← Daily AI news

Daily AI news, Friday, 6 November 2026

Daily AI / Tech / VC Briefing: Friday 6 November 2026

Anthropic logo and brand image
Image: Anthropic (anthropic.com)

Good afternoon, here's today's briefing. Anthropic's IPO talks are being repriced down after a brutal tech selloff, OpenAI is closing the first slice of a smaller private round while landing JPMorgan as a Codex and cyber customer, and SpaceX has named a date for its first attempt to catch a Starship with the tower.

LEAD: Anthropic IPO talks slip toward $1.2–$1.3 trillion

Background: Anthropic has been racing toward what could be the first major public listing by a frontier model lab, with a roadshow once expected to start the week of 9 November and trading as early as the week of 16 November. Earlier investor chatter had put a valuation as high as $2 trillion, then roughly $1.4–$1.5 trillion after a financial briefing.

What happened: Bloomberg reported on Friday that some prospective investors are now discussing about $1.2–$1.3 trillion, citing a person familiar with the matter. No final decisions have been made; Anthropic declined to comment. That follows a Wall Street Journal report on Thursday that major investors were sceptical of $1.5 trillion, with one floating a figure closer to $1 trillion, and a Reuters exclusive the same evening saying four people familiar with investor talks doubted whether the roadshow could still start next week for a November listing. Investors want a bigger discount, are watching for Anthropic's still-blank October revenue figure in the S-1, and note the lab's quarterly losses and huge compute commitments. Thursday's Nasdaq 100 plunge of about 6.4% — roughly $2.3 trillion of market value — was the largest single-day decline on record for that index, according to the Bloomberg write-up, and has clouded the outlook for a deal some had sized near $100 billion.

Why it matters: This is the first real public-market stress test for a frontier AI lab. A clean listing at a grounded price would reassure the AI trade; a delay or a scramble on valuation would tighten the funding window for everyone else still private, including OpenAI.

Sources: Bloomberg via Investing.com · Reuters via Investing.com · Yahoo Finance IPO guide

OpenAI: SoftBank and Coatue scale back; first funding tranche due today

Background: Yesterday's briefing noted secondary reports that OpenAI had cut a private raise from about $30 billion toward $20 billion and a talked-about valuation from about $1.4 trillion to $1.25 trillion. SoftBank remains one of OpenAI's largest backers and a Stargate partner.

What happened: NYT DealBook, as relayed today, says OpenAI is seeking up to $20 billion at a $1.25 trillion valuation after major backers signalled smaller cheques than before. A UAE investor group including MGX is expected to provide the largest commitment, in the single-digit billions. SoftBank is in talks to invest hundreds of millions — far below its earlier multi-billion cheques — and Coatue, which also backs Anthropic, has discussed a similar amount. The raise is split into two parts, with the first tranche due to close on Friday and the rest expected in coming weeks. SoftBank's stake is estimated around 13% after nearly $65 billion invested over time.

Why it matters: Even OpenAI's closest capital partners are moderating new exposure while Anthropic prices a public book. That is a signal about risk appetite in private AI markets, not just about one round's size.

Sources: Crypto Briefing / DealBook · Reuters / TradingView on SoftBank

Competitor move: OpenAI puts Codex and cyber models inside JPMorgan

Background: JPMorgan already runs an internal LLM Suite with OpenAI and Anthropic models, and has worked with Anthropic since April on Project Glasswing for defensive cybersecurity. OpenAI has been pushing Codex and specialised Daybreak cyber models as enterprise products.

What happened: At the Wall Street Journal Tech Conference on Wednesday, OpenAI president Greg Brockman said OpenAI has a strategic partnership with JPMorgan Chase to deploy Codex and specialised cybersecurity models across engineering, security and broader enterprise work. JPMorgan has used OpenAI models since 2024; by early 2026 its LLM Suite had about 150,000 weekly active users, and the bank has cited roughly $2 billion a year in AI-driven savings. The new deal does not replace Anthropic's Glasswing work — it puts both labs competing for the same bank's cyber and coding budget.

Why it matters: Winning the largest US bank as a coding-agent and cyber customer is the kind of enterprise proof OpenAI wants ahead of any 2027 listing talk, and it raises the bar for Anthropic, Cursor and Cognition in regulated industries.

Sources: Crypto Briefing · AI Affairs on Brockman

SpaceX targets 13 November for Starship Flight 15 and the first ship catch

Background: SpaceX has caught the Super Heavy booster three times, but the ship returns from orbital velocity and has never been caught. Flight 14 on 28 September was the first Starship to reach orbit. Yesterday's briefing covered Goldman and Apollo shopping pieces of SpaceX's roughly $40 billion AI-chip lease — that financing story is unchanged.

What happened: SpaceX vice president of launch Kiko Dontchev said at the UP.Summit on Thursday that Flight 15 is targeting Friday 13 November for the company's first attempt to catch the ship with the launch tower, according to Teslarati citing Ars Technica's Eric Berger, who was there. It is the first date given directly by a SpaceX executive; earlier FAA planning slides had listed no earlier than 8 November, then 12 November. Hardware for Booster 22 and Ship 42 has completed static fires. The FAA has not yet listed a Flight 15 licence. Elon Musk previously put the odds of a successful ship catch at "more likely than not," and has said a successful catch would enable a Flight 16 that reflies both stages.

Why it matters: Full tower recovery of an orbital-class upper stage is the step SpaceX needs before a rapid Starship cadence — and a public milestone while SPCX trades through the AI-market chop.

Sources: Teslarati

Microsoft tightens the leash on Nebius, IREN and Nscale

Background: Yesterday's briefing covered Microsoft declining to expand its Texas deal with UK AI-cloud firm Nscale. "Neoclouds" — specialised GPU landlords — have been a fast way for hyperscalers to add capacity.

What happened: Bloomberg, relayed by Reuters on Thursday, says Microsoft has tightened oversight of some data-centre projects from Nebius, IREN and Nscale, is stepping away from deals it sees as risky, and is debating internally whether to sign new neocloud agreements while focusing more on its own campuses. IREN said in earnings it was pushing back some Microsoft AI-infrastructure delivery dates after network-design change requests, without financial penalties for those delays. Nebius and IREN shares fell about 14% and 17% on the day; CoreWeave about 10%. A Microsoft spokesperson told Reuters its approach mixes owned sites, leases and third-party partners. The Nscale Texas option and Essex pull-out remain as reported yesterday.

Why it matters: Hyperscaler caution on rented GPU capacity compounds the midterms data-centre backlash: power, permits and counterparty risk are now pricing into AI infrastructure equities.

Sources: Reuters via Investing.com

Traversal raises $120 million at a $1 billion valuation for AI site-reliability agents

Background: Coding agents have sped up how much software ships; site-reliability engineers then inherit more incidents. Splunk and Oxford Economics have estimated enterprise downtime costs up more than 50% in two years.

What happened: Traversal, which builds autonomous AI agents that diagnose, remediate and help prevent production outages, raised a $120 million Series B at a $1 billion valuation, co-led by Kleiner Perkins, Redpoint Ventures and Khosla Ventures, with Sequoia, Battery and Amex Ventures participating. SiliconANGLE reports revenue up more than nine-fold and monthly active users up about 20× in the past year, with customers including PepsiCo, DigitalOcean, Nubank and Harvey. CEO Anish Agarwal says general-purpose frontier models still struggle on SRE tasks; Traversal leans on causal machine learning over telemetry, logs and traces. The company says it will accelerate hiring in engineering, go-to-market and R&D.

Why it matters: Capital is following the "outer loop" of AI software — not just writing code, but keeping AI-accelerated systems running — which is the next battleground after coding agents.

Sources: SiliconANGLE · Business Wire

Sentra raises $150 million to lock down data before AI agents touch it

Background: As companies roll out copilots and autonomous agents, each agent becomes a new identity that can reach sensitive data at machine speed. Traditional perimeter tools were built for people, not fleets of non-human identities.

What happened: Sentra, an agentic data-security company based in New York and Tel Aviv, announced a $150 million strategic investment led by Capital One Ventures, which takes a board seat. Existing backers include Bessemer, Zeev Ventures, Standard Investments, Munich Re Ventures and Key1 Capital. Sentra says it grew more than 300% year over year. Its pitch is to classify sensitive data where it lives, map what humans and AI agents can reach, enforce least privilege for non-human identities, and contain unexpected agent access — without copying customer data into Sentra's own cloud. Lyft's CISO is among the named customers.

Why it matters: The agent boom is creating a security market that sits beside coding tools: who and what can see the data those agents need. Capital One putting strategic money and a board seat behind that thesis is a clear enterprise signal.

Sources: Sentra · Axios Pro

Also noted

Cursor shipped "Build Live Apps" for cloud agents — scaffold, host, security-check and publish live web apps behind Cursor SSO — covered in today's agents briefing; no separate hire of the day beyond yesterday's Cat Wu move to OpenAI. China's Sand.ai raised about $200 million for generative video. No material new Cognition product or named key-hire story cleared the bar today.

← Thursday, 5 November 2026 Saturday, 7 November 2026 →

Get the weekly sift

The best new AI and agent releases, with one-line verdicts. Occasional picks, no spam.

Privacy policy: your email goes to our email provider, Resend. Unsubscribe any time.

News is editorial and never paid for. Sponsored content on ToolSift is always labelled.