AI IPO Wave: What's the deal with OpenAI's 'secret application'?

In the early hours of June 9th, OpenAI announced it has submitted a "secret" S-1 filing to the SEC, initiating its IPO process. The target valuation is up to $1 trillion. It could go public as early as September. Combined with SpaceX, which is set to list on June 12th (valuation $1.77 trillion), and Anthropic, which has already completed its confidential filing (private market valuation $965 billion) — within one month, the three AI giants are set to transition from the private market to the public market, with a combined $4.5 trillion about to be unveiled.

This scale is unprecedented in the history of U.S. stock market IPOs. During the 1999 internet stock rush, 475 companies went public throughout the entire year, and their combined scale wasn't as concentrated as these three today.

What is a "Secret Filing"?

The formal name is Confidential Filing, which can be loosely translated as "保密申报" in Chinese.

It's not a "secret"; the SEC is fully aware throughout the process, but it's kept confidential from the market and media. The company submits the S-1 document to the SEC for review, but this document is not made public. It only becomes officially public 21 days before the company decides to actually go public. This mechanism was established by the 2012 U.S. "JOBS Act" (Jumpstart Our Business Startups Act). The original intent was clear — to allow small and medium-sized tech companies to communicate privately with the SEC before an IPO, without having to disclose all financial details, business models, and risk factors to competitors from the start.

Initially, it was only open to "emerging growth companies" (annual revenue below $1 billion). In 2025, the SEC expanded its scope, and now large companies can also use it.

OpenAI benefits from confidential filing in three ways:

First, timing flexibility. Submitting does not mean actually going public. The company can proceed with the SEC review process while observing the market. If the timing seems unfavorable, it can withdraw at any time, and the market remains completely unaware.

Second, competitors cannot see. The S-1 contains a large amount of sensitive data — detailed revenue composition, cost structure, customer lists, sources of losses. With traditional public filing, this information is copied by competitors as soon as it's submitted. Confidential filing cuts out most of this time window.

Third, it avoids the awkwardness of a "post-filing decline". Between the submission of a traditional public S-1 and the actual listing, the company's stock price (private valuation) and market sentiment are exposed. If something unexpected happens in between, the company's image can be damaged. Confidential filing reduces this "exposure period" to 21 days.

In plain terms — this is about leaving a way out for oneself, not a commitment.

Why are all three crowding into this month?

SpaceX, OpenAI, and Anthropic did not coordinate this — they are fierce competitors. But their concentration in this month is driven by the same market signals.

Three signals:

First, valuations have peaked. SpaceX at $1.77 trillion, OpenAI at $1 trillion, Anthropic at $965 billion. This scale is already too large for the private market to handle; few funds can lead another round of $10+ billion financing alone. The only entity that can absorb this scale is the public market.

Second, the AI valuation environment is just right. In the first half of 2026, global capital flowed out of and then back into tech stocks, and the market is still buying into the AI narrative. This is a valuation window — waiting longer could mean encountering changes in the Fed's pace, geopolitical risks, or an AI cycle correction, making it impossible to lock in the valuation.

Third, mutual pressure. After SpaceX's June 12th listing, the public market's appetite for "giant AI companies" will be partially consumed. OpenAI and Anthropic must follow closely, not letting SpaceX alone absorb all the capital.

This is the reason for the accelerated pace of these three since late May. This is not corporate strategy; it's market rhythm forcing their hand.

The 1999 Internet Bubble

To understand this IPO wave, we must go back to 1999.

That year, 475 companies went public in the U.S. stock market, the vast majority being internet companies. The Nasdaq index rose from 750 in 1995 to 5048 in March 2000 — a more than fourfold increase in 5 years. The market was then filled with narratives like "new economy," "paradigm shift," and "zero marginal cost."

Then, in March 2000, the bubble burst. The Nasdaq fell 75% over the next two and a half years, wiping out $5 trillion in market value.

The most painful statistic came later: of the 473 companies that went public in 1999, only 66 were still operating by 2018. 86% of the companies had either gone bankrupt, been acquired, or delisted.

But note — there are several fundamental differences between the 1999 episode and the 2026 one.

A Few Real Differences

I can tell you a few real differences —

Difference one: Profitability. The vast majority of internet companies that IPO'd in 1999 did not have sustainable revenue models. Many companies had "annual losses of $100 million, valuation of $5 billion." In 2026, these three — SpaceX Q1 revenue $4.694 billion (but loss $4.276 billion), OpenAI Q1 revenue $5.7 billion, Anthropic annualized revenue crossing $47 billion. They have real revenue, just not yet profitable at scale.

Difference two: Technological substance. Many 1999 IPOs were about "putting a traditional business on a .com" — pets.com, webvan.com were like that. The 2026 trio are companies truly changing computing/communications/logistics infrastructure — this is infrastructure-level technology, not a marketing gimmick.

Difference three: Customer base. The customers of 1999 internet companies were mainly "consumers not yet accustomed to online shopping." Among the customers of the 2026 trio — SpaceX has NASA + the Pentagon + global telecoms, OpenAI and Anthropic have Fortune 500 companies. These are mature customers paying real money.

Difference four: The nature of the money. At the peak of the 1999 bubble, retail investors following the trend constituted a huge proportion of IPO subscriptions. The early investors in the 2026 trio are all top-tier institutions (Sequoia, Greenoaks, Founders Fund, etc.). Retail buying happens after the listing.

Back to OpenAI's confidential filing.

It sends two layers of signals to the market —

The surface layer is: OpenAI is preparing for an IPO, targeting a $1 trillion valuation.

The deeper layer: OpenAI choosing confidential filing is about leaving itself the option "not to go." If in the coming months SpaceX's listing performs poorly, AI valuations correct, or a macro black swan event occurs, it can quietly withdraw.

This "leaving room" posture itself is the most telling signal to read in the AI listing wave.

Not everyone believes this valuation can hold. Including OpenAI itself.