AI/Nov 20 edition/Reporting & analysis
BusinessSafetyPolicy

BusinessMarkets, money & strategy

Anthropic raises $81.3 billion in record IPO as founders keep majority voting control

Anthropic sold about 980 million shares at $83 in the largest IPO on record, but a founder-held share class keeps 50.1% of votes. Critics question whether its Long-Term Benefit Trust can meaningfully check founders or slow risky model releases.

From the The Verge · AI article of November 19, 2026 and cited sources: The offering priced at $83, above the expected $70–78 range, and raised $81.3 billion, ahead of SpaceX's $75 billion. [2] [3] [8]

SOURCES[2] CNBC: Anthropic IPO raises over $81 billion cnbc.com[3] Fortune: Anthropic's dual-class share structure fortune.com[8] Yahoo Finance: Anthropic Raises $81.3 Billion in Record IPO finance.yahoo.com

AI-assisted, not human-edited. Methodology Corrections Who owns this site

KEY FACTS
What happened
The offering priced at $83, above the expected $70–78 range, and raised $81.3 billion, ahead of SpaceX's $75 billion. [2] [3] [8]
Source caveat
On the eve of trading, a CAIS-led statement signed by 92 people, including two Nobel laureates, called Anthropic not a safe AI company. [6] [7]
Original source
The Verge · AI article, independent perspective
Brief by
AT Pulse research desk AI-assisted; publisher and contact on the desk page
When
November 19, 2026 article published
Evidence
10 sources across 10 sites

Reach: not measured yet

STK202_DARIO_AMODEI_CVIRGINIA_D
Image: The Verge — Original article ↗
THE CORE IDEAS4 TAKEAWAYS
01

The offering priced at $83, above the expected $70–78 range, and raised $81.3 billion, ahead of SpaceX's $75 billion. Private placements of about $26 billion bring the total to more than $107 billion. Public Class A shares are capped at 49.9% of the vote, while founder-held Class F stock carries 50.1%, so outside shareholders cannot outvote the founders. [2] [3] [8]

02

Anthropic presents its Long-Term Benefit Trust as the check on founder power, but its authority appears narrow. According to The Verge's reading of the S-1, the trust can appoint and remove directors but cannot delay or halt a model release. Governance experts and a former employee doubt it has meaningful power. A former trustee's move into an Anthropic executive role also raises questions about trustee independence. [1] [3] [9]

03

In February, Responsible Scaling Policy v3.0 replaced the company's commitment to pause training with a conditional delay, citing competitors that apply weaker safeguards. Oversight now depends on risk reports every three to six months, a non-binding safety roadmap, and external review when reports are heavily redacted. GovAI identifies the reliance on self-reporting as a remaining weakness, while arguing that an honest policy is better than commitments the company would not keep. [4] [5] [10]

04

On the eve of trading, a CAIS-led statement signed by 92 people, including two Nobel laureates, called Anthropic not a safe AI company. The signatories accuse it of using safety branding to build power and of lobbying against SB 1047, the RAISE Act and the AEGIS Act. The reviewed reporting does not independently corroborate the lobbying claims. [6] [7]

WHY IT MATTERS

Under the reported structure, public investors get economic exposure without control. Safety oversight relies on the company's own risk reports and on a trust that has limited say over releases.

Read the full assessment

Enterprise buyers and policymakers may need to judge Anthropic by those reports, not by the old pause pledge.

Sources are verified, so I'm writing up the dossier now.

Executive brief

Anthropic's IPO raised $81.3 billion at $83 a share, the largest IPO on record, ahead of SpaceX's $75 billion. Even so, outside shareholders can't take control of the company. Class A public shares are capped at 49.9% of the vote, and a single founder-held share carries 50.1% (CNBC; Fortune). Anthropic presents its Long-Term Benefit Trust as the check on the founders, but The Verge reports that governance experts doubt it has real power. On the eve of trading, 92 people signed a statement calling Anthropic "not a safe AI company." Signatories included two Nobel laureates.

What changed and event timeline

  1. RSP v3.0 drops the pause pledge

    Anthropic replaced its commitment to halt training with a conditional "delay," citing competitors that don't apply the same safeguards (;).

  2. Multi-class share structure detailed

    The share classes are Class A (public), Class F (founders, 50.1% of votes), Class T (the trust, which elects directors), Class E (employees) and Class N (strategic partners, no votes) ().

  3. "MisAnthropic Statement" published

    The Center for AI Safety (CAIS) and a broad coalition accused Anthropic of using safety branding to build power and of lobbying against SB 1047, the RAISE Act and the AEGIS Act (;).

  4. IPO prices above range

    About 980 million shares sold at $83, above the expected $70–78 range. The deal implies a fully diluted value of about $1.5 trillion ().

  5. Trading begins as ANT on Nasdaq

    About $26 billion in private placements brings the total raised in the transaction above $107 billion ().

Capabilities and access

  • The IPO doesn't come with a new model release. The company's main products are Claude and Claude Code (CNBC).
  • Revenue for the nine months to Sept. 30 was $33.24 billion.
  • Google, Microsoft, Nvidia and AMD are investing $26 billion together at the IPO price.
Read the full section
  • The IPO doesn't come with a new model release. The company's main products are Claude and Claude Code (CNBC).
  • Revenue for the nine months to Sept. 30 was $33.24 billion. The net loss was $104 billion, mostly from changes in the fair value of convertible notes. The annualized revenue run rate reached about $81 billion by the end of October (CNBC).
  • Google, Microsoft, Nvidia and AMD are investing $26 billion together at the IPO price.

Technical analysis for researchers and developers

No model architecture or benchmark is involved.

Read the full section

No model architecture or benchmark is involved. The technical substance is in the safety-governance machinery:

  • RSP v3.0 replaced a fixed "stop at threshold" rule with three mechanisms: risk reports every 3–6 months that cover internally deployed models, a non-binding Frontier Safety Roadmap, and external review when reports are heavily redacted (GovAI).
  • Implication: Outsiders can now check Anthropic's safety commitments only through the company's own reports. GovAI names this reliance on self-reporting as a remaining weakness.
  • Board oversight: According to The Verge's reading of the S-1, the trust can appoint and remove directors but can't delay or stop a model release.

Claims and evidence

  • How many directors the trust controls is disputed. The reviewed sources don't settle the difference.
  • The RSP was weakened. Reporting confirms this (Semafor).
  • The lobbying allegations come from the statement's authors. No independent reporting in the reviewed sources corroborates them.
Read the full section
  • How many directors the trust controls is disputed. The Verge says the trust picks four of seven directors, and one of the four effectively needs founder approval. Fortune says Class T elects "at least 6 of 13" directors, and the S-1 describes Class T directors as a board majority. The reviewed sources don't settle the difference.
  • The RSP was weakened. Reporting confirms this (Semafor). The pause became conditional on leaders judging that Anthropic is ahead and that catastrophic risk is significant (NYU Shanghai RITS).
  • The lobbying allegations come from the statement's authors. No independent reporting in the reviewed sources corroborates them.
  • Anthropic didn't comment to The Verge.

Context and prior work

  • Anthropic introduced the Long-Term Benefit Trust in 2023 (The Verge).
  • Experts compare it to OpenAI's nonprofit board before Sam Altman's ouster.
  • The structure resembles founder-controlled companies like Meta and SpaceX (Fortune).
Read the full section
  • Anthropic introduced the Long-Term Benefit Trust in 2023 (The Verge).
  • Experts compare it to OpenAI's nonprofit board before Sam Altman's ouster. That board did have authority over operational and safety decisions, including the power to delay a release; Anthropic's trust doesn't.
  • The structure resembles founder-controlled companies like Meta and SpaceX (Fortune).
  • Class F's founder control ends if fewer than three of the seven co-founders remain in the founders' LLC for 90 days. Class T has no expiration date but can be amended or redeemed.

Limitations, safety and contested findings

  • Law professor Ann Lipton questions whether trustees have the AI expertise to judge when to intervene.
  • The Verge reports several safety staff left in 2026 with public warnings. One former employee said the trust has no meaningful power.
  • Dario Amodei has said self-improving AI must be pursued "very carefully, if at all."
Read the full section
  • Expertise and independence. Law professor Ann Lipton questions whether trustees have the AI expertise to judge when to intervene. The trustees are Ben Bernanke, Richard Fontaine, Christopher Stubbs and Neil Buddy Shah (chair). Former trustee Mariano-Florentino Cuéllar joined Anthropic as an executive in August, which shows a trustee can be hired away (The Verge).
  • Staff departures. The Verge reports several safety staff left in 2026 with public warnings. One former employee said the trust has no meaningful power.
  • Recursive self-improvement. Dario Amodei has said self-improving AI must be pursued "very carefully, if at all." The statement's signatories say Anthropic is racing to build it (CAIS).
  • The other view. GovAI concluded that an honest RSP beats commitments that wouldn't be honored in practice.

Business and practitioner implications

  • Investors get economic exposure but no control. Public shareholders can't outvote the founders.
  • Enterprise buyers who rely on Anthropic's safety branding should track the risk reports and external reviews rather than the pause pledge, which no longer exists.
  • Compute is the binding constraint. Amodei called compute "the deepest business dilemma" the company faces.
Read the full section
  • Investors get economic exposure but no control. Public shareholders can't outvote the founders.
  • Enterprise buyers who rely on Anthropic's safety branding should track the risk reports and external reviews rather than the pause pledge, which no longer exists.
  • Compute is the binding constraint. Amodei called compute "the deepest business dilemma" the company faces. Lipton argues the need for money will be the real check on Anthropic's priorities (The Verge).
  • Policy watchers should expect the lobbying allegations to come under scrutiny now that Anthropic is public.
FOLLOW THE EVIDENCE

The source trail.

Sources (10)
A LITTLE LESS NOISE. A LOT MORE CONTEXT.

Stay curious.
Follow the evidence.

Independent perspectives, the original sources, and room for the questions that don't have easy answers.

How we build the brief
Connect with us

Find us where you already read.