🧭 Anthropic IPO Prices at $82 Per Share — 5.1% Above Range, Valuing the Company at Approximately $1.48 Trillion Fully Diluted
After market close on November 18, Anthropic and its underwriters — Goldman Sachs and Morgan Stanley leading — set the final IPO price at $82 per share, clearing the stated $70–$78 range and landing 5.1% above the ceiling. At this price, the sale of 825 million Class A shares raises approximately $67.6 billion in gross proceeds. The fully diluted share count implies a market capitalisation of roughly $1.48 trillion at pricing — below the crypto-futures-implied $1.7 trillion that had circulated earlier in the day, but substantially above the most optimistic institutional price-discovery signals from last week. Class A shares carry one vote each; existing investors hold Class B (10 votes) and Class C (no votes) shares. The Class C structure covers the $26 billion in private placement commitments from Google, Nvidia, Microsoft, TSMC, and three other technology partners. Nasdaq trading under the ticker ANT opens at 09:30 ET on November 19.
What $82 means in context
- Largest US tech IPO since Meta (2012): At $67.6 billion in gross proceeds, the offering surpasses Alibaba's $25 billion 2014 US listing and trails only Saudi Aramco ($29.4 billion) among all-time global IPOs in raw proceeds — though Aramco's valuation multiple was far lower. By market cap at pricing, Anthropic immediately enters the global top-25 public companies.
- Underwriter greenshoe: Goldman Sachs and Morgan Stanley hold a 30-day over-allotment option to purchase up to an additional 123.75 million shares at $82. Full exercise would add roughly $10.1 billion and push gross proceeds above $77 billion.
- Lock-up cliff: Existing investors — including Amazon ($8 billion committed), Google ($3 billion), and Spark Capital — are subject to a 180-day lock-up expiring mid-May 2027. The $26 billion Class C private placement partners have a separate 90-day lock-up. These cliff dates are the first structural stress-test for the post-IPO price.
- Revenue basis: Anthropic's last reported annualised revenue run-rate (October 2026) was approximately $5.4 billion. At $82/share the price-to-annualised-revenue multiple is approximately 274×, consistent with a high-growth AI infrastructure company priced on future earnings, not current cash flows.
- Public Benefit Fund trigger: With pricing confirmed, the commitment to transfer 5% of fully diluted shares to the independent Public Benefit Fund is now legally effective. At $82 per share, that represents approximately $1.21 billion in notional equity at opening.
What to watch on day one
An IPO open significantly above the offer price (a "pop") benefits retail allocants and signals underwriter conservatism; a price that falls below $82 on day one would be unusual given 7x oversubscription but is not unprecedented in volatile markets. The more important signal for enterprise buyers is what happens to the stock in the 60–90 days after the greenshoe option expires: that is when early investors who received allocations typically begin rebalancing, and it sets the reference price for the first 180-day lock-up expiry in May 2027.
IPO
ANT
Nasdaq
$82/share
$1.48 trillion
Goldman Sachs
Morgan Stanley
greenshoe
Public Benefit Fund
🧭 Claude Code 2.1.326 — Remote Control Streaming, Code Mode + Agent Teams, and Sandbox Write Hardening
Claude Code 2.1.326 shipped on November 18 with three developer-facing improvements and one security hardening change. The headline feature is live streaming for Remote Control: responses from claude remote-control now stream token-by-token to the mobile app or browser interface as they are written, rather than arriving as a single block after completion. This makes long remote-control sessions dramatically more usable — you can follow Claude's reasoning in real time and interrupt mid-output if the direction is wrong, rather than waiting several seconds for a full response before correcting course.
Key changes
- Code Mode + Agent Teams simultaneously: Previously, enabling Code Mode (where Claude writes raw JavaScript to compose its own tool calls) was mutually exclusive with running Agent Teams. As of 2.1.326, the two modes can operate together. Existing sessions that had both configured will now launch directly in Code Mode; teams deployed via the API will inherit any Code Mode configuration from the initiating session.
- Sandbox write hardening: The macOS, Linux, and WSL sandbox now blocks writes to sensitive home-directory paths by default:
~/.ssh/, ~/.gnupg/, ~/.aws/credentials, ~/.config/gh/, and similar credential stores. File tools will refuse writes to files whose names or paths match a known credential-file pattern (e.g. .credentials.json, id_rsa). This addresses a class of prompt-injection attacks where a malicious tool response attempts to overwrite SSH keys or cloud credentials.
- Hook decision propagation fix: Hooks returning
"decision": "block" now reliably hold tool execution in several edge cases that previously allowed the tool call to proceed. This was a correctness bug — if your permission hooks were blocking fewer calls than expected, this release fixes that.
- Performance improvements: Lower idle CPU usage in long sessions; faster keystroke rendering in high-latency environments; improved resize behaviour when terminal windows are narrow.
Breaking change for Bash permission scopes
Some Bash commands that were previously allowed by a broad "bash" permission now require an explicit, narrower permission string. If you have automation that runs Claude Code unattended (e.g. via cron or CI), audit your .claude/settings.json allowedTools entries after upgrading. The specific commands affected are listed in the changelog; the most common impact is on curl and wget invocations that now require "bash(curl:*)" rather than a bare "bash" entry.
Claude Code
2.1.326
Remote Control
streaming
Code Mode
Agent Teams
sandbox
security
🧭 Anthropic Cuts Coordinated Vulnerability Disclosure Window from 90 to 30 Days — Citing AI-Accelerated Security Research
Anthropic published a revised Coordinated Vulnerability Disclosure (CVD) policy on November 16, effective 1 January 2027. The core change: high- and critical-severity vulnerabilities must now be fully patched and publicly disclosed within 30 days of private notification — down from 90 days. The stated rationale is direct: "Frontier models have become much better at discovering vulnerabilities." The 90-day standard was designed for a manual-research era; Anthropic argues that the same AI capabilities it ships are now routinely used by security researchers (and attackers) to find bugs far faster than human-only teams could previously, and that the standard window is too long to protect users under the new threat model.
Full timeline changes
- High/critical severity: 90 days → 30 days from private notification to patch + public disclosure
- All other severity levels: 90 days → 45 days
- Escalation to external coordinator (if maintainer non-responsive): 30 days → 15 days
- Post-patch technical detail embargo: 45 days → 30 days (the window between patch release and publishing a full write-up)
- Actively exploited vulnerabilities: 7-day window is unchanged
What this means for teams building on Claude APIs
If you use Claude APIs in a product and receive a Coordinated Vulnerability Disclosure from Anthropic about a component in your stack, you now have 30 days (for high/critical) to patch and disclose — not 90. That is a significant operational tightening. Teams without a documented incident-response runbook for third-party CVD notifications should build one now, before January 1. Anthropic is offering free access to its OSS Scanner and the Claude for Open Source programme to help maintainers move faster. For security teams already operating under Google Project Zero's 90-day policy, this effectively halves the runway on the most severe class of issues.
# Practical checklist for January 2027 readiness
# ─────────────────────────────────────────────────
1. Add Anthropic Security to your incident-response distribution list
2. Document who owns CVD triage in your on-call rotation
3. Set a 25-day internal deadline (5-day buffer before the 30-day hard limit)
4. Enroll in Anthropic's OSS Scanner (free) for proactive detection
5. Test your patch-and-deploy pipeline can hit a 25-day target on a drill
CVD
vulnerability disclosure
security
30-day window
AI-accelerated research
incident response
OSS Scanner