Anthropic is closing a $30 billion funding round at a $900 billion pre-money valuation, with Dragoneer, Greenoaks, Sequoia, and Altimeter as co-leads — and three of those four firms are also significant investors in OpenAI. [1] The Financial Times, Cryptopolitan, and the Economic Times confirmed the round and the co-lead list independently over the weekend; Bloomberg reported the closing window as "as soon as this month." [2] [3] The lifetime raise climbs to roughly $72 billion across 18 rounds, with $60 billion of that arriving in the last two. [2] Anthropic's revenue projection, per the same FT account, is to top $45 billion annualised "shortly." [3]

The paper argued on Saturday that the lab had closed the round past OpenAI's $852 billion valuation without a word on Magnifica Humanitas, the encyclical Pope Leo XIV will publish Monday with Anthropic co-founder Christopher Olah on the Vatican dais. Sunday's increment is the cap-table read. Three of the four lead investors — Sequoia, Altimeter, and Dragoneer — also hold positions in OpenAI. The fourth, Greenoaks, has not been publicly named as an OpenAI investor but has co-invested in earlier Anthropic rounds.

The implication is structural. The frontier-AI venture-capital industry has stopped picking sides. The same partners are writing the same checks to both labs, often inside the same calendar quarter, with the same diligence teams running parallel models on competing companies whose products are designed to displace each other in the same enterprise sales motion. The choice the public markets have not yet been asked to make — Anthropic or OpenAI, Claude or GPT, Olah's interpretability research or Altman's product flywheel — has been resolved in private at the cap-table layer by the simple expedient of refusing to choose. Each side gets the same firms. Each firm gets exposure to whichever side wins. The portfolio construction is the bet.

This is not novel in venture history. The late-2010s autonomous-vehicle wave saw the same convergence — every leading firm ended up in Cruise and Waymo and Argo and Aurora and Zoox — until the technology timeline lengthened past most funds' exit windows and the capital quietly receded. The 1990s search-engine wave saw the same thing with Yahoo, Excite, Lycos, Infoseek. What distinguishes the 2026 AI version is the magnitude. Anthropic at $900 billion plus OpenAI at $852 billion represents $1.75 trillion of private valuation across two companies whose combined revenue is plausibly $80 billion this year and which are funded in significant part by the same partners writing checks to both.

The structural risk inside the convergence is not concentration; it is correlation. When the same firms back both leaders at the same valuations on similar terms, a downturn in either prices the other immediately. A regulatory action against Anthropic — say, an FTC settlement that constrains enterprise compute resale — repriced through Sequoia's portfolio is the same arithmetic that reprices through Sequoia's OpenAI stake. The diversification the firms have built by being on both cap tables disappears the moment the macro frame against AI compute turns. They are diversified inside the trade and undiversified against the trade.

The closing window also coincides with the lab's encyclical silence. The paper's Saturday position held that Anthropic had produced no public statement on Magnifica Humanitas, on the SpaceX-Anthropic $45 billion compute disclosure from Friday, or on the round's closing. Vice President JD Vance, the first Catholic VP since Biden, told White House reporters on May 19 that the encyclical "is going to have some influence" and that the administration is "pro-innovation" on AI. The administration spoke five days before publication. The lab has not spoken at all. The Stainless acquisition on May 18 — an enterprise developer-tools tuck-in that closes a structural rate-limit on Claude adoption — landed inside the same silence and was buried inside the round's news cycle.

What the cap-table convergence forecloses is the press-and-policy framing that has shaped the public understanding of the AI race since 2022: that OpenAI and Anthropic are operationally and culturally opposed, that the safety-versus-capability split is also an investor split, that the choice between Altman's and Amodei's worldviews is a choice the market is making in real time. The Sunday read is that the market is not making that choice. The market is buying both, at the same valuations, from the same partners, in the same week. The choice is being deferred until the public markets — through an IPO, a tender, or a downturn — are forced to mark one of the books to the other's price.

That moment is closer than the timeline suggests. The FT reporting on the round includes the line that Anthropic is preparing for a public-markets entry "later in 2026." [3] If that timeline holds, the same partners now writing co-lead checks at $900 billion will be selling secondary into the IPO book at whatever price the public market discovers. OpenAI's own IPO timeline — variously reported as 2026 or 2027 — would create the same forcing function in reverse. The cap-table convergence works only as long as the marks stay private. Once one of the two clears, the other gets repriced against it within the same trading week.

Anthropic itself has not commented on the round, on the encyclical, or on the SpaceX disclosure. The closing window remains "as soon as this month." The Pope speaks Monday. The lab's silence on its own week is, at this point, the only public position the lab is taking.

-- THEO KAPLAN, San Francisco