# Anthropic's S-1 and OpenAI's Foundation: two trillion-dollar bets, one week

URL: https://www.thedeepfeed.ai/posts/2026-06-05-anthropic-s1-openai-foundation-two-bets/
Category: Business
Published: 2026-06-05
Author: the-deep-feed
Tags: anthropic, openai, ipo, ai-funding, valuations, agi
Kind: deep

> On June 1, 2026, Anthropic filed confidentially to go public at a $965B valuation while OpenAI's nonprofit Foundation published its case for spending $25B on 'AI resilience.' Two leading labs, same day, opposite bets: one privatizes the upside, the other pre-commits capital to the downside.

## TL;DR

- On **June 1, 2026**, two issuer pages went up the same day: Anthropic announced a **confidential S-1 filing** with the SEC, and OpenAI's nonprofit Foundation published Wojciech Zaremba's case for a **$25 billion 'AI resilience'** bet. One lab filed to sell shares; the other's nonprofit made its case for spending against catastrophe.
- Anthropic's run to the door was fast and verified: a **$65B Series H at $965B post-money** (May 28, co-led by Altimeter, Dragoneer, Greenoaks, Sequoia), its **first profitable quarter** on **$10.9B Q2 revenue**, and the first frontier lab to file. It eclipsed **OpenAI's $852B** standing mark.
- The contradiction is the story. Anthropic, [in one investor's words](https://x.com/faraz0x/status/2061765526561402956), 'refused the Pentagon, locked away its most dangerous model, and bet everything on safety' — and **is about to answer to shareholders**.
- Prediction markets price the race precisely: Polymarket gives Anthropic **73%** to IPO first and **92%** to hold the higher valuation on June 30. The $1.25T-by-December market sits at **88% yes**.
- The two bets are not opposites by accident. They are two answers to the same question — what do you do with the capital AGI throws off before AGI arrives? Anthropic returns it to shareholders. OpenAI's Foundation pre-commits a slice of it to the scenario where the technology goes wrong.

On June 1, 2026, two pages went up on two corporate websites, dated the same day, that together describe the entire shape of the AI industry's relationship to money. On one, **Anthropic** announced it had confidentially submitted a draft registration statement to the Securities and Exchange Commission — the paperwork that begins a public offering. On the other, OpenAI's nonprofit Foundation published an essay by co-founder Wojciech Zaremba making the case for spending $25 billion on what it calls "AI resilience," the work of preparing for the ways advanced AI could go wrong.

![A split diptych divided by a red line: a coin into a rising stack on the left, a coin into a sealed vault on the right — upside versus downside](/post-images/2026-06-05-anthropic-s1-openai-foundation-two-bets/two-bets-hero.jpg)

Read together, the two pages are a split-screen of the same moment — the financial mirror of the [land-grab playing out one layer down](/posts/2026-06-03-agent-governance-land-grab/), where the same labs are buying the control plane for how agents get deployed. One lab is racing to convert the value of the AGI bet into public-market liquidity and shareholder upside. The other's nonprofit parent is publicly pre-committing capital to the downside scenario. Both companies believe they are building the most consequential technology in history. They have arrived at opposite conclusions about what to do with the money it is already generating.

This is not a story about which lab is right. It is a story about what the two filings reveal when you put them next to each other — and the numbers, cross-checked against the issuers' own pages, are stranger and sharper than the timeline's "Anthropic beat OpenAI to Wall Street" framing captured.

# The 96-hour window

The events that matter happened in a tight cluster, and the dates are worth stating precisely because the simultaneity is the point.

On May 28, Anthropic [closed a $65 billion Series H at a $965 billion post-money valuation](https://www.anthropic.com/news/series-h), co-led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital — the four lead investors named on Anthropic's own newsroom page. Bloomberg's framing was that the round [eclipsed OpenAI](https://www.bloomberg.com/news/articles/2026-05-28/anthropic-raises-at-965-billion-valuation-eclipsing-openai), whose standing mark from its March raise was $852 billion. For the first time, the smaller, safety-focused lab was worth more on paper than the company that defined the category.

Four days later, on June 1, Anthropic [confidentially filed its draft S-1](https://www.cnbc.com/2026/06/01/anthropic-ipo-s1-prospectus.html), becoming the first frontier AI lab to begin the public-offering process. CNBC confirmed the filing and noted that OpenAI "is readying its own confidential filing" — the rival was close behind but, on the day, behind.

That same June 1, the OpenAI Foundation published [Zaremba's "Resilience in the age of AI"](https://www.openaifoundation.org/news/resilience-in-the-age-of-ai), framing the Foundation's $25 billion commitment to AI-resilience work. The pairing was not invented by commentators after the fact. Alex Heath, who [profiled Zaremba "stepping out"](https://sources.news/p/openais-quiet-co-founder-steps-out) to make the public case, bundled both events into a single newsletter — his subhead read, in effect, that Zaremba was making the resilience case while "Anthropic readies for IPO." The two stories were one story to the reporter closest to them.

| Date | Anthropic | OpenAI |
|---|---|---|
| May 28 | Closes $65B Series H at $965B post-money | (standing mark: $852B from March) |
| Jun 1 | Confidential S-1 filed with SEC | Foundation publishes $25B "AI resilience" case |
| Jun (ongoing) | Reportedly targeting a fall debut | Readying its own confidential filing |

One company spent the window filing to sell shares. The other's nonprofit spent it explaining why it will spend billions on the possibility that the product fails catastrophically. Same four days, same industry, opposite postures.

# By the numbers, cross-checked

The figures driving this story circulate on financial X in a slightly inflated form, and the gap between the verified numbers and the repeated ones is itself instructive.

The verified core is remarkable enough. Anthropic's Series H was **$65 billion at a $965 billion post-money valuation** — both confirmed on the issuer's page, both corroborated by Reuters, Bloomberg, and TechCrunch. CNBC reported that Anthropic was on track for **$10.9 billion in Q2 revenue**, its [first profitable quarter](https://www.cnbc.com/2026/05/20/anthropic-revenue-explosive-growth-ipo-profitable-quarter.html), up from $4.8 billion in Q1. A lab that was at a roughly $61.5 billion valuation thirteen months earlier had, by [PitchBook's accounting](https://pitchbook.com/news/articles/anthropic-bests-openai-in-valuation-race-hitting-965b-with-series-h), bested OpenAI in the valuation race.

Then there is the number that is everywhere and nowhere: the **$47 billion revenue run-rate**. It appears in nearly every viral post about the filing. It does not appear on any Anthropic page. It is a reported, social-media-propagated figure, and it should not be confused with the verified $10.9 billion quarterly number, which annualizes to roughly $43.6 billion at that quarterly pace. The distinction matters for a publication that takes its numbers seriously: $10.9 billion for Q2 is confirmed; $47 billion in run-rate is reported. The viral version rounds up and smooths over, and the precise version is the one worth printing.

The same discipline applies to OpenAI's side. The **$25 billion AI-resilience bet** is verified on the [Foundation's own page](https://www.openaifoundation.org/news/resilience-in-the-age-of-ai), and traces to the [March 24 recapitalization update](https://www.openai.com/index/update-on-the-openai-foundation/) that pledged at least $1 billion in grants over twelve months across four program areas. OpenAI's reported September debut at a $1 trillion-plus valuation, by contrast, is not issuer-confirmed — CNBC confirms only that OpenAI is "readying its own confidential filing," not the date or the price. The trillion-dollar target is a market expectation, not a company statement.

# The velocity that made the filing possible

The number that makes the rest of the story legible is not the valuation. It is the speed at which Anthropic reached it. The lab set a roughly $61.5 billion mark in early 2025 and a $380 billion mark in February 2026. By late May 2026 it closed at $965 billion — a 154% step-up from the February mark in about 105 days. ![Four steeply rising bars climbing left to right with the tallest final bar in red: a valuation up 154% in about 105 days](/post-images/2026-06-05-anthropic-s1-openai-foundation-two-bets/valuation-climb.jpg)

The revenue curve underneath it is steeper still. Anthropic's reported annual revenue run-rate went from about $1 billion in early 2025 to roughly $30 billion by April 2026, and one widely-circulated post marked the moment the smaller lab passed the larger one on the top line:

> Anthropic just officially surpassed OpenAI in revenue for the first time. January 2025: Anthropic ARR = $1 billion. April 2026: Anthropic ARR = $30 billion. $1 billion to $30 billion in 15 months.
>
> — [@NoLimitGains](https://x.com/NoLimitGains/status/2049213286311436597), Apr 28, 2026

That trajectory is what makes a confidential S-1 a rational next step rather than a vanity one. A company growing revenue thirty-fold in fifteen months and posting its first profitable quarter has the financials a public-market underwriter wants to see. The valuation followed the revenue, and the filing followed the valuation.

There is a contrarian read of the raise worth surfacing, because several analysts reached for it independently. The $65 billion is not really valuation theater, the argument goes; it is a compute bet. Anthropic's own page describes the proceeds as funding research and expanded Claude capacity, and outlets framed the Series H as ["really a compute bet"](https://implicator.ai/anthropic-beats-openai-to-a-confidential-ipo-filing-at-965-billion/), capital raised primarily to secure the GPUs and data-center capacity that frontier training and inference now demand. On that reading, the IPO is less about cashing out and more about access to the deepest possible capital pool to keep buying compute. The valuation is the headline; the compute is the use of proceeds.

# The contradiction at the center

Anthropic was founded by people who left OpenAI over how fast and how commercially the frontier should be pushed. Its public identity is built on caution: it positions itself as the lab that takes catastrophic risk most seriously, that publishes the most safety research, that holds capabilities back. The single most quotable line in the entire discourse around the filing put the resulting tension exactly:

> The AI era just went to Wall Street. Anthropic filed for an IPO today. Confidential S-1 submitted to the SEC. The company that refused the Pentagon, locked away its most dangerous model, and bet everything on safety is about to answer to shareholders.
>
> — [@faraz0x](https://x.com/faraz0x/status/2061765526561402956), Jun 2, 2026

![A seesaw balancing a red shield of safety and restraint against a group of figures pulling for growth: the tension a public listing institutionalizes](/post-images/2026-06-05-anthropic-s1-openai-foundation-two-bets/safety-vs-shareholders.jpg)

"About to answer to shareholders" is the crux. A public company has a fiduciary obligation to its investors. The safety-first lab is about to acquire a quarterly-earnings cadence, an activist-investor surface, and a stock price that will react to every decision to slow down, hold back, or refuse a contract on principle. The same posture that built Anthropic's brand, declining lucrative work and gating its most capable models, is the posture that public-market discipline most aggressively punishes. The filing does not resolve that contradiction. It institutionalizes it.

OpenAI's move is the mirror image, and it is at least as strange. The company that has been criticized for moving fast and commercializing aggressively is the one whose nonprofit parent is publicly committing $25 billion to the proposition that the technology might go badly wrong and that society needs to be made resilient to it. Zaremba's framing of [resilience as "the work of our generation"](https://www.openaifoundation.org/news/resilience-in-the-age-of-ai) is the language of an organization hedging its own product. One reading is genuine mission. Another is reputational insurance bought with a slice of the upside. The honest answer is probably that it is both, and that the two are not separable.

The cleanest way to hold the two bets together is this: both labs are sitting on capital that the AGI thesis is generating right now, before AGI itself has arrived. Anthropic's answer is to formalize that capital into public equity and return its growth to shareholders. OpenAI's answer, through its Foundation, is to ring-fence a portion of it against the scenario where the whole thesis produces something dangerous. Privatize the upside, or pre-fund the downside. Those are the two available moves when you believe you are building something world-altering and profitable at the same time, and the two leading labs each picked one.

# What the prediction markets see

The most precise read on the race does not come from a press release. It comes from people betting real money, and Polymarket priced the contest with unusual granularity in the days around the filing.

On the headline question of which lab goes public first, Polymarket gave **Anthropic 73% against OpenAI's 27%**. On which company holds the higher valuation at the end of June, the market was even more lopsided: **Anthropic 92%, OpenAI 8%**. The longer-dated markets show how far the expectations run. A market on whether Anthropic's valuation hits $1.25 trillion by December 31 sat at **88% yes**. The market on Anthropic's IPO closing market cap assigned near-zero probability to any band below $100 billion — the crowd does not merely expect the company to go public, it expects it to debut very large.

These are not idle numbers. They are the market's verdict that Anthropic's first-mover filing was a real lead, not a symbolic one, and that the valuation gap over OpenAI is expected to persist rather than snap back. (One widely-shared post cited 81% on the IPO-first market; the live figure on June 3 was 73% — prediction-market odds move, and the lower, current number is the honest one to quote.)

Notably, the prediction markets have a line on the IPO race and no line at all on the resilience bet. There is no market on whether OpenAI's $25 billion will be well spent, or on whether AI-resilience work will matter. The thing that is legible to bettors is the horse race to Wall Street. The thing that is not is the question Zaremba's essay is actually about. That asymmetry, where the upside is bettable and the downside is not, is its own quiet comment on which of the two bets the market knows how to price.

# Who actually owns the trillion-dollar bet

There is a structural fact that complicates both filings and that most of the coverage skipped: neither lab is independent of its largest backers, and those backers are public companies whose shareholders already hold the exposure. The point was made cleanly in a widely-shared explainer of how retail investors can get a position before either IPO:

> Neither Anthropic nor OpenAI is public yet. Retail investors can get exposure via these public companies with major stakes: Microsoft (~27% of OpenAI plus partnership); Amazon (a major Anthropic investor plus cloud deals, and an OpenAI stake); Alphabet (a large committed investor in Anthropic); Nvidia (heavy OpenAI investor and chips for both).
>
> — [@grok](https://x.com/grok/status/2061959986717118854), Jun 2, 2026

(Those specific stake figures are indicative and should be checked against each company's filings before being treated as exact, but the structure is real.) The consequence is that the "Anthropic versus OpenAI" race is partly a proxy for "Amazon and Alphabet versus Microsoft and Nvidia." Anthropic's primary infrastructure and capital partners are Amazon and Google; OpenAI's is Microsoft, with Nvidia supplying both. When Anthropic's valuation eclipses OpenAI's, it is also a scoreboard for which hyperscaler placed the better bet — and an IPO turns that private scoreboard into a public, daily-marked one.

This is why the compute-bet reading and the IPO are not in tension. A lab that raises $65 billion mostly to buy compute, from investors who are also its cloud providers, and then goes public, is building a flywheel: public capital funds compute purchases that flow back to the hyperscaler-investors who helped raise the capital. The IPO is the mechanism that opens that flywheel to the broadest possible pool of money. Going public is not the end of the compute bet. It is how the compute bet scales.

# Two answers to one question

![A road forking at a red signpost: one path to an open-market gateway, one to a protective dome](/post-images/2026-06-05-anthropic-s1-openai-foundation-two-bets/fork-in-road.jpg)

It would be easy to write this as a morality tale: the safety lab sells out, the commercial lab finds religion. That reading is too clean and probably wrong. The more useful frame is that both labs are responding rationally to the same unprecedented situation, and the responses only look contradictory because the situation is.

Anthropic's filing is the act of a company that has discovered it can be both safety-focused and extraordinarily profitable, and has concluded that the way to fund the former at scale is to formalize the latter. A $10.9 billion profitable quarter changes what is possible; it lets a lab argue that public-market capital is the fuel for more safety research, not a betrayal of it. Whether the quarterly-earnings treadmill actually permits that is the open question the S-1 does not answer, and the one worth watching after the debut.

OpenAI's Foundation bet is the act of an organization that has internalized the criticism that it moves too fast, and has decided to put nonprofit capital behind the hedge. Twenty-five billion dollars toward AI resilience is real money and a real signal, regardless of whether it is mission or insurance. It is also, conveniently, the kind of commitment that a company readying its own IPO would want on the record.

The two pages dated June 1 are best read not as a contest but as a disclosure. They reveal that the people closest to the AGI bet have looked at the capital it is throwing off and split, cleanly, on what to do with it. One camp says return it to owners and let the market allocate. The other says set a piece aside for the day the technology proves more dangerous than profitable. Both camps are staffed by people who believe they are building the same thing. That they cannot agree on what to do with the proceeds is the most honest thing either filing tells you about how uncertain the destination still is.

## Sources

- [Anthropic — raises $65B Series H at $965B post-money valuation](https://www.anthropic.com/news/series-h)
- [CNBC — Anthropic confidentially files IPO prospectus with SEC](https://www.cnbc.com/2026/06/01/anthropic-ipo-s1-prospectus.html)
- [CNBC — Anthropic revenue growth, $10.9B Q2, first profitable quarter](https://www.cnbc.com/2026/05/20/anthropic-revenue-explosive-growth-ipo-profitable-quarter.html)
- [Bloomberg — Anthropic raises at $965B valuation, eclipsing OpenAI](https://www.bloomberg.com/news/articles/2026-05-28/anthropic-raises-at-965-billion-valuation-eclipsing-openai)
- [Reuters — Anthropic raises $65 billion, now valued at $965 billion](https://www.reuters.com/business/anthropic-raises-65-billion-now-valued-965-billion-2026-05-28/)
- [TechCrunch — Anthropic raises $65 billion, nears $1T valuation ahead of IPO](https://techcrunch.com/2026/05/28/anthropic-raises-65-billion-nears-1t-valuation-ahead-of-ipo/)
- [OpenAI Foundation — Resilience in the age of AI (Wojciech Zaremba)](https://www.openaifoundation.org/news/resilience-in-the-age-of-ai)
- [OpenAI Foundation — update on the OpenAI Foundation (Bret Taylor)](https://www.openai.com/index/update-on-the-openai-foundation/)
- [Sources (Alex Heath) — OpenAI's quiet co-founder steps out](https://sources.news/p/openais-quiet-co-founder-steps-out)
- [PitchBook — Anthropic bests OpenAI in valuation race at $965B](https://pitchbook.com/news/articles/anthropic-bests-openai-in-valuation-race-hitting-965b-with-series-h)
- [Implicator.ai — Anthropic beats OpenAI to a confidential IPO filing](https://implicator.ai/anthropic-beats-openai-to-a-confidential-ipo-filing-at-965-billion/)

---

Canonical: https://www.thedeepfeed.ai/posts/2026-06-05-anthropic-s1-openai-foundation-two-bets/
Site: https://www.thedeepfeed.ai
Full corpus: https://www.thedeepfeed.ai/llms-full.txt