Anthropic’s IPO Filing Puts OpenAI on the Defensive

Alex Kantrowitz said Anthropic had confidentially submitted a draft Form S-1 registration statement to the U.S. Securities and Exchange Commission for a proposed initial public offering of common stock. As described in the source, the filing does not force an immediate listing. It starts the SEC process and gives Anthropic the ability to move when it chooses.

Kantrowitz also noted that Anthropic’s announcement was unusually sparse: two paragraphs on its blog, with no apparent pre-briefing of news publications. He called it a “mic drop.”

For MG Siegler, the timing mattered because OpenAI had reportedly been expected to file soon, possibly the prior Friday, though no public confirmation had emerged. Siegler kept the ambiguity intact: OpenAI may already have filed confidentially without that becoming known, or Anthropic may have “pulled a quick one” and moved ahead first.

That sequencing matters because a large AI IPO is not only a financing event; it is also a narrative contest. The first company to reach public investors can help establish the comparison set, frame the growth story, and put rivals in the position of explaining why their numbers deserve equal or better treatment.

If Anthropic reaches the public market before OpenAI, Siegler argued, OpenAI faces a “narrative nightmare.” Earlier, the public-market pitch for OpenAI could lean on ChatGPT’s consumer scale and top-line leadership, while explaining heavier spending and a longer path to profitability as the cost of pursuing a larger opportunity. Siegler said newer reports complicate that story: Anthropic is said to be growing quickly, may have reached some level of profitability, and now appears to have put itself in position to list first.

Kantrowitz compared the dynamic to book publishing, where similar books can lead one publisher to announce a publication date to clear the field while another keeps its plans quiet. He said he would not be surprised if OpenAI had also filed and was “trying to move in silence.” Siegler agreed that if OpenAI had filed first, it might now feel pressure to say so, if only to avoid the appearance that Anthropic beat it. His reaction compressed the competitive implication: “your move OpenAI.”

MG Siegler’s central concern was not simply who files first. It was that Anthropic and OpenAI are becoming more comparable at the same time that the reported comparisons between them have become less favorable to OpenAI.

SpaceX and OpenAI may also be compared by investors, especially because SpaceX now has xAI, but Siegler treated that comparison as imperfect. SpaceX has a rocket business and X; OpenAI does not. Anthropic and OpenAI, by contrast, are much more direct competitors. Their businesses are not identical, but Siegler said they are converging because OpenAI increasingly has to pursue the areas where Anthropic has found traction: developers, coding, and eventually agentic work.

That makes public-market positioning more difficult. Investors looking for comparable companies would be weighing two AI businesses against each other, not a frontier AI lab against a space company with an AI arm. Siegler said the most recent fundraising comparisons “increasingly do not look good for OpenAI,” especially if Anthropic has moved ahead on top-line growth while improving its bottom-line story.

OpenAI’s likely counter, Siegler said, would be consumer scale. It could point, in his telling, to 900 million monthly active users and argue that ChatGPT remains the far larger consumer product. But even that number carried a caveat: Siegler said it “has not shifted in a while,” and OpenAI would presumably prefer to announce that it had crossed one billion. He also said Sundar Pichai had announced Gemini at 900 million monthly active users at I/O, making OpenAI’s user-count narrative less clean than it once was.

The public-market question, then, becomes less about whether OpenAI is an important company and more about whether investors, comparing it directly with Anthropic, would still prefer its story. Siegler suggested OpenAI may need to “zoom back ahead” before listing, perhaps through Codex, if it wants to offset the narrative created by Anthropic’s reported growth and developer traction.

Alex Kantrowitz raised the financing question left open by the filing. He said Anthropic had just raised $65 billion and OpenAI had just raised $122 billion, then asked why either would need to go public. Was the burn so large that public markets were necessary, and what funding sources remained after that? The source does not specify the composition, timing, or terms of those raises, so the figures should be read as Kantrowitz’s stated framing rather than a breakdown of cash proceeds.

MG Siegler’s answer was partly structural and partly practical. A public company can have easier access to debt and can use its stock in a more liquid way. He did not claim to know every granular reason, but said the high-level reason may be simple: the IPO window is open, and companies that know they will need to keep raising capital may not want to wait until it closes.

Anthropic may look better positioned because of reported profitability, but Siegler said it also faces a capacity problem. OpenAI, he said, has played up the claim that Anthropic has not had enough capacity to meet demand. To fix that, Anthropic is striking infrastructure deals, including with SpaceX and other “neo-clouds.” Those deals are expensive. Siegler referred to a report that Anthropic could pay around $15 billion a year to lease SpaceX’s Colossus data center.

That kind of spending changes the meaning of a large private raise. A company can raise tens of billions and still need more if it immediately commits huge sums to compute capacity. Siegler did not predict exactly when Anthropic would list, noting that it was June and many variables could change before year-end. His broader point was that AI demand, data-center scarcity, and capital-market timing make the filing rational even for a company that recently raised a very large amount of money.

MG Siegler added one complication he described as controversial and still playing out: index inclusion. As SpaceX prepares to list, he said there had been changes, or efforts to change the rules, that could affect whether very large newly public companies need to wait before being added to major indices.

If companies such as SpaceX, OpenAI, or Anthropic were eligible for index inclusion sooner, investor-demand mechanics could change. Large mutual funds and other index-linked funds could have to rebalance and buy shares automatically. In that conditional scenario, Siegler said, traditional comparisons between OpenAI and Anthropic might matter less at the margin because index-driven demand could support the stocks regardless of the finer points of each company’s financial profile.

Siegler framed SpaceX as the near-term test case. He said it was expected to go public within days, possibly by around the 11th, and would test both public appetite and the index-inclusion dynamic. He described a possible world in which SpaceX’s market cap could move “way past two trillion” quickly because of automatic buying.

That possibility does not erase the strategic problem for OpenAI. Siegler presented it as a roundabout argument for why OpenAI might still go public around the same time as Anthropic, not as the core case. The core case remained sequencing: Anthropic’s filing gives it optionality, makes OpenAI’s timing more visible, and may force investors to compare the two companies on terms that are no longer clearly favorable to OpenAI.