Breaking News — Anthropic Files S-1 Confidentially with the SEC
On June 1, 2026 (Eastern Time), Anthropic officially confirmed that it had confidentially submitted a draft registration statement (Form S-1) to the SEC in preparation for an IPO of common stock. In a statement, the company said, "This gives us the option to go public after the SEC completes its review. An actual IPO will depend on market conditions and other factors," emphasizing that this is a process to secure an option, not a commitment to listing. Major outlets including CNBC, Bloomberg, CNN, and Axios all reported the news simultaneously.
The "confidential submission" referred to here is a system established under the U.S. JOBS Act that allows emerging growth companies to work with the SEC privately before disclosing financial information to the public, giving them the opportunity to finalize the prospectus. As Axios notes, companies can receive regulatory feedback before disclosing information to investors, allowing the review process to proceed without exposing sensitive figures to the market. The public version of the S-1 is typically released approximately 15 days before the roadshow (the tour of presentations to institutional investors) begins, with listing and pricing following roughly one month later. In other words, this "confidential submission" is the starting gun of a countdown toward what could be the IPO of the century — but with pricing and share count still undetermined, it is best understood as precisely that: a starting signal.
The weight of this news lies in its timing. Elon Musk's SpaceX filed its confidential application with the SEC in April, has since submitted its public prospectus, and is on the verge of a roadshow, with a mid-June listing being discussed. OpenAI was also reported to have confidentially filed its S-1 in late May (CNBC reported on May 20 that a filing could come "as early as Friday"), with a September listing said to be the target. Anthropic's move means it has now entered an unprecedented race — one in which, sandwiched between these two companies, "companies each valued at roughly $1 trillion are heading to public markets in the same calendar year for the first time in history."
How Each Paper Covered It — The Starting Gun of the "3 Trillion Dollar IPO Race"
The tone of initial reports varies subtly in emphasis depending on the outlet. Bloomberg headlined it as "Confidential Filing Amid Surging Claude Demand, in Competition with OpenAI," framing it in the context of product demand and the race against OpenAI. CNN positioned it as "one of the most anticipated market debuts in years," while CNBC built its story methodically around the company's official statements and rapid financial growth. Axios took a more macro view of the race, highlighting that "for the first time in history, three companies valued at over $1 trillion are heading to public markets simultaneously." Reuters-affiliated reporting (distributed by MarketScreener and others) focused on investor sentiment, noting that "investors are making a major bet on the future of AI," while crypto media and some aggregator outlets grouped SpaceX, OpenAI, and Anthropic together under the umbrella term "the $3 trillion AI IPO race."
What is notable is that every outlet placed the "valuation reversal" at the center of the narrative. TechCrunch reported that Anthropic was "closing in on $1 trillion ahead of IPO," treating the fact that Anthropic reached a $965 billion valuation in its Series H — surpassing OpenAI's $852 billion valuation from late March — as an event that reshapes the entire landscape of the race. A company that, just months earlier, had been seen as the runner-up chasing OpenAI had now taken the lead on both fundraising and IPO preparation. It is precisely this narrative of reversal that elevated the initial reports from routine procedural news to "the IPO of the century."
The Reality of Quarterly Profitability――Quarterly Revenue Exceeding 1 Trillion Yen and Persistent Skepticism
The biggest reason this IPO filing is capturing market attention is that Anthropic is on the verge of achieving "profitability" — a milestone that has long remained out of reach for major generative AI players. According to multiple reports (CNBC), the company recorded $4.8 billion in revenue in Q1, with Q2 projected to reach $10.9 billion; if achieved, that would represent approximately $559 million in operating profit — the company's first quarterly operating profit since its founding. The annualized revenue run rate had swelled to $47 billion as of May, a dramatic leap from roughly $10 billion for full-year 2025 just months earlier. Figures that had been reported in stages — $14 billion at the start of the year, $30 billion in April — have now been revised even higher.
Improvements in unit economics further support expectations of profitability. Per CNBC's reporting, gross margins are around 40%, and while the company spent roughly $0.71 of every dollar of revenue on compute in Q1, that ratio is expected to fall to approximately $0.56 in Q2. The explosive revenue growth is being driven largely by coding demand, including Claude Code, a developer assistance tool that runs in the terminal.
At the same time, growing warnings caution against taking these profitability figures at face value. First, these numbers are not audited final results but projections prepared for fundraising purposes, and Anthropic itself has noted that it may not be able to sustain full-year profitability as data center investment accelerates in the second half of the year. Second, prominent tech industry critic Ed Zitron raised sharper doubts in a piece titled "The Lie of Anthropic's Profitability," arguing that: the company may be receiving a temporary discount on SpaceX-linked compute (the so-called Colossus, discussed below) specifically in Q2 — the quarter projected to show a profit — artificially compressing costs; the "operating profit" may be presented using non-GAAP metrics closer to EBITDA rather than GAAP accounting standards; and prepaid token purchases may be recognized as immediate revenue. He invoked comparisons to WeWork as a warning. These are Zitron's inferences and not facts acknowledged by the company, but they represent the market's prevailing caution that "the substance of this profitability claim should await scrutiny of the public S-1." Given the volatility in reporting, what investors should examine is not the flashy headline figures but the audited financial statements filed with the SEC.
Strengths Against Competitors — A Dual Moat in Enterprise and Coding
The rationale behind Silicon Valley VCs pouring massive capital into Anthropic lies not in consumer-facing brand recognition, but in its dominant share of the enterprise market. According to a report compiled by major VC firm Menlo Ventures in December 2025, Anthropic holds 40% of enterprise LLM spending, surpassing OpenAI's 27%. In the developer-focused coding model market in particular, Claude commands an estimated 54% share, far ahead of OpenAI's 21%. This represents the largest single-vendor gap across all LLM use cases, demonstrating the depth of the company's moat.
This strength is backed by concrete products. "Claude Code," made generally available in May 2025, reached $1 billion in annualized revenue by November of that year — at a faster pace than ChatGPT — and doubled that figure to $2.5 billion by February 2026. In a February 2026 survey by Pragmatic Engineer of 15,000 developers, Claude Code garnered 46% support as the "most loved tool," with 71% of developers who use AI agents daily naming Claude Code as their primary tool. Corporate customers exceeded 300,000 as of October 2025, and enterprise contracts for Claude Code reportedly quadrupled entering 2026.
Technological and brand differentiation is also noteworthy. Since its founding, Anthropic has championed "Constitutional AI" — a method of training models based on a codified set of principles rather than human feedback — building a reputation as the company most committed to safety and responsible AI. In January 2026, it published a revised "Claude's Constitution," explaining that models should understand not just "what to do" but "why they should behave that way," enabling generalization to novel situations. Its top-tier model "Claude Opus 4.7," released in April of that year, is considered the company's strongest in software engineering, instruction following, and practical task execution, and features safety mechanisms that automatically detect and block high-risk cyber use cases. For large enterprises on edge about regulatory and litigation risk, this "cautious and predictable behavior" has become a decisive factor in adoption.
Furthermore, the supply-side moat of secured computational resources is substantial. The company receives massive compute supply from hyperscalers such as Amazon and Google, and its Series H is reported to have come with approximately $15 billion in hyperscaler commitments, including $5 billion from Amazon. Ironically, SpaceX — which competes with Anthropic for an IPO spotlight — is also a supplier of compute resources to Anthropic, making the relationship between the two companies impossible to reduce to simple rivalry (detailed further in the next section).
SpaceX, OpenAI, Anthropic — Is There Enough Capital in the Market for Simultaneous IPOs?
This is the central question of this article: when three companies valued at around $1 trillion each attempt to absorb capital from the public markets in rapid succession during the same second half of 2026, does the market have enough money to absorb it?
The scale of demand is unprecedented. SpaceX is targeting a raise of $40–80 billion (with some analyses citing up to $75 billion) at a valuation of $1.75–2 trillion — which alone would be roughly three times the all-time IPO record set by Saudi Aramco in 2019 (approximately $26 billion). OpenAI and Anthropic are each expected to raise tens of billions of dollars as well, bringing the combined new equity supply from all three companies to an estimated $135 billion or more by one estimate, and over $200 billion by another that frames it more broadly as "total demand on public markets." Figures vary across reports depending on how the scope is defined, but in any case the magnitude is in a different league. For context, the entire U.S. IPO market raised just $45 billion over all of 2025. Goldman Sachs projects that IPO proceeds in 2026 will reach approximately $160 billion — four times the prior year — and yet even that figure may fall short of the combined demand from these three companies alone.
Even so, the major Wall Street banks are broadly optimistic. JPMorgan, Goldman Sachs, and Morgan Stanley view the liquidity impact of these mega-IPOs on the market as manageable. One basis for this view is the roughly $8 trillion sitting in U.S. money market funds — so-called "dry powder" — against which SpaceX's $75 billion raise represents just about 1%. As long as global investors continue to allocate capital to the AI theme, the IPO window is seen as well-supported by ample liquidity.
Skepticism persists, however. A prominent strategist at Bank of America warned that the listings of all three companies could push the technology sector's share of the S&P 500 past a historical threshold of 48%, exceeding the concentration levels seen at the peak of past bubbles. Valuations themselves are a source of serious doubt: one analysis characterized SpaceX's multiple of roughly 95–107 times 2025 revenues and OpenAI's multiple of over 75 times as "difficult to justify within a conventional return-on-earnings framework alone." To help bridge this supply-demand gap, Elon Musk has announced a plan — unprecedented in IPO history — to allocate up to 30% of SpaceX shares to retail investors, itself evidence that even the parties involved recognize institutional investors alone cannot absorb the offering. The staggered timing of the three listings — June (SpaceX), September (OpenAI), and October (Anthropic) — also reflects a pragmatic effort to avoid market indigestion. Ultimately, whether all three companies can complete their listings on schedule and at their intended valuations will itself serve as a grand stress test of whether the public markets can swallow trillion-dollar-valued AI infrastructure stocks.
Silicon Valley VC Perspective: From "VC Investment" to "Equity Investment"
What follows is the intrinsic perspective of Silicon Valley VCs rarely articulated in other news outlets. The current IPO race is simultaneously a story about AI companies and a reflection of an inflection point for the venture capital industry itself.
The most emblematic move is that of Sequoia Capital. As revealed by Financial Times reporting in January 2026, the firm has invested in both OpenAI and xAI while also backing competitor Anthropic — breaking the decades-old unwritten VC rule of "one horse per race." In the words of those close to the deal, "this is a case where the company has grown so large that it has become an equity investment rather than a venture investment." Indeed, Sequoia reportedly relinquished its information rights in OpenAI in order to lock in its position in Anthropic, making a judgment call that prioritized "exposure (ownership) over information" — the exact opposite of traditional VC operating principles. One commentary called this "the death of the non-compete," noting that AI majors are no longer being treated as early-stage software startups but as "sovereign-grade infrastructure" more akin to utilities or aerospace companies. Training next-generation models requires over $100 billion (approximately ¥15.5 trillion) in compute resources per cluster, and with capital requirements having shifted by an order of magnitude, the traditional VC framework for evaluating investments simply can no longer contain them.
This structural shift is reflected in the numbers. U.S. VC investment in Q1 2026 reached a record $297 billion (approximately ¥46 trillion), with 81% concentrated in AI. Sequoia itself established a new $7 billion (approximately ¥1.1 trillion) fund in April to expand its AI investments. Brad Gerstner, CEO of Altimeter Capital — which led the Series H round — stated: "Claude's latest evolution has driven large-scale adoption among the world's most demanding organizations. This momentum positions Anthropic as the leader of next-generation AI." He makes no attempt to conceal that these massive sums are being deployed not as early-stage bets but as "pre-IPO bridges" into established oligopolies.
From a VC perspective, Anthropic's web of intertwined relationships is particularly revealing. SpaceX is simultaneously a competitor in the race to IPO and a supplier of compute resources. In its S-1, SpaceX disclosed a compute contract with Anthropic totaling $45 billion (approximately ¥7.0 trillion) at $1.25 billion (approximately ¥194 billion) per month, supplying Anthropic with over 220,000 GPUs and more than 300 megawatts of capacity across Colossus I and II in Memphis. SpaceX itself has transformed into an AI conglomerate, having absorbed xAI in an all-stock transaction valuing the combined entity at $1.25 trillion (approximately ¥194 trillion) in February 2026 (with xAI formally integrated as the "SpaceXAI" division in May). In other words, the compute cost discounts underpinning Anthropic's path to profitability (the very point questioned by Citron), the listing value of SpaceX, and the rationale for Sequoia betting on both — all of it circulates within the same computational resource ecosystem. The real bet for VCs has shifted from "which single company will win" to "how broadly can we position ourselves across the tectonic shift that is AI infrastructure itself." This is the capital logic of Silicon Valley flowing beneath the surface of the three simultaneous IPOs.
Future Focus — Public S-1, Roadshow, and Fall Listing
So when and what kinds of new developments can we expect to observe going forward? The confidential filing is merely a starting point, from which several verifiable milestones will follow in sequence.
The most immediate thing to watch is the SEC review process. A confidential S-1 review typically takes 60 to 90 days and involves multiple rounds of comments. Once complete, the public S-1 will be disclosed approximately 15 days before the roadshow begins — this moment is the single most important inflection point, where the audited financial statements, risk factors, and the true substance of profitability that this article has repeatedly qualified will finally be laid bare. According to various reports, Anthropic has retained Wilson Sonsini Goodrich & Rosati (the prestigious firm that handled the IPOs of Google and LinkedIn) as legal counsel, with Goldman Sachs, JPMorgan, and Morgan Stanley named as underwriters. In terms of a realistic timeline, the prevailing view is that the public S-1 will come in July or August, the institutional investor roadshow led by CEO Dario Amodei and others will follow in September, pricing and listing will occur in October, and NASDAQ is the likely exchange. However, Anthropic itself has not committed to any specific dates, and everything depends on market conditions.
Equally worth tracking is the dynamics of the race as a whole. If SpaceX, which is leading the pack, successfully lists in mid-June with a strong debut, it will provide a tailwind for the two that follow — OpenAI (targeting September) and Anthropic (targeting October). Conversely, a stumble could pour cold water on the schedules and valuations of all three companies. In that sense, SpaceX's pricing and debut in June will serve as the first litmus test for Anthropic's IPO. The variables that investors and VCs should be measuring over the coming months are clear: the length of SEC comment rounds, the actual gross margins and quality of profitability shown in the public S-1, the structure of the lockup period, the allocation ratio to retail investors, and whether the cash sitting idle in money market funds actually flows into the new equity supply. Only when all of these pieces are in place will it become clear whether the "IPO of the century" becomes a historic success story — or a symbol marking the peak of the AI bubble. Anthropic's confidential filing is, without question, a meaningful first step in that long process of verification.