Teahose.
SIGN IN
NEW HERE — WHAT TEAHOSE DOES
We read the entire AI & tech firehose — so you don't have to.
PODPodcastsAll-In, No Priors, Acquired…
NEWNewslettersStratechery, Newcomer…
PAPPapersPhysical AI research
PHProduct Huntdaily launches
VCInvestor ScoutSequoia, a16z, Benchmark…
CLAUDE DISTILLS →
7 reads, 30 sec each — free, 6 AM ET.
+ a live graph of the companies, people & themes underneath.
HOME/STRICTLYVC/Wall Street's AI Rout Erases Mor…
NEWS
// NEWSLETTER ISSUE
STRICTLYVC

Wall Street's AI Rout Erases More Than $4 Trillion in Market Value

DATE November 21, 2026SOURCE STRICTLYVCPARTICIPANTS CONNIE LOIZOS
// KEY TAKEAWAYS4 ITEMS
  1. 01Theme: A Violent AI Market Reset, Even as Private Capital Keeps Flowing
  2. 02Theme: Open-Weight Models and "Skip the Frontier" as the Enterprise Cost Play
  3. 03Theme: The AI Agent Stack Is Spawning "Picks-and-Shovels" Startups, Especially Governance and Security
  4. 04Theme: Nvidia and China: A Compounding Governance and Regulatory Overhang
In this episode
// SUMMARY

1. Key Themes

Theme: A Violent AI Market Reset, Even as Private Capital Keeps Flowing

The public-market rout is severe, and Nvidia is its epicenter.

"Wall Street's AI rout erased more than $4 trillion in market value from Nasdaq companies this week, wiping out the index's 2026 gains as Nvidia plunged 14% Friday and even Anthropic's record-breaking IPO failed to restore investor confidence."

"The company's performance today capped a seventh straight weekly decline — the stock's longest losing streak since 2008 — and left the chipmaker down 44%, or roughly $2.5 trillion in market value, over that span. It is currently trading at a remarkably low 9.3 times forward earnings."

The selloff is already choking the IPO pipeline for AI infrastructure.

"DayOne Data Centers has reportedly put plans for a U.S. IPO that could have raised as much as $5 billion on hold amid growing investor concerns about data-center and AI valuations, though the Singapore-based company could revive the listing in 2027 if market conditions improve."

Private markets are decoupled, at least for now.

"Databricks... is reportedly in early talks to raise several billion dollars at a valuation of about $225 billion, up from $190 billion in August."

"Pragmatik Labs, a Shanghai startup founded this year that is building AI agents for digital and physical environments, raised a round of tens of millions of dollars at a $5 billion valuation, up from $2 billion in August."

Anthropic's IPO was historically large but did not stabilize sentiment.

"Anthropic raised $93.5 billion in its IPO after underwriters exercised the full 'greenshoe' option, bringing total proceeds to $119.5 billion including a concurrent $26 billion private placement."

Theme: Open-Weight Models and "Skip the Frontier" as the Enterprise Cost Play

Enterprises are looking for ways to avoid paying frontier-model prices.

"That's the enormous opportunity being chased by companies like Fireworks AI, as enterprises seek ways to avoid paying frontier prices for their AI needs."

The pitch is cheaper, faster, and more defensible.

"The Bay Area-based outfit specializes in open-weight models, which anyone can download, and helps customers tune them on their own data so they run faster and cheaper than the models of Anthropic, OpenAI, and other heavyweights, and are harder for rivals to copy."

Data leakage to frontier labs is a growing driver.

"There are also growing concerns that frontier labs absorb too much of their customers' know-how, even if only through the prompts used to get answers, a worry Microsoft CEO Satya Nadella raised in July."

Theme: The AI Agent Stack Is Spawning "Picks-and-Shovels" Startups, Especially Governance and Security

A cluster of seed-to-Series-A companies targets the control, testing, and security layer for agents.

"Metano... provides centralized discovery, behavioral monitoring, and security controls for AI agents operating across enterprise systems, raised a $12 million seed round led by Sierra Ventures."

"Disseqt AI, a one-year-old Dublin startup that tests and monitors companies' AI agents before and after launch, enforcing policies and keeping compliance audit trails, raised an $8 million round."

Agents are also moving into vertical workflows and developer tooling.

"Modern Industrials... builds custom AI agents that automate quoting, order entry, and other workflows for manufacturers and distributors, raised a $6 million seed round."

"Hoplite... lets software teams run multiple AI coding agents in separate cloud workspaces, raised a $5.5 million round led by Patrick Collison and John Collison."

Platform players are moving too.

"Apple appears to be building an AI agent that can operate a Mac on a user's behalf, including browsing the web, working across apps, running commands, and writing and debugging code, according to a job listing first spotted by Bloomberg."

Theme: Nvidia and China: A Compounding Governance and Regulatory Overhang

Regulatory exposure is mounting at the same time as the stock collapses.

"California Attorney General Rob Bonta has subpoenaed Nvidia as part of an investigation into whether the company or former CEO Jensen Huang helped China obtain advanced chips or confidential information and whether Nvidia misled investors about those risks."

"The Commerce Department approved sales of more than 879,000 Nvidia and AMD AI chips to Chinese companies as of May, including some firms designated by the Pentagon as 'Chinese Military Companies,' according to documents released by the House Foreign Affairs Committee."

Leadership-conduct reporting adds another risk layer.

"The New York Times reports that Jensen Huang spent decades using humiliation as a management tool at Nvidia, berating employees with personal insults and violent threats and prompting workplace complaints and at least two settlements with former senior executives who raised concerns about his conduct."

2. Contrarian Perspectives

Most Companies Don't Need Frontier Models

At the moment of peak frontier-lab enthusiasm, the Slush conference conversation was about the opposite. Fireworks' co-founder argues the premium isn't justified for most use cases:

"Fireworks co-founder Chenyu Zhao and many others believe companies are gravitating toward platforms like his because frontier models are pricey and, in most cases, he says, unnecessary unless you need the very latest and greatest generalized intelligence."

Supporting evidence: the same day Anthropic had the largest IPO on record, "far away at the Slush startup conference in Helsinki, the topic du jour was open source everything." Fireworks' customer list (Uber, Shopify, Notion) and its $17.5 billion valuation suggest real demand, though the disclosed breaches temper the case.

Private AI Valuations Are Rising While Public AI Valuations Collapse

The divergence is itself a contrarian signal. Nvidia trades at "a remarkably low 9.3 times forward earnings" after losing 44%, while private rounds reprice sharply upward (Databricks $190B to ~$225B; Pragmatik Labs $2B to $5B in three months). Either public markets are overcorrecting or private marks are lagging, and the article implicitly forces the question.

3. Companies Identified

Nvidia

Description: AI chipmaker, until recently a market bellwether. Why mentioned: Center of the AI rout and a California investigation. Quotes:

"Nvidia plunged 14% Friday... a seventh straight weekly decline — the stock's longest losing streak since 2008." "It is currently trading at a remarkably low 9.3 times forward earnings."

Anthropic

Description: Frontier AI lab, now public. Why mentioned: Record-breaking IPO; researcher claims a mathematical breakthrough with Claude. Quotes:

"Anthropic raised $93.5 billion in its IPO after underwriters exercised the full 'greenshoe' option, bringing total proceeds to $119.5 billion including a concurrent $26 billion private placement."

Fireworks AI

Description: Four-year-old Bay Area inference platform specializing in open-weight models, backed by Sequoia and Benchmark. Why mentioned: Case study in the "skip the pricey models" thesis, and in the security risks of AI infrastructure vendors. Quotes:

"Fireworks runs AI models for other businesses... Customers have included Uber, Shopify, and Notion." "Fireworks has also gotten its first black eye, having disclosed two security incidents this fall." "Fireworks... was assigned a $17.5 billion valuation in summer."

Databricks

Description: 13-year-old San Francisco enterprise data and AI software company. Why mentioned: Reportedly raising at ~$225B despite the public rout. Quotes:

"...reportedly in early talks to raise several billion dollars at a valuation of about $225 billion, up from $190 billion in August."

Pragmatik Labs

Description: Shanghai AI-agent startup founded this year. Why mentioned: 2.5x valuation step-up in three months. Quotes:

"...raised a round of tens of millions of dollars at a $5 billion valuation, up from $2 billion in August."

CrowdStrike / Zscaler

Description: Cybersecurity leaders. Why mentioned: Potential $32B+ acquisition. Quotes:

"CrowdStrike is reportedly in talks to acquire Zscaler in a deal that could value the zero-trust cybersecurity provider at more than $32 billion."

DayOne Data Centers

Description: Singapore-based data-center operator. Why mentioned: Shelved IPO as a symptom of the AI-infrastructure selloff. Quotes:

"...put plans for a U.S. IPO that could have raised as much as $5 billion on hold amid growing investor concerns about data-center and AI valuations."

Acutronic Group

Description: 53-year-old Zurich/Austin maker of small jet engines and test equipment for aerospace and defense. Why mentioned: $100M round in a defense-adjacent hardware niche. Quotes:

"...raised a $100 million equity round co-led by InterAlpen Partners and Hale Capital Partners."

Metano

Description: San Jose agent-security startup founded this year. Why mentioned: Enterprise agent governance. Quotes:

"...provides centralized discovery, behavioral monitoring, and security controls for AI agents operating across enterprise systems, raised a $12 million seed round."

Disseqt AI

Description: Dublin AI-agent testing and compliance startup. Why mentioned: Agent assurance and audit trails. Quotes:

"...tests and monitors companies' AI agents before and after launch, enforcing policies and keeping compliance audit trails, raised an $8 million round."

Hoplite

Description: SF/Dublin startup for parallel AI coding agents. Why mentioned: Notable angel cap table. Quotes:

"...raised a $5.5 million round led by Patrick Collison and John Collison, with... Y Combinator, Paul Graham, and Des Traynor also buying in."

Modern Industrials

Description: NYC startup building custom AI agents for manufacturers and distributors. Why mentioned: Vertical agents for industrial workflows. Quotes:

"...automate quoting, order entry, and other workflows for manufacturers and distributors, raised a $6 million seed round co-led by Glasswing Ventures and TMV, with Andreessen Horowitz also chipping in."

Touchmark

Description: SF marketplace for reserved AI inference capacity. Why mentioned: A market for compute capacity, relevant given the volatility in AI infrastructure. Quotes:

"...operates a marketplace for buying, selling, and managing reserved AI inference capacity, raised a $3 million seed round."

Archer

Description: London/NY AI recruiting-agent company. Why mentioned: $10M round. Quotes:

"...uses AI agents to match job candidates with employers and qualify applicants for recruiters, raised a $10 million round led by Volition Capital."

InPrime Finserv

Description: Bengaluru small-business lender. Why mentioned: $10.9M Series B serving underbanked borrowers. Quotes:

"...lends to small-business owners and self-employed borrowers underserved by traditional banks, raised a $10.9 million Series B led by FMO."

Eight by Zero

Description: Montreal employee-data integration startup. Why mentioned: Insurance and HR integration layer seed round. Quotes:

"...connects insurers, third-party administrators, employers, and HR systems through a shared employee-data integration layer, raised a $2.1 million seed round."

IntelliProve

Description: Ghent startup turning cameras into contactless vital-sign monitors. Why mentioned: $3M seed. Quotes:

"...turns cameras in everyday devices into contactless health monitors that measure vital signs from facial scans."

Europa Fusion

Description: Secretive London pulsed-power fusion startup. Why mentioned: Backed by Daniel Ek. Quotes:

"...pursuing a simulation-heavy approach to pulsed-power fusion, has received backing from Spotify founder Daniel Ek."

SE Ventures

Description: Schneider Electric's €1B venture arm. Why mentioned: Expanding its accelerator to Europe. Quotes:

"...offering selected early-stage energy and industrial intelligence startups a €100,000 SAFE plus access to Schneider Electric and AVEVA experts, customers, and partners; applications close Jan. 15."

Uber / Notion / Shopify

Description: Fireworks customers. Why mentioned: Uber and Notion were affected by the Fireworks breaches. Quotes:

"Uber says hackers stole the names, addresses, birth dates and driver's license numbers of about 150,000 drivers through a vendor that TechCrunch identified as Fireworks." "Notion says AI chat transcripts from fewer than 1% of its workspaces were exposed, but it has disconnected from Fireworks."

Apple

Description: Consumer-tech giant. Why mentioned: Reportedly building an on-device agent. Quotes:

"Apple appears to be building an AI agent that can operate a Mac on a user's behalf."

Take-Two Interactive / Rockstar Games

Description: Publisher of Grand Theft Auto VI. Why mentioned: Record entertainment launch. Quotes:

"Grand Theft Auto VI generated $3.79 billion in preorder and first-day sales, making it the biggest entertainment launch ever."

Fruit of the Loom Japan

Description: Apparel brand. Why mentioned: Sold-out product launch. Quotes:

"...launched a $240 machine-washable cotton 'T-suit'... has already sold out at most stores carrying it."

4. People Identified

Jensen Huang

Description: Former Nvidia CEO. Why mentioned: Subject of California investigation and NYT reporting. Quotes:

"...whether the company or former CEO Jensen Huang helped China obtain advanced chips or confidential information." "Jensen Huang spent decades using humiliation as a management tool at Nvidia."

Rob Bonta

Description: California Attorney General. Why mentioned: Issued the Nvidia subpoena. Quotes:

"California Attorney General Rob Bonta has subpoenaed Nvidia."

Chenyu Zhao

Description: Co-founder of Fireworks AI. Why mentioned: Articulates the anti-frontier-pricing thesis. Quotes:

"...frontier models are pricey and, in most cases, he says, unnecessary unless you need the very latest and greatest generalized intelligence."

Satya Nadella

Description: Microsoft CEO. Why mentioned: Raised concerns about frontier labs absorbing customer know-how. Quotes:

"...a worry Microsoft CEO Satya Nadella raised in July."

Levent Alpöge

Description: Anthropic mathematician. Why mentioned: Claims to have used Claude to prove the Riemann hypothesis. Quotes:

"...claims to have used Claude to prove the 167-year-old Riemann hypothesis, publishing the argument in Lean, a programming language and proof assistant that can formally check each step of a mathematical proof."

Tim Genewein

Description: Google DeepMind researcher. Why mentioned: Forecasts AGI-to-superintelligence timeline. Quotes:

"...argues that AI could progress within the next decade from AGI... to superintelligence, which would outperform large groups of human experts across virtually all tasks."

Patrick Collison and John Collison

Description: Stripe founders. Why mentioned: Led Hoplite's round. Quotes:

"...raised a $5.5 million round led by Patrick Collison and John Collison."

Paul Graham, Des Traynor

Description: Y Combinator co-founder; Intercom co-founder. Why mentioned: Hoplite backers. Quotes:

"...Y Combinator, Paul Graham, and Des Traynor also buying in."

Daniel Ek

Description: Spotify founder. Why mentioned: Backer of Europa Fusion. Quotes:

"...has received backing from Spotify founder Daniel Ek."

Sean Cairncross

Description: Former White House national cyber director. Why mentioned: Fired after a string of outages. Quotes:

"President Trump fired White House national cyber director Sean Cairncross days after a string of technology outages disrupted U.S. air travel, banking, and national security systems."

Rahul Dravid

Description: Investor in Disseqt AI (named only in the funding line). Why mentioned: Participant in the round. Quotes:

"...additional participation from Sure Valley Ventures, Delta Partners, Enterprise Ireland, and Rahul Dravid."

5. Operating Insights

Vet Your Vendors: Your Infrastructure Provider's Breach Is Your Breach

Fireworks' incidents cascaded to customers, costing them customer data and trust. Uber lost driver PII "through a vendor," and Notion disconnected entirely.

"Notion says AI chat transcripts from fewer than 1% of its workspaces were exposed, but it has disconnected from Fireworks... Fireworks will have to win back the confidence of customers like them."

Takeaway: for AI vendors especially, audit secrets handling, cloud-provider exposure, and what data (prompts, billing, datasets) you're sending them.

Match Model Tier to the Task and Own Your Differentiation

Tuning open-weight models on proprietary data can cut cost and latency and reduce leakage of your know-how.

"...helps customers tune them on their own data so they run faster and cheaper... and are harder for rivals to copy."

Takeaway: reserve frontier models for tasks needing "the very latest and greatest generalized intelligence" and evaluate open-weight alternatives for the rest.

Build Governance Into Agent Deployments from Day One

The funding cluster (Metano, Disseqt) signals that discovery, monitoring, policy enforcement, and audit trails are becoming prerequisites.

"...enforcing policies and keeping compliance audit trails."

Takeaway: before scaling agents, inventory them, monitor behavior, and keep auditable records.

6. Overlooked Insights

A Marketplace for Reserved AI Inference Capacity Is Emerging

Touchmark's seed round hints at inference compute becoming a tradable, financialized commodity, which could matter more as AI infrastructure valuations reset.

"...operates a marketplace for buying, selling, and managing reserved AI inference capacity."

Defense and Industrial Hardware Is Still Attracting Large Checks

In the middle of an AI-centric news cycle, a 53-year-old jet-engine maker quietly raised $100 million, and Schneider's venture arm is pushing "industrial intelligence" startups.

"Acutronic Group, a 53-year-old Zurich and Austin company that manufactures small jet engines and specialized test equipment for aerospace and defense systems, raised a $100 million equity round."