Investing/Nov 14 edition/Reporting & analysis
StocksMarketsRegulation

StocksEquities, earnings & valuations

Jensen Huang resigns as Nvidia CEO as METR audit finds Anthropic's AI progress has nearly doubled in pace

Jensen Huang resigned as Nvidia CEO after a second affair was reported, and CFO Colette Kress became interim CEO. Days earlier, an 11-week METR audit estimated that Anthropic's model progress has nearly doubled in pace, mostly because AI is accelerating Anthropic's own research.

From the Abnormal Returns article of November 13, 2026 and cited sources: Huang resigned on November 12. [1] [4] [5] [7]

SOURCES[1] Abnormal Returns: Friday links: lying about lying abnormalreturns.com[4] TechCrunch: Nvidia's Jensen Huang resigns techcrunch.com[5] CNBC: Nvidia CEO Jensen Huang resigns, live updates cnbc.com

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

KEY FACTS
What happened
Huang resigned on November 12. [1] [4] [5] [7]
Original source
Abnormal Returns article, independent perspective
Brief by
AT Pulse research desk AI-assisted; publisher and contact on the desk page
When
November 13, 2026 article published
Evidence
11 sources across 11 sites
39/100RisingHow it’s measured

YouTube video · 15.1K views · 308 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.

THE CORE IDEAS4 TAKEAWAYS
01

Huang resigned on November 12. The night before, the board had moved to remove him after a reported affair with an Nvidia vice president, which came on top of an earlier affair linked to a suspected foreign intelligence effort. Kress is interim CEO and remains CFO, and Harvey C. Jones became board chair. Shares fell nearly 12% early in the day and closed down 4.3% at $160. Commentators argued that the decisive problem was Huang telling the board the first affair was his only one. [1] [4] [5] [7]

02

METR estimated that Anthropic's progress is running about 1.9 times its historical rate (95% CI 1.5–2.8). It attributed roughly 80% of that speedup to AI doing Anthropic's own R&D and found convincing evidence that more training compute was not the cause. On Anthropic's internal capabilities index, yearly gains rose from about 15 points to about 29. [3] [8]

03

Forecasters shortened their timelines in response. The AI Futures Project now expects a superhuman AI researcher in 2027 and puts the odds of top-expert-dominating AI arriving before 2028 at 35–65%. Daniel Kokotajlo moved his median forward by 19 months. The Decoder reports that Anthropic may have crossed the acceleration warning threshold in its own Responsible Scaling Policy around mid-July. [6] [8] [9]

04

The DOJ, the Commerce Department, Congress and outside counsel for Nvidia's board are all investigating, and none has published findings. A Commerce rule blocking Chinese remote access to frontier chips has been reported, but no text has been released. [5] [11]

WHY IT MATTERS

METR found the speedup was not driven by more training compute, and leading forecasters moved their timelines forward.

Read the full assessment

Implication: capability gains could outpace chip buildouts, while Nvidia faces leadership and regulatory uncertainty with no formal succession plan.

Executive brief

Nvidia's founder lost his job this week. For AI practitioners, though, the bigger signal came a week earlier. On November 5, METR published an independent audit estimating that Anthropic's model progress is running about 1.9× faster than its historical rate (95% CI 1.5–2.8×), with roughly 80% of that speedup coming from AI doing Anthropic's own research (METR). Anthropic redacted most of the evidence METR rated as most informative. Then on November 12, Jensen Huang resigned as Nvidia CEO after an affair scandal tied to a suspected foreign intelligence effort. CFO Colette Kress is interim CEO (TechCrunch). The Abnormal Returns roundup links coverage of both stories.

What changed and event timeline

  1. METR publishes its Anthropic audit

    Over 11 weeks (Aug 12–Oct 29), METR estimated that Anthropic's progress had accelerated 1.9× since October 2025. It attributed about 80% of that to AI ().

  2. Huang apologizes

    Huang apologized for being "unfaithful" to his wife. This followed reporting that he had been the target of a foreign intelligence effort ().

  3. Forecasters shorten their AI timelines

    The AI Futures Project now expects a superhuman AI researcher in 2027. It puts the odds of top-expert-dominating AI arriving before 2028 at 35–65% and names METR's report as the main reason ().

  4. A second affair is reported

    The New York Times reported an alleged affair with Nvidia VP Kimberly Powell, and the board moved that night to remove Huang ().

  5. Huang resigns

    Kress became interim CEO while staying CFO. Harvey C. Jones became board chair (). NVDA closed at $160, down 4.3% ().

  6. Commentary turns to the board

    M.G. Siegler argued that the decisive problem was Huang telling the board the first affair was his only one, which he called "lying about lying" (;).

Capabilities and access

  • METR worked on devices supplied by Anthropic.
  • The models in the analysis run from Claude 3 Opus through Opus 4.5 and 4.6, Mythos Preview, Haiku 5.5 and three unreleased internal models.
  • Outsiders cannot access the redacted pipeline-stage estimates, bottleneck analyses or forecasts.
Read the full section
  • METR worked on devices supplied by Anthropic. It used Anthropic's internal capabilities index (AECI), about 15 staff interviews, self-reported productivity data and pipeline metrics (METR).
  • The models in the analysis run from Claude 3 Opus through Opus 4.5 and 4.6, Mythos Preview, Haiku 5.5 and three unreleased internal models. No new product launched.
  • Outsiders cannot access the redacted pipeline-stage estimates, bottleneck analyses or forecasts.

Technical analysis for researchers and developers

  • Method: METR fit a piecewise linear regression with one breakpoint to AECI scores, with trend data going back to October 2023.
  • Attribution: METR found convincing evidence that the speedup did not come from scaling training compute. The 80% AI share carries substantial uncertainty.
  • Self-reported uplift: Staff estimates rose from about 1.2× in January 2025 to about 12× in October 2026.
Read the full section
  • Method: METR fit a piecewise linear regression with one breakpoint to AECI scores, with trend data going back to October 2023. Gains rose from about 15 to about 29 AECI points per year (The Decoder).
  • Attribution: METR found convincing evidence that the speedup did not come from scaling training compute. The 80% AI share carries substantial uncertainty.
  • Self-reported uplift: Staff estimates rose from about 1.2× in January 2025 to about 12× in October 2026. Self-reports are not the same as measured output.
  • Reproducibility: It is low. The underlying data are internal and mostly redacted, and The Decoder reports that one graph is labelled "FAKE DATA" because of the redactions.

Claims and evidence

  • METR reports 1.9× (CI 1.5–2.8×). The AI Futures post cites "about 1.7×" and a plausible range of 1.3–3× (AI Futures).
  • Anthropic may have crossed its own warning threshold. METR's report itself does not assess RSP thresholds.
  • The size of Nvidia's drop depends on the time of day. TechCrunch reported an early fall of nearly 12%.
Read the full section
  • The headline multiplier differs by source. METR reports 1.9× (CI 1.5–2.8×). The AI Futures post cites "about 1.7×" and a plausible range of 1.3–3× (AI Futures).
  • Anthropic may have crossed its own warning threshold. The Decoder says the doubling trigger in Anthropic's Responsible Scaling Policy may have been crossed around July 15, and that internal figures showed 1.97× while the public August report said less than 2× (The Decoder). METR's report itself does not assess RSP thresholds.
  • The size of Nvidia's drop depends on the time of day. TechCrunch reported an early fall of nearly 12%. The podcast hosts said about 13–14% intraday (video 15:48). CNBC reports a 4.3% loss at the close.

Context and prior work

  • In the spring, METR researchers had put the chance of a near-term intelligence explosion below 10%, according to Kevin Roose (video 25:00).
  • Kokotajlo moved his median for a superhuman AI researcher forward by 19 months (FourWeekMBA).
  • A Commerce Department rule to block Chinese remote access to frontier chips has been reported, but there is no published text.
Read the full section
  • In the spring, METR researchers had put the chance of a near-term intelligence explosion below 10%, according to Kevin Roose (video 25:00).
  • Kokotajlo moved his median for a superhuman AI researcher forward by 19 months (FourWeekMBA). Earlier versions of the AI 2027 timeline models drew a detailed public critique (titotal).
  • A Commerce Department rule to block Chinese remote access to frontier chips has been reported, but there is no published text. The House passed the Remote Access Security Act 369–22 in January (Introl).

Limitations, safety and contested findings

  • Possible overestimate: METR says hiring or organizational changes could explain part of the speedup, and that Anthropic's 2023–25 baseline growth rate was unusually low.
  • Possible underestimate: METR also says recent acceleration may be higher and that AECI may understate real-world impact.
  • Incomplete audit: According to The Decoder, Anthropic cancelled the compute-allocation check partway through. METR still rates acute catastrophic risk as low.
Read the full section
  • Possible overestimate: METR says hiring or organizational changes could explain part of the speedup, and that Anthropic's 2023–25 baseline growth rate was unusually low.
  • Possible underestimate: METR also says recent acceleration may be higher and that AECI may understate real-world impact.
  • Incomplete audit: According to The Decoder, Anthropic cancelled the compute-allocation check partway through. METR still rates acute catastrophic risk as low.
  • Host disclosure: Casey Newton disclosed that his fiancé works at Anthropic (video 26:44). The hosts backed Kelsey Piper's call for Congress to see the unredacted report (video 36:31).
  • Nvidia investigations: The DOJ, the Commerce Department (looking into Megaspeed), Wachtell Lipton for the board, and Congress are all investigating (CNBC). No findings have been published.

Business and practitioner implications

  • Key-person risk: Nvidia had no formal succession plan, and Huang managed about 60 direct reports (TechCrunch).
  • Neoclouds and open models: The hosts argue that neoclouds face less Nvidia financing under Kress and that open-model advocates have lost a champion (video 12:36; 7:33).
  • Planning around the METR result: If AI is speeding up AI research, capability gains may arrive faster than new compute alone would predict.
Read the full section
  • Key-person risk: Nvidia had no formal succession plan, and Huang managed about 60 direct reports (TechCrunch).
  • Neoclouds and open models: The hosts argue that neoclouds face less Nvidia financing under Kress and that open-model advocates have lost a champion (video 12:36; 7:33). Both points are commentary, not reported fact.
  • Planning around the METR result: If AI is speeding up AI research, capability gains may arrive faster than new compute alone would predict. Because the evidence is redacted, treat the 1.9× figure as a wide range when planning.
FOLLOW THE EVIDENCE

The source trail.

Sources (11)
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.