SpaceX filed for the largest IPO in American history — seventy-five billion dollars at a one-point-seven-five trillion dollar valuation. The filing is not just a liquidity event. It is the stress test that every great infrastructure buildout must eventually face.
SpaceX filed a confidential S-1 registration statement with the SEC on April 1. Target valuation: one point seven five trillion dollars. Target raise: seventy-five billion. Expected listing: June, on the Nasdaq. Goldman Sachs, JPMorgan, Bank of America, and Morgan Stanley are underwriting. Up to thirty percent of shares are earmarked for retail investors.
Seventy-five billion dollars is three and a half times the largest American IPO in history — Alibaba's twenty-one point eight billion on the NYSE in 2014. It is nearly three times Saudi Aramco's twenty-five point six billion in Riyadh in 2019. No offering in any market has approached this scale.
The filing followed the merger. In February, SpaceX absorbed xAI in a share exchange that valued the combined entity at one point two five trillion dollars — the largest merger of all time. Two months later, the IPO prices the same entity forty percent higher. The markup is not based on new revenue. It is based on the belief that public markets will pay a premium for access to an asset class that has been private for twenty-three years.
The Concentration
The SpaceX filing did not arrive in isolation. It arrived at the peak of the most concentrated capital allocation cycle in venture history.
Global venture capital hit two hundred ninety-seven billion dollars in the first quarter of 2026 — an all-time record. Eighty-one percent went to artificial intelligence. Four deals consumed sixty-four percent of all global venture funding: OpenAI at one hundred twenty-two billion, Anthropic at thirty billion, xAI at twenty billion, and Waymo at sixteen billion. Four companies. Two hundred forty-six billion of the two hundred ninety-seven billion in late-stage funding went to five hundred eighty-four deals — a two hundred five percent increase from the prior year, concentrated in fewer hands.
OpenAI closed its round at an eight hundred fifty-two billion dollar valuation with twenty-five billion in annualized revenue. Nine hundred million weekly active users. Fifty million paying subscribers. Enterprise revenue above forty percent of total and accelerating toward parity by year-end. Anthropic reached three hundred eighty billion at nineteen billion in annualized revenue, with Claude Code alone generating two and a half billion.
The combined private valuations of SpaceX, OpenAI, and Anthropic exceed three trillion dollars. Each is expected to go public. The question is not whether the capital markets can absorb these companies. The question is what the capital markets will reveal about them.
The Pattern
Every great infrastructure buildout follows the same financial arc. Private capital funds the construction. The construction exhausts private capital. The builders go public.
The railroads followed this arc in the 1870s. Union Pacific listed on the NYSE in September 1870, sixteen months after completing the transcontinental railroad. Railroad stocks came to dominate American exchanges — nine of eleven original Dow Jones Transportation Average components were railroads. One hundred seventy thousand miles of track were built between 1871 and 1900 through massive speculation. The Panic of 1873 was triggered by railroad overbuilding. The companies failed. The infrastructure persisted.
The telecoms followed the same arc in the 1990s. The Telecommunications Act of 1996 triggered an IPO frenzy — hundreds of new entrants, banker-driven roll-ups, over eighty million miles of fiber laid. Global Crossing hit a forty-seven billion dollar valuation by 1999 and filed for bankruptcy by January 2002 with twelve point four billion in debt. Eighty-five percent of the fiber was still dark by 2005. WorldCom's false claim that internet traffic doubled every hundred days — the reality was roughly doubling every year — drove the overbuild that bankrupted the builders and created the backbone for cloud computing, streaming, and broadband.
The pattern is structural. The infrastructure survives the financial collapse that follows its construction. The question is never whether the infrastructure is valuable. It is whether the companies that built it capture that value — or whether they are the bridge between two eras of ownership.
The Stress Test
Private markets and public markets impose different disciplines. The difference is not transparency — private companies disclose to their investors. The difference is adversarial scrutiny.
Public markets have short sellers. Private markets do not. Public markets have quarterly earnings calls where analysts ask uncomfortable questions about unit economics. Private markets have board meetings where aligned investors ask encouraging questions about growth. Public markets have fiduciary duty running in both directions — officers to shareholders, analysts to clients. Private markets have fiduciary duty running in one direction, toward a small group of sophisticated participants who have already decided to believe.
The three trillion dollars in combined private AI valuations were built in an environment without short sellers, without quarterly reporting cadence, without the adversarial price discovery that forces companies to justify their economics to people who profit from finding flaws. The IPO is the moment that changes.
SpaceX has real revenue — Falcon 9 launches, Starlink subscriptions, government contracts. The xAI merger is the wildcard. xAI was valued at two hundred fifty billion on the basis of Grok and a one-hundred-thousand-GPU data center. Public market investors will want to see revenue attribution between the SpaceX business that generates cash and the xAI business that consumes it. Private investors agreed to a blended valuation. Public investors will demand a decomposition.
The same decomposition awaits OpenAI and Anthropic. Twenty-five billion in revenue sounds like validation until investors ask what the margins are. Anthropic's nineteen billion sounds like proof until the capital expenditure required to maintain frontier capability is subtracted. The private market rewarded revenue growth. The public market will interrogate profitability — or at least a credible path to it.
The Question
This journal has tracked the AI infrastructure buildout from the six-hundred-fifty-billion-dollar bet through the supply chain that followed, the physical constraints that pushed back, and the first genuine supply shock that sorted the cycle's winners from its telecoms. The Offering is the next chapter: the moment the builders must justify the build to an audience that can sell.
The question that resurfaces is the one this journal has carried since March: is this 1999 telecom or 1870s railroad? If AI unit economics — revenue per GPU-hour — improve before the capital expenditure cycle peaks, it is railroad. Real demand meeting real infrastructure. If unit economics deteriorate while spending accelerates, it is telecom. Supply creating its own narrative, demand following later or never.
The SpaceX filing is the first move. OpenAI's IPO will follow. Anthropic's will follow that. Within eighteen months, the three companies that absorbed the majority of private AI capital will be publicly traded, publicly scrutinized, and publicly priced by an adversarial market.
The infrastructure will persist regardless. It always does. The fiber that Global Crossing laid in bankruptcy became the backbone of Netflix and AWS. The track that Union Pacific built through the Panic of 1873 still carries freight.
The question has never been whether the infrastructure is real. The question is who owns it on the other side.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)