Model

Anthropic cuts Fable 5.2 prices 20% on IPO roadshow eve, and Google and Broadcom shareholders carry the bill

Anthropic discounted its flagship the day its IPO quiet period ended, for buyers who had already stopped climbing. The compute bill behind the discount is owed chiefly to Google and chipmaker Broadcom, which leaves their shareholders carrying the exposure.

Vincent JiangVincent Jiangthree-minute read
Share
Dario Amodei, Anthropic's chief executive, speaking at the White House luncheon on super intelligence
1 / 6Slide 1 of 6
Anthropic chief executive Dario Amodei at President Trump's luncheon on super intelligence at the White House, September 29, 2026.

Anthropic$350B — Anthropic, private, latest valuation $350B put its newest frontier models on the table on Monday, the same day the quiet period around its IPO ran out, the stretch after a filing in which US rules bar a company preparing to list from promoting its own shares 1 2 3. Fable 5.2, the general-release flagship, handles harder software engineering, robotics control and long-running projects, and can coordinate more than 1,000 AI agents at once 1.

Mythos 5.2, the same model with fewer safeguards for cybersecurity and life sciences work, reaches a widening set of vetted organizations; Washington restricted both lines in June and let access back only after Anthropic added blocking classifiers 1. Inside the launch is a price cut: Fable 5.2 lists at $8 per million input tokens and $40 per million output, down 20% from Fable 5.1's $10 and $50 1.

Buyers had already stopped climbing

The cut lands against a demand problem the data had already marked. Spending data released in August by Ramp, the corporate card firm, showed Anthropic's priciest model drawing just 11% of what businesses spend on its models, a level Ramp called "a new upper bound for how much businesses are willing to spend" 1.

The bill behind the discount runs the other way. Anthropic's amended IPO filing shows third-quarter revenue of nearly $17 billion against $9.9 billion of computing expense, which turned a $974 million second-quarter operating profit into a $567 million operating loss 4.

Fable 5.2 cuts the price 20% on both ends of the meter

The roadshow's flagship product arrived 20% cheaper on both ends

  • Fable 5.1
  • Fable 5.2
$0/M$10/M$20/M$30/M$40/M$50/MInput tokens$8/M$10/M-20%Output tokens$40/M$50/M-20%
Data
Fable 5.1Fable 5.2Change
Input tokens$10/M$8/M-20.0%
Output tokens$50/M$40/M-20.0%
List API price per million tokens, input and output, as announced November 9, 2026. Source: Bloomberg.1

Pacing is not like pausing

In mid-September, chief executive Dario Amodei publicly asked top AI firms to slow their most advanced work, days after an Anthropic researcher quit saying the people building AI believe it could kill all humans by the end of the decade 1. No new Fable or Mythos had shipped since early September, until Monday 1.

Dianne Penn, who leads research product management, bridged the gap in four words: "Pacing is not like pausing" 1. A Democratic Congress expected to push AI safety rules now has the reversal on the record 1.

Whose money absorbs the cut

Formal marketing opens on Tuesday, with the listing reportedly targeted for the week of November 16 on Nasdaq under the ticker ANT, and no price range on file as of Monday 3 5. What is on file is $895.4 billion of committed spend, substantially compute and infrastructure 4.

Google is owed at least $121.3 billion, and chipmaker Broadcom+4.96% — Broadcom, up 4.96 percent today holds $161.2 billion of equipment leases, the largest single arrangement in the filing; the leases supply TPU-based accelerator hardware for data centers run on Anthropic's behalf 4 6. Broadcom's chief executive expects Anthropic to become "our largest XPU customer in 2027"; XPU is shorthand for accelerator chips like those 4 6. Holders of AVGO and GOOGL must read the 20% cut as demand softness or a volume play; no document states which 4.

Broadcom and Google hold the two largest claims on Anthropic's $895B spend

The two largest claims on Anthropic's compute bill are Broadcom and Google

$0B$50B$100B$150B$200BBroadcom equipment leases$161.2BLargest single arrangement in the filingGoogle$121.3BSpaceX AI unit$120.5BAmazon AWS$110BMicrosoft Azure$44.9BNscale$44.6BLambda$41.3B
Data
Value
Broadcom equipment leases$161.2B
Google$121.3B
SpaceX AI unit$120.5B
Amazon AWS$110B
Microsoft Azure$44.9B
Nscale$44.6B
Lambda$41.3B
Committed compute and infrastructure spend by counterparty, from the November 2 S-1 amendment; AWS and Azure are contractual minimums, the SpaceX figure a maximum. Sources: 24/7 Wall St; GeekWire.4,6

Cisco signs on the same day

Cisco−0.15% — Cisco, down 0.15 percent today said the same day that it is adding Claude Mythos to its AI Defense product so security teams can red-team, meaning attack-test, corporate AI agents 7 8. Engagements run only against registered targets, under a named operator, with no open-ended prompts 7. The tell is the IPO price range, still blank in the filing 5. Until it prints, IPO buyers are pricing a flagship that just got 20% cheaper 1.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Loading AVGO

Reader comments

0 comments

    Model

    Anthropic hired a UK lobby firm the week it priced its $1.4 trillion IPO

    The same week Anthropic set its $70-to-78 IPO range, the UK lobbyist registry showed it paying for political access in three countries. Goldman's $685 billion revenue case runs on data centers that voters and regulators must permit, and buyers fund that insurance without seeing its cost.

    Model

    Mistral built its enterprise stack in eight days and priced none of it

    Between October 29 and November 5 the Paris lab published its biggest model's weights, expanded its agent platform and bought a fifth startup, all funded by a September mark of €21 billion. Not one piece of the stack carries a published price, while free Chinese open-weight models still outscore it.

    Model

    xAI's Grok 4.8 entered training 57 days late with no price card; Anthropic pays SpaceX $1.25 billion a month

    The 57-day wait bought a serving stack xAI wrote itself, and the delay ends with no scores and no price. For a SpaceX shareholder, the unpublished price card for 4.8 is the next receipt.

    Model

    Meta's Muse Ember costs $0.14 a task, a third of Claude Haiku, as Anthropic prices its IPO

    Meta's smallest model does the work agents repeat all day at $0.14 a task, matching Anthropic's Haiku at a third of the cost. Anthropic prices its IPO this week on about $81 billion of run-rate token revenue, and the floor under that business just moved.

    Model

    AfterQuery's agent training data lifted OpenAI's small model from 3.1% to 17% on Terminal-Bench

    The November 4 run, showcased on Thinking Machines' Tinker, sits beside rival Mercor's proof on the same page, giving labs a direct comparison of measured benchmark uplift for the first time. It is the claim beneath AfterQuery's reported $3.2 billion valuation, now competing in public for labs' data budgets.

    Model

    Google DeepMind's Kohli promised a phase transition; the receipts are an RSA-1024 result and Antigravity posts

    DeepMind's chief scientist for Google Cloud promised a glimpse of a "phase transition" in scientific discovery within weeks. The numbers behind it are already on the table: RSA-1024 cracks, an hour-long drug-design loop, and a blog cadence that funnels straight into Google Antigravity.

    Get the newsletter

    The day's AI and market news, in your inbox every weekday.