out the GPU compute that Musk has spent years stockpiling. Tesla — not directly in this story but structurally adjacent — consumes silicon, training data, and inference compute. The parts talk to each other in ways that a spreadsheet cannot easily quantify.

The Anthropic deal makes this concrete. Anthropic is widely considered one of the most capable AI labs in the world, backed by Amazon to the tune of several billion dollars. Yet it is paying Musk's xAI $1.25 billion every month for access to compute. That is not a partnership between equals. That is a dependency — and dependencies in technology translate directly into leverage.

The IPO dynamic sharpens the problem. SpaceX going public would bring scrutiny, yes — quarterly earnings calls, institutional shareholders, a market cap that demands explanation. But it would also give Musk access to a new tier of capital, at public-market valuations, for a company that already generates reliable revenue from government contracts. Every dollar raised at the IPO valuation is a dollar that can be deployed into xAI, into Starlink expansion, into the next layer of the stack. Public shareholders become unwitting co-investors in an infrastructure empire that Musk controls.

OpenAI's Uncomfortable Parallel

The irony of OpenAI's IPO timing is difficult to miss. Musk sued the organization in 2024, arguing that its transition from a nonprofit research lab to a capped-profit commercial entity betrayed the founding compact. The lawsuit failed. But its shadow falls over the IPO nonetheless. OpenAI was created, in part, to ensure that artificial intelligence did not fall under the control of any single individual or corporation. It was structured as a nonprofit with a commercial arm, a governance arrangement designed to prevent exactly the kind of concentrated power that Musk is now assembling.

The IPO would complete the inversion. Sam Altman's organization, which spent years arguing that it was different from a ordinary tech company, is preparing to list itself on public markets. The governance safeguards that distinguished it from Musk's Tesla-centric empire are being stripped away by the requirements of investor liquidity and public capital. Whether that represents a failure of OpenAI's original mission or simply the inevitable result of competitive pressure is a legitimate question. What is not a question is that the AI safety organization most closely associated with responsible development is now structurally identical to the company it was founded to prevent.

The Compute Trap

There is a structural logic to the current moment that deserves examination. AI labs are not software companies in the traditional sense. They are compute companies — their output is determined by how many chips they can afford to train on, and their competitive position is determined by who controls access to those chips. This is not a secret. Nvidia's market capitalization, the GPU rental economics of AWS and Azure, the compute agreements that have defined the last three years of AI dealmaking — all of it reflects the same underlying reality.

Musk understood this earlier than most. By positioning xAI not merely as a model developer but as a compute provider — selling capacity to Anthropic at $1.25 billion per month — he has inserted his infrastructure into the supply chain of his competitors. The labs that depend on him for training compute are, by definition, labs that cannot easily exit that relationship without absorbing a massive cost increase. The vertical stack is not just a commercial advantage; it is a moat.

This is the real story beneath the IPO filings. The largest public offering in history and the AI industry's compute supply chain are not separate topics. They are two facets of the same consolidation pattern — a handful of individuals and their affiliated entities capturing control over the critical nodes of the next generation of technology infrastructure, from the chips that train models to the satellites that distribute them.

What Comes Next

The regulatory apparatus has not caught up. SpaceX holds a dominant position in US launch capacity; xAI is one of three or four labs capable of training frontier models; Starlink controls the broadband connectivity of large swaths of the globe. None of these entities faces meaningful antitrust scrutiny in the traditional sense, because they operate in different formal markets. But the interconnection between them — the compute agreements, the government contracts, the shared control by a single principal — describes a concentration of economic and political power that the existing framework was not designed to address.

Whether that matters depends on what you believe AI is for. If the technology is primarily a consumer product — chatbots, image generators, copilots — then the competitive dynamics of the current market may be sufficient to discipline any single actor. If, however, AI is the general-purpose technology its proponents claim — the infrastructure layer for economic growth, military capability, and political influence for the next half-century — then the structure of ownership at this moment in time will shape the structure of power for decades. The IPO filings landing this week are not just capital markets events. They are architecture.

This desk tracked SpaceX, xAI, and OpenAI's parallel IPO tracks across four wire services on 20 May 2026. The dominant framing in US financial media emphasized the scale of the SpaceX offering and its implications for Musk's personal wealth. Monexus placed equal weight on the Anthropic compute agreement and the structural convergence of three Musk-adjacent entities toward a shared infrastructure position.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

https://t.me/rnintel/12345