VC & Startups/Nov 21 edition/Reporting & analysis
ExitsValuationsM&AFunding

ExitsIPOs, listings & liquidity

Anthropic ends its Nasdaq debut up 17% at nearly $1.8 trillion while Nvidia and AMD shares struggle

Anthropic closed its first trading day at $97.17, up from its $83 IPO price. The same week, Nvidia's CEO resigned without warning and a short-seller attack hit AMD, so the record listing did not lift the broader AI chip trade.

From the Newcomer article of November 20, 2026 and cited sources: Anthropic listed on Nasdaq on Nov. 19 at $83 a share. [3] [8]

SOURCES[3] CNN: Anthropic IPO cnn.com[8] Bloomberg: Anthropic debut bloomberg.com

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

KEY FACTS
What happened
Anthropic listed on Nasdaq on Nov. 19 at $83 a share. [3] [8]
Original source
Newcomer article, independent perspective
Brief by
AT Pulse research desk AI-assisted; publisher and contact on the desk page
When
November 20, 2026 article published
Evidence
16 sources across 14 sites
35/100RisingHow it’s measured

YouTube video · 9K views · 90 likes

Reach blends public engagement (views, likes, comments, shares, stars) from the platforms where the story or its original source lives: YouTube, Hacker News, GitHub, Substack and similar. A video and the article written from its transcript share one score, taken from the video. Platforms with no data are left out rather than counted as zero, and the age of a story never lowers its score.

Illustration from Newcomer: Anthropic ends its Nasdaq debut up 17% at nearly $1.8 trillion while Nvidia and AMD shares struggle
Image: Newcomer — Original article ↗
THE CORE IDEAS4 TAKEAWAYS
01

Anthropic listed on Nasdaq on Nov. 19 at $83 a share. It opened at $111.50, reached $115.10 and closed at $97.17, giving it a value close to $1.8 trillion. It was the largest IPO on record. [3] [8]

02

On Nov. 12, Nvidia's board said Jensen Huang was resigning effective immediately and named CFO Colette Kress interim CEO. Five days later Nvidia reported revenue of $115.8 billion against $109.1 billion expected and raised its fiscal 2028 growth guidance to 75%. The stock was still down 24% for the month. [4] [5] [11]

03

AMD shares fell almost 10% after Hunterbrook reports accused the company of circular customer financing and of letting firmware reach a blacklisted Chinese firm. AMD says the claims misrepresent its business. Newcomer compares chipmakers investing in their own customers to the early-2000s telecom bubble. Citing Crunchbase, it says Nvidia and NVentures have made a record 121 investments in 2026. [1] [9] [16]

04

Palo Alto Networks agreed to buy Applied Compute in an all-stock deal valued at $4.5 billion. Applied Compute's platform post-trains and serves open-weight models. The price is about 1.4 times the $3.25 billion valuation reported for its last private round, a $350 million raise led by Elad Gil. [7] [15]

WHY IT MATTERS

Filings and exchange data back the IPO pricing and Nvidia's results.

Read the full assessment

The implication is inference: if chipmakers cut back on investing in their own customers, AI startups would lose a major funding source, and late-stage valuations could face tougher scrutiny.

Sources are verified; writing the dossier now.

Executive brief

Anthropic listed on Nasdaq on Nov. 19, 2026 and closed its first day at $97.17, up 17% from its $83 IPO price and worth close to $1.8 trillion (CNN). The listing did not lift the rest of the AI trade. Nvidia lost its founder-CEO, a short-seller report knocked almost 10% off AMD, and a Wall Street pledge of about $500 billion in data-center loans has not paid out anything, according to the Wall Street Journal (Newcomer). In the same week, Google's Gemini 4 Neon placed third on an independent intelligence index. That weakens the argument that DeepMind had dropped out of the frontier race.

What changed and event timeline

  1. Huang out at Nvidia

    The board announced Jensen Huang's resignation, effective immediately. CFO Colette Kress became interim CEO and Harvey C. Jones became the company's first board chair ().

  2. Earnings beat, stock lags

    Revenue was $115.8B against $109.1B expected, and fiscal 2028 growth guidance rose to 75%. Shares were still down 24% for the month ().

  3. Gemini 4 Neon ships

    This is the middle model of the Gemini 4 family. It costs $3/$15 per million input/output tokens and has a 1M-token context window ().

  4. Palo Alto buys Applied Compute

    The all-stock deal is valued at $4.5B. Applied Compute builds a platform for post-training and serving open-weight models ().

  5. Anthropic debuts

    Shares opened at $111.50, reached $115.10, then closed at $97.17. It was the largest IPO on record (;).

  6. Hunterbrook targets AMD

    Two short-seller reports accused AMD of circular customer financing and of firmware reaching a blacklisted Chinese firm. AMD says the claims misrepresent its business. Shares fell almost 10% (;).

Capabilities and access

  • Gemini 4 Neon is available in the Gemini API (AI Studio), Gemini for Business, Antigravity (paid tiers), the Gemini app for Pro/Ultra subscribers, and the Fairwind program for cybersecurity defenders. Google has published a model card.
  • Applied Compute sells post-training and inference for open-weight models. Palo Alto is positioning it for secure deployment on public and sovereign clouds (PANW).
Read the full section
  • Gemini 4 Neon is available in the Gemini API (AI Studio), Gemini for Business, Antigravity (paid tiers), the Gemini app for Pro/Ultra subscribers, and the Fairwind program for cybersecurity defenders. Cached input costs $0.15 per million tokens (Google). Google has published a model card.
  • Applied Compute sells post-training and inference for open-weight models. Palo Alto is positioning it for secure deployment on public and sovereign clouds (PANW).

Technical analysis for researchers and developers

  • Google's announcement does not describe Neon's architecture.
  • Artificial Analysis scores Neon (xhigh) at 61 on its Intelligence Index.
  • Neon produces about 15% more output tokens per task than Carbon, but generates them at about 110 tokens/s versus 57.
Read the full section
  • Google's announcement does not describe Neon's architecture.
  • Artificial Analysis scores Neon (xhigh) at 61 on its Intelligence Index. That puts it behind GPT-6.2 Astra (max) and Claude Fable 5.2 (max), and one point ahead of Gemini 4 Carbon (Artificial Analysis).
  • Neon produces about 15% more output tokens per task than Carbon, but generates them at about 110 tokens/s versus 57. Tasks finish roughly 40% faster (Artificial Analysis).
  • The cost comparison uses Carbon's discounted rates, which end Nov. 30. Recompute per-task cost after that date before relying on it.
  • In the podcast, Gary Marcus argues that recent capability gains come from neuro-symbolic tooling rather than from pure LLMs. As evidence he cites the leaked Claude Code and Grok-4's announcement data (19:00).

Claims and evidence

  • Nvidia's financials and leadership change appear in SEC filings, and CFO Dive reports an SEC disclosure probe. Newcomer says Kress plans to scale back Huang's dealmaking.
  • Google's Neon results are mostly self-reported, for example 65.2% on Terminal-Bench 4.1 and 0.0% prompt-injection success on Gray Swan's benchmark.
  • Applied Compute deal figures:
Read the full section
  • Nvidia's financials and leadership change appear in SEC filings, and CFO Dive reports an SEC disclosure probe. Newcomer says Kress plans to scale back Huang's dealmaking. Earnings coverage reviewed by Fortune does not report that.
  • Google's Neon results are mostly self-reported, for example 65.2% on Terminal-Bench 4.1 and 0.0% prompt-injection success on Gray Swan's benchmark. The index ranking has independent support from Artificial Analysis and the Vals Index (#2 of 43, 70.87%).
  • Applied Compute deal figures:
  • The $4.5B price is company-announced.
  • The $350M round led by Elad Gil at a $3.25B valuation comes from BankInfoSecurity. It is not in a filing.
  • The $100M revenue run rate is self-reported.
  • Palo Alto's target of more than $1B in ARR by 2029 is a company projection.
  • Hunterbrook is a short-seller with a position against AMD. Its numbers come from that report and have not been independently confirmed: $161.1B in share promises, about $28B in guarantees, and $15.8B of cash (Nacionale).

Context and prior work

  • Newcomer compares chipmakers investing in their own customers to the telecom bubble of the early 2000s.
  • SemiAnalysis had declared that DeepMind was no longer a frontier lab, after Gemini 3.5 Pro was delayed.
  • Marcus says he has argued since August 2023 that labs building near-identical models face commodity economics with "no moat."
Read the full section
  • Newcomer compares chipmakers investing in their own customers to the telecom bubble of the early 2000s. Per Crunchbase, Nvidia and NVentures have made a record 121 investments so far in 2026 (Newcomer).
  • SemiAnalysis had declared that DeepMind was no longer a frontier lab, after Gemini 3.5 Pro was delayed.
  • Marcus says he has argued since August 2023 that labs building near-identical models face commodity economics with "no moat." He likens the business to airlines (09:52).

Limitations, safety and contested findings

  • Rebellion Research argues that Hunterbrook overstates its evidence against AMD.
  • Marcus's market claims are his own reading and have no corroboration in the reviewed sources:
  • In data-center referendums, he says 43 of 44 went against new data centers (02:12).
Read the full section
  • Rebellion Research argues that Hunterbrook overstates its evidence against AMD.
  • Marcus's market claims are his own reading and have no corroboration in the reviewed sources:
  • In data-center referendums, he says 43 of 44 went against new data centers (02:12).
  • He says OpenAI asked the UAE for $30B and got $11B (07:10).
  • He says Anthropic's Q3 growth fell to 47% quarter over quarter, which he calls an omission from its original S-1 (25:39).
  • On safety: Marcus calls for a narrow, temporary pause on agents with open-ended internet and credential access (34:15).

Business and practitioner implications

  • Startups: if Nvidia and AMD slow their customer investing, startups lose a major funding channel.
  • Model buyers: Neon offers near-frontier scores at mid-tier prices, which strengthens the case for routing between models. Re-benchmark costs after Nov. 30.
  • Security vendors: Palo Alto paid about 1.4× Applied Compute's reported last private valuation. That signals strong demand for secure fine-tuning of open-weight models.
Read the full section
  • Startups: if Nvidia and AMD slow their customer investing, startups lose a major funding channel.
  • Model buyers: Neon offers near-frontier scores at mid-tier prices, which strengthens the case for routing between models. Re-benchmark costs after Nov. 30.
  • Security vendors: Palo Alto paid about 1.4× Applied Compute's reported last private valuation. That signals strong demand for secure fine-tuning of open-weight models.
  • Coding: Marcus puts the productivity gain at about 30%. He calls that real but far below what was promised (11:51).
  • Neoclouds: credit markets are already pricing in risk. Marcus says CoreWeave is borrowing at about 15% (02:57).
FOLLOW THE EVIDENCE

The source trail.

Sources (16)
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.