US AI Company Product Matrix · Updated 11-20

Anthropic

Anthropic's play is"Enterprise APIs + coding": ~85% of revenue from APIs and enterprise services; Claude Code takes the #1 spot in AI coding; enterprise adoption overtook OpenAI for the first time in April. No cash-burning consumer war, no hardware — MCP defines the agent protocol. Quietly eating the B2B side.

Anthropic lists on Nasdaq Nov 19 (ticker ANT): priced at $83, opened at $111.50 (+34%), intraday high ~$115.10, low $92.72, closed at $97.17 (+17.07%); volume over 846M shares; final offering of 980M shares raised $81.34B in public proceeds — the largest IPO ever (above SpaceX's $75B); underwriters hold a 30-day over-allotment option on 147M shares, worth ~$93.5B if fully exercised; a further $26B Class N private placement brings public + private to ~$107.3B; ~13.22B shares outstanding after the offering, ~$1.28T market cap at the close (11-19)

Key numbersAt a glance

~$81B
Annualized revenue run rate (S-1/A 11-10: revenue of $33.2B in the first three quarters, Q3 ~$16.96B, +47% QoQ, ~$81B as of Oct 31; total compute purchase and lease commitments ~$900.9B)
$97.17 (+17.07%)
IPO first-day close (priced at $83, opened at $111.50, +34%); pricing weighed above the $70–$78 range (FT 11-11, citing five people familiar: strong first-day roadshow demand; valuation target could rise above $1.5T, ~$83 per share on a fully diluted basis, total raise possibly over $100B; Anthropic has not confirmed, latest filing still $70–$78; the original range covers 825M Class A shares offered, raising up to ~$64.4B, Nasdaq ticker ANT; seven strategic investors — Google, NVIDIA, AMD, Microsoft, Broadcom, Micron, TSMC — privately placing $26B of non-voting Class N shares at the IPO price; pricing / listing date not set; 11-12: the order book was fully covered on day one at the $70–$78 range (confirmed by CNBC), with sources describing multiple-times oversubscription; FT and Bloomberg report banks are considering lifting the valuation above $1.5T, while the filed S-1/A terms are unchanged); 11-13: on the third roadshow day, orders reached 6× the shares on offer (Reuters, citing people familiar; Anthropic declined to comment, order counts may still change); pricing expected Nov 18, Nasdaq listing Nov 19 (ticker ANT); Barron's adds: October revenue annualized at an $81B run rate, Q3 revenue $17B, adjusted margin 15%, long-term target 35%, up to 20% of shares for retail investors; 11-16: orders reached nearly 7× the shares on offer (Bloomberg, up from ~6× a week earlier), pricing Nov 18 and Nasdaq listing Nov 19 unchanged; seven supplier-investors (Google, NVIDIA, AMD, Broadcom, Micron, Microsoft, TSMC) are privately placing $26B in total (Google and NVIDIA ~$10B each); 11-17: third S-1/A filed, Nasdaq has approved the listing (ticker ANT); pricing range unchanged at $70–$78, 825M Class A shares offered (company issuing ~801M, existing shareholders selling ~23.65M), underwriters' over-allotment 123.75M shares; post-IPO Amazon holds 14.7%, Google 13.7% (of shares outstanding); concurrent $26B private placement (Google and NVIDIA $10B each, AMD $2.5B, Broadcom / Micron / Microsoft $1B each, TSMC $0.5B); net proceeds ~$84.6B at the $74 midpoint ($93.7B if the over-allotment is fully exercised); annualized revenue ~$81B as of Oct 31; the service condition date for ~67M RSUs moves from Nov 20 to Nov 19 (expected first trading day); 11-18 (US evening, ~05:26 Beijing time Nov 19): Reuters first reported pricing at $83/share, cross-cited by Bloomberg, CNBC and TechCrunch citing people familiar; Anthropic declined to comment, and no Form 424B4 pricing filing had been seen by the window's close, so $83 is reported pricing pending official filing; Nasdaq listing Nov 19 at 09:30 ET (ticker ANT); ~$68.5B on the original 825M shares (offer size reported raised but the final share count undisclosed, so ~$68.5B is not the final raise); concurrent $26B Class N private placement (AMD, Broadcom, Google, Micron, Microsoft, NVIDIA, TSMC Partners); valuation ~$1.5T (fully diluted, an estimate) / ~$1.08T (on ~13.05B shares outstanding), both pending official filing; Q3 revenue $17.0B, $33.2B in the first nine months, consistent with the figures already recorded; 11-19 first day: priced at $83, opened at $111.50 (+34%), intraday high ~$115.10, low $92.72, closed at $97.17 (+17.07%), volume over 846M shares; Form 424B4 filed: final offering of 980M shares (company issuing 956,351,354 + existing shareholders selling 23,648,646), public proceeds $81.34B — the largest IPO ever (above SpaceX's $75B); underwriters' 30-day over-allotment option on 147M shares, ~$93.5B if fully exercised; a further $26B Class N private placement (AMD, Broadcom, Google, Micron, Microsoft, NVIDIA, TSMC Partners), public + private ~$107.3B; ~13.22B shares outstanding after the offering, market cap near $2T at the open (fully diluted) and ~$1.28T at the close; the prospectus also discloses ~$900.9B in infrastructure spending commitments through 2045 (Alphabet at least $121.3B, Amazon $110B, Broadcom equipment leases $161.2B)
$8 / $40
Fable 5.2 pricing per million tokens (was $10 / $50, down 20%); fast mode $16 / $80, up to 2.5× speed; 1M context, up to 128K output
$2.5B+
Claude Code ARR (Feb 2026), ~1/5 of revenue
34.4%
Enterprise adoption (Ramp 04-2026, first time ahead of OpenAI)
85%
Revenue from APIs and enterprise services; subscriptions only ~15%
$19B
TeraWulf 20-year compute lease total (401MW, Kentucky), backed by Google credit support
$567M
Q3 operating loss (disclosed in the S-1/A on 11-12; the prior quarter earned $974M)
$4.3B
New Microsoft compute commitment (up to, disclosed 11-12)
$1.3B
Operating loss in the first nine months of 2026 (S-1 financial details summarized Nov 14, from SEC filings; revenue $33.2B in the same period vs $2.5B a year earlier)

Product layersAll entries

1 · Model layer

The Claude 5.5 family
Claude Fable 5.2 / Mythos 5.2Launched
Released Nov 9: two versions of the same underlying model — Fable 5.2 for all users, Mythos 5.2 only for members of cybersecurity / life-sciences verification programs (some safety blocks removed); output speed up to 30% faster, new fast mode; 1M context, up to 128K output, safety blocks down ~55% vs 5.1 (all Anthropic's own claims)
Pricing $8 / $40 per million tokens (was $10 / $50, down 20%); fast mode $16 / $80, up to 2.5× speed; API IDs claude-fable-5-2 / claude-mythos-5-2; live now on the Claude API, Claude Code, and AWS / Google Cloud / Azure; Pro users can count Fable toward up to 50% of their weekly limit until 23:59 PT Nov 16, then pay by usage credits (API pricing); cross-reported by multiple outlets; 60.0% on CursorBench (Fable 5.2 max), second to Gemini 4 Carbon's 60.4%; 65 on the Artificial Analysis Intelligence Index, first of 74 models
Claude Sonnet 4.6Launched
Released Nov 8 (official blog): the strongest Sonnet yet; 1M-token context in beta, plus context compression (beta); in early Claude Code testing users preferred it over Sonnet 4.5 ~70% of the time and over Opus 4.5 59% of the time; OSWorld 72.5% (Sonnet 4.5: 61.4%), all Anthropic's own test figures
Same pricing as Sonnet 4.5: input $3 / output $15 per million tokens; now the default model on claude.ai and Claude Cowork for Free and Pro; API ID claude-sonnet-4-6; free tier also gains file creation, connectors, skills, and compression
Claude Opus 5.5Launched
Current flagship, released late September; tops the Artificial Analysis AI Index at 58
Pricing ~$4 / $20 per million tokens (inferred from Sonnet's "half price")
Claude Sonnet 5.5Launched
Released Sept 28 (US Pacific); +30% speed, -30% per-task cost; AI Index 56; first Sonnet-level cybersecurity hardening
$2 / $10 per million tokens, cache reads $0.20/M
Claude Haiku 5.5Launched
Officially released Oct 7; third model in the 5.5 family, built for high-frequency classification / summarization / extraction (customer service, voice agents, in-app assistants); first Haiku with built-in hardening against high-risk cybersecurity requests; new adjustable "effort" parameter
Tiered pricing: short-context input $0.10 / output $0.50 per million tokens, a steep cut vs Haiku 4.5 (press figures 75%–90%, matching GPT-6 Luna pricing); live on Claude Platform, AWS, and Google; Reuters
Claude Fable 5.1Launched
Mythos-class, the most widely released strongest model; adaptive thinking
$10 / $50; released 2026-09-01
Opus 5 / Sonnet 5 / Opus 4.7Launched
Opus 5 (July 24, $5/$25); Sonnet 5 (June, $2/$10); Opus 4.7 (GA April 16, 3× image resolution)
Fable / Opus / Sonnet all with 1M context; Haiku 200K

2 · Consumer apps

Productivity-suited; goes head-to-head with Workspace
Claude AppLaunched
Primary web / iOS / Android / desktop chat entry point, with Artifacts
No reliable public MAU figure; topped the App Store in March for a day on a wave of ChatGPT uninstalls (one-off event); Oct 4: Claude began asking users to opt in to voice-conversation data for model training — off by default, separate from chat/code training toggles, can be turned off or deleted anytime
Claude supported-region policyLaunched
Effective Nov 12: Claude stops serving people and entities in mainland China, Hong Kong, Macao, Russia, Iran, North Korea, and elsewhere, extending to companies majority-owned by people or entities from those regions (wherever their employees are based); Taiwan is on the supported list
An availability fact; supported-region policy update
Claude HTML-in-CanvasLaunched
Reported Nov 16: Claude can now place real web elements — text, buttons — inside complex canvas graphics and 3D effects, with content still selectable, clickable, and readable by screen readers; the demo page has 15 examples (e.g. a knitting-pattern page, a 3D-cube website); requires Chrome/Edge 157 or Claude desktop app 5.11492, Safari/Firefox not yet supported, and users must ask for it explicitly in the prompt
The Decoder, Nov 16
"One Claude"In Preview
Rolling out since mid-September: chat Claude merged with desktop Cowork; new Docs / Slides, exportable to Google Docs / Word / PPTX
Pro / Max first; goes head-to-head with Google Workspace; Docs / Slides / Design GA for all tiers (including free) since Oct 8
Claude for Google WorkspaceLaunched
Announced Oct 6, now in public beta for all paid Claude plans: a sidebar inside Docs / Sheets / Slides reads selected content and edits files directly; three beta connectors create/edit Google files from Claude chat
Inherits Google sharing permissions and enterprise compliance controls; cross-reported by multiple secondary outlets the same day (official announcement page not directly captured)
Claude Dashboards + MotionIn Preview
Released Oct 8: Dashboards (beta, paid tiers) connects Redshift / BigQuery / Databricks / Snowflake / Salesforce and more, generating live data dashboards from natural language, every figure expandable to its underlying query; Motion (beta, Team / Enterprise) turns reports and charts into editable-code animated explainer videos (exportable as MP4); Docs / Slides / Design also exited beta — GA for all tiers including free
Reuters Oct 8; also announced on Claude's official X account; 45M+ documents, presentations, and designs have been created with the suite so far
Claude for Small BusinessLaunched
Launched May with 15 workflows; expanded Sept 15 to 43 workflows + 27 connectors (Shopify / Stripe / Salesforce, etc.)
900K+ installs since launch
Midterm election safeguardsLaunched
Announced Oct 14 (ahead of the November midterms): in partnership with nonprofit Democracy Works, US users can look up official voting information and election results inside Claude this month; voter-registration lookups surface a TurboVote link; new abuse-detection classifiers deployed (including foreign-actor detection)
Reuters; company disclosed that ~0.51% of 250K US Claude conversations from Sept 21–27 touched elections/voting
Claude DuetIn Preview
Released Oct 21 as an agentic research preview: daily briefings, background app checks, tasks completed automatically on cloud VMs
iOS + Android + claude.ai; waitlist for US Pro / Max subscribers only; powered by Claude Opus 5.5; no extra charge (counts against existing plan allowance); Reuters
Claude personal finance (Beta)Launched
Released Nov 5: US Pro / Max / Ultra users link bank, credit-card, loan, and investment accounts via Plaid; Claude can read balances, credit limits, and up to 2 years of transactions; read-only — no transfers, money management, or investing, and not used for model training; rolling out this week, US only; up to 16 banks linked; Free / Team / Enterprise excluded
Announced by Anthropic Nov 5; cross-reported by multiple outlets

3 · Developer platform

MCP is the "HTTP" of the agent era
FlashProve verifiable inferenceIn Development
Open technical spec and code (Apache 2.0, incl. a fast C++ verifier and a Lean formal verifier) built with EPFL, MIT, StarkWare, and Succinct: a zkSNARK cryptographic proof for every response, with the serving side first publishing a model-weight commitment (fingerprint), so verifiers can confirm a response came from the claimed model without a GPU or seeing the weights; tested on Meta Llama 4 Maverick (400B MoE, 4-bit public converted weights, weights themselves not in the release package); ~110 tokens/sec proving on a single H100, ~19MB proof per 1,024-token request, ~17s verification on a 24-core CPU; covers only ≤1,024-token contexts and greedy decoding on integer weights, and does not prove the hardware did no other training; Anthropic calls it mid-progress, not a finished research agenda, with no deployment planned in Claude or any lab
Interesting Engineering Nov 4 (TechCrunch via Nogentech, Think Facility); code on GitHub since Oct 30, widely reported Nov 4
Claude APILaunched
All models distributed via AWS / Google Cloud / Azure
Revenue base: ~85% from APIs and enterprise services
Claude API version 2026-10-21Launched
Reported Oct 22: the first official API version number since 2023; Messages API drops temperature/top_p/top_k sampling parameters (assistant turns passing documents, user turns passing citations now return 400 errors); SDK and ant CLI ship major versions with the new version as default
Breaking change developers must handle; panews.io
Message Threads public betaIn Preview
Entered public beta Nov 3: long-conversation optimization for the Messages API — the agent sends the full conversation on the first request, then only new messages plus the previous response ID, instead of resending all history each round; Anthropic says median time-to-first-token for 500K–1M-token requests drops ~35%, and median long-session request bodies shrink from ~800KB to 1.8KB
Threads live in encrypted server-side memory, expire after 5 minutes with no request, then clients must resend the full conversation; token usage, billing, and rate limits unchanged since Claude still reads the full history each round; available for all Claude models across the native platform, AWS, Microsoft Foundry, and Google Cloud; most Claude Code users have it on by default; zero-data-retention and HIPAA customers excluded for now; Gadgets360 and The Decoder Nov 3 (citing the official blog); 35% is a vendor-measured figure
MCPLaunched
Released Nov 2024; donated Dec 9, 2025 to the Agentic AI Foundation under the Linux Foundation (co-founded with OpenAI, Block)
10K+ public servers; v2 stateless revision released July 28, 2026
Agent Skills donationLaunched
Announced Oct 23 on the official blog: donated to the Agentic AI Foundation under the Linux Foundation — the name, logo, GitHub organization, and agentskills.io domain; Anthropic's second foundation move after MCP
70+ agent platforms integrated (ChatGPT, Cursor, GitHub Copilot, VS Code); 1M skills, 280M installs
Agent toolsLaunched
Computer Use, remote MCP connectors, 10 finance agents
Claude Security in private beta: 50+ enterprise partners, 10K+ critical vulnerabilities identified
Cyber Verification ProgramLaunched
Expanded Oct 6 into a three-tier Defense / Red Team / Specialized structure (folded into Project Glasswing), opening Claude Opus 5.5, Sonnet 5.5, and Mythos 5.1 to vetted security professionals with relaxed cybersecurity-blocking classifiers
Glasswing partners found at least 129K verified software vulnerabilities (33K critical/high-severity) from April–July; Reuters
OSS ScannerLaunched
Released Oct 8: a free opt-in service that uses the strongest models (incl. Claude Mythos) to run periodic code-security audits on eligible open-source projects; reports are fully model-generated, sent straight to maintainers' inboxes without human review; sign up by filing a PR to Anthropic's oss-scanner GitHub repo
Its models found 29,000+ candidate vulnerabilities in six months (~6,000 human-verified), producing 584 security advisories as of Oct 2; The Hacker News, Pondera AI, oossa, and others reported Oct 9
OpenTelemetry Trace ExportIn Preview
Entered public beta Oct 14: developers can export traces (latency, token counts, tool calls) from Claude Managed Agents, the Messages API, and the Claude SDK to any OTLP-compatible observability tool (Grafana / Sentry / Splunk / New Relic), following the OpenTelemetry GenAI semantic conventions; the SDK's optional full-content mode captures prompt/response text
Unite.AI and CryptoBriefing cross-reported
Self-hosted Claude Managed AgentsIn Preview
Announced Nov 12, entering private preview: enterprises can run the agent framework, sandboxes, credentials, and session history in their own cloud or on-premises, while model calls still go through Anthropic (default) or Bedrock, Google Cloud, Microsoft Foundry, and others; no public pricing or GA date, available only by applying through the customer team
Announced by Anthropic Nov 12; reported by aiagentupdate Nov 13
Linked Billing BetaIn Preview
Launched Nov 17: customers can link their Claude organization to a third-party app (e.g. Factory, Linear, Ramp) and pay directly for Claude API usage inside that app, without sharing an API key; customers can set monthly caps and unlink at any time; only for organizations paying Anthropic directly
Anthropic official announcement, reported by multiple outlets

4 · Claude Code

The #1 seat in AI coding
Claude CodeLaunched
ARR over $2.5B (Feb 2026), ~1/5 of Anthropic's revenue; enterprise subscriber count up 4× this year
1M-token context; leads SWE-bench Pro; 67% win rate on blind code-quality reviews
Claude Code Code Review GALaunched
Exited research preview and went GA Oct 15 for Claude Team and Enterprise organizations: review speed ~3× faster (20 minutes → 6–10 minutes), with new Quick / Standard / Deep review-depth tiers
Eligible Enterprise organizations can claim at least $10,000 in promotional credits before 2027-01-15; official blog + Unite.AI
Market positionPress Coverage
JetBrains survey (Aug 2026): most-used at work 31% vs Copilot 21%; Pragmatic Engineer survey: startup penetration 75% vs Copilot 9% (the 9% is "favorite" share)
ZDNet survey of 138: 76% use Claude Code daily vs 35% for Codex
Internal adoptionLaunched
Since May 2026, 80%+ of merged code written by Claude; engineers' daily merged lines 8× vs Aug 2024
Claude leads 26% of AI R&D work
Claude Code ModsLaunched
A TypeScript plugin mechanism launched Oct 1: rewrite prompts, intercept/retry tool calls, replace built-in features (official features like /diff now ship as Mods); Team / Enterprise get the preloaded sec-default security Mod first
Officially stated Mods do not run in a sandbox — trust the source; v2.1.288 also fixed a dangerous rm-command bypass
Claude Code time-to-first-token optimizationLaunched
Announced Oct 17: local sessions 4.07s → 1.33s (~3×), cloud sessions 16.99s → 6s (~2.8×), web / Cowork 3.36s → 2.17s, API 1.53s → 1.01s
Achieved via conversation-data compression + pre-started cloud sessions; announced by the official @ClaudeDevs account, confirmed by two tech outlets
Claude Code v2.1.321Launched
Released Nov 9: Fable 5.2 becomes the default Fable model; remote-control gains a --settings file parameter; fixes MCP OAuth refresh-token redirect leaks and other sandbox/permission vulnerabilities; Opus 4.8 fast mode to be removed Nov 30
Havoptic version tracker, single source; no direct official release notes found
Claude Code v2.1.322Launched
Released Nov 10: fast mode extended to Google Vertex AI (global / us / eu regions, falling back to standard speed on refusal)
Version-tracker reporting

5 · Enterprise & distribution

Adoption overtakes OpenAI for the first time
Barclays strategic partnershipLaunched
Expanded Oct 1: target 50% of software developers using Claude Code by end of 2026, most software engineers by 2027
Barclays Global Markets processes ~120K client emails daily with Claude; Colleague Knowledge Assistant has 16K+ UK employees using it, 1M+ searches so far
JPMorgan expands Claude Code deploymentLaunched
Announced Nov 9 by Anthropic: JPMorgan extends Claude Code from engineering teams to product development, infrastructure operations, risk assessment, and asset & wealth management, running inside the containerized Devspace platform; agents have no direct credentials / system access, and production changes require human-approval gates
Anthropic official article
FIS × Anthropic anti-money-laundering agentPress Coverage
The AML-investigation agent built with FIS went into full commercial availability for US banks Oct 15: it pulls together core banking / payments / data systems to assist AML investigations; design partner Amalgamated Bank is deployed; previewed in May as a "financial crime AI agent"
PYMNTS citing the company press release (single source, not cross-verified)
Ramp AI IndexPress Coverage
April 2026 enterprise adoption 34.4%, first time ahead of OpenAI (32.3%)
Enterprise LLM market share 32% vs OpenAI 25%
Cloud & capitalLaunched
Amazon has invested $8B in total, Google $2B
The foundation of three-cloud distribution
Developer platform joins Google Cloud billing & IAMLaunched
Oct 22: Claude API + Agent tools can be used through Google Cloud credentials / billing / existing commitments, operated by Anthropic with data processed by Anthropic
Material availability fact; runtimewire
Compute procurementPress Coverage
Pays SpaceX $1.25B/month for compute through May 2029
Press-reported; awaiting official confirmation
Microsoft compute commitmentLaunched
Disclosed Nov 12: Microsoft commits up to $4.3B in additional compute
IPO-related disclosure; reported by aiagentupdate and opentools on 11-12
Claude Frontier AcademyLaunched
Launched Oct 2 as an enterprise-engineer training program: $100M committed, training 10,000 Frontier Deployed Engineers by end of 2027; Accenture / Bain / Capgemini / Deloitte / McKinsey / Morgan Stanley / Novo Nordisk among the first to join; classes in San Francisco, New York, and London
Three-day intensive + exam earns the Claude Resident Engineer badge, then 12 weeks of on-site deployment work; Anthropic calls it "a depth of investment no AI company has made"
Claude for Government formally GALaunched
Exited July public beta and is now formally open to US federal and state agencies; runs in a FedRAMP High authorized environment; prepaid-allowance billing with hard spending caps, department admins can set per-user-group spending and model limits, quota alerts, audit logs, and two-person approval for sensitive operations
New capabilities ship at the same pace as the commercial version; session records stay on agency-managed devices; TechRepublic and others reported (Oct 2)
Genesis Mission three-year $150M partnershipLaunched
Pledged Oct 8 at the White House OSTP "Science: A New Golden Age" summit: $150M over three years supporting the federal Genesis Mission, giving 15+ federal agencies (NASA, NIH, NSF, etc.) and hundreds of research projects Claude, Claude Code, API credits, and technical support, focused on fusion energy, quantum computing, and other priority areas
Anthropic official Newsroom announcement (Oct 8); White House says 11 industry partners pledged $2.4B in total (NVIDIA $1B, AMD $500M, OpenAI $200M, Anthropic and Google $150M each)
HCLTech Global Premier PartnershipLaunched
Announced Oct 12: a dedicated Anthropic business unit and Center of Excellence (CoE), deploying Claude for enterprises via Frontier Deployed Engineers in a secure sandbox; HCLSoftware, as an Anthropic ISV partner, embeds Claude and Claude Code into BigFix, AppScan, and other enterprise products (cybersecurity / automation / business-process apps)
Service teams build industry solutions on Claude, Cowork, and Claude Code; hcltech.com official announcement + Morningstar / PR Newswire + unite.ai cross-confirmed
Workday × Anthropic new partnershipLaunched
Announced Oct 13 at Workday Rising 2026: Claude's reasoning connects with Sana AI agents and Workday's enterprise HR / finance data and business rules; users can start work in either Sana or Claude, with both sides exchanging information and tasks via MCP; data stays inside Workday
Announced via official Workday and Anthropic PR
Singapore office + MDDI AI safety MOULaunched
Announced Oct 14: Singapore office opens (fifth in Asia-Pacific); AI safety memorandum of understanding signed with Singapore's Ministry of Digital Development and Information (MDDI), covering AI safety, model evaluation, testing standards, incident response, and adoption research
Signed Oct 13 during Singapore International Cyber Week by co-founder Ben Mann and MDDI Permanent Secretary Chng Kai Fong, witnessed by Minister Josephine Teo; Unite.AI, Tech in Asia, The Darwin cross-reported
EverWind × Canada's first data centerLaunched
Lease signed Oct 16: Anthropic's first data center in Canada, at Point Tupper, Nova Scotia
EverWind is investing C$10B+ in parallel (incl. wind power); Reuters
Claude Security enterprise updateIn Preview
Entered public beta Oct 21: continuous scanning, fix tracking, per-project threat modeling, new APIs
$25,000/month free allowance through Dec 31; unite.ai
Claude Enterprise: Chat + Cowork unified GALaunched
Announced Oct 23 on Anthropic's official X: Chat and Cowork merge into one conversation inside Claude Enterprise, with cloud task execution folded into the same flow; web, desktop, and mobile; start in chat, switch to Anthropic-managed isolated cloud sandboxes when code or files need running, then return to chat for review; tasks continue in the cloud after closing the laptop and resume across web / desktop / mobile
Enterprise admins control it centrally; Oct 23 official X announcement; off by default for existing organizations, on by default from Nov 16 (admins can enable early or keep it off); zero-data-retention, HIPAA, and customer-managed-encryption-key organizations excluded; new Efficient mode limits the merged experience to Sonnet / Haiku to control cost; PYMNTS citing Anthropic's Oct 23 announcement
Associated Press paid licensing agreementLaunched
Reported Nov 5: Anthropic pays the Associated Press for real-time election data for Claude to show during the midterms; amount and term undisclosed; AP had already signed with OpenAI and Google
First paid publisher licensing deal; Anthropic had previously refused licensing on fair-use grounds — a reversal of position; Adweek first reported, Dealroom and Emergent confirmed Nov 5
TeraWulf compute lease backed by GoogleLaunched
Google signed three credit-support agreements Nov 4 backing Anthropic's 20-year data-center lease in Hawesville, Kentucky (401MW, total contract lease ~$19B); if Anthropic defaults or goes bankrupt, Google can take over the lease or pay a termination fee in exchange for corresponding equity in project subsidiary Raylan; backing takes effect in phases as data halls are delivered (end of 2027 to Q2 2028) and ends after Anthropic holds an investment-grade rating for three months
The lease itself was signed in July; CoinDesk, Reuters, The Inference reported Nov 5 (citing TeraWulf's Nov 4 8-K filing)
Claude Enterprise security suiteLaunched
Announced Nov 5: inference hooks now available — prompt hooks and pre-tool-call hooks send requests to the customer's own DLP server for adjudication before generation / tool calls (default 5-second timeout, allow/deny binary decisions), supporting Netskope, Palo Alto Networks, Proofpoint, Zscaler, or self-hosted servers; covers claude.ai, Cowork, Claude Code, Slack, Chrome, and more; Security Center (beta), centralized admin.claude.com (beta), and Cowork EDR (beta) still in preview; not available via Bedrock / Google Cloud access
Announced by Anthropic Nov 5; aiweekly

6 · Subscription tiers

Subscriptions are only ~15% of revenue
Personal subscriptionsLaunched
Free $0; Pro $20/month (5-hour sliding-window rate limit); Max 5x $100/month; Max 20x $200/month
The 50% usage bonus expired Sept 14, compensated with a +25% base quota
Team / EnterpriseLaunched
Team Standard $25/user/month (annual, 5+ users); Team Premium $100/user/month annual or $125 monthly (includes Claude Code); Enterprise custom
Feb 2026: restricted subscription OAuth tokens for third-party commercial use (cracking down on "token arbitrage")

7 · Hardware

Explicitly not doing
No hardware roadmapLaunched
Anthropic has never shipped hardware; no smart-glasses / phone roadmap
The only pure-software company of the six

TimelineMilestones

Radar zone · Unconfirmed (not in the main layers)

5-year, $200B cloud/chip procurement deal with Google(Market rumor, no official confirmation) · Exact Claude App MAU(No reliable public figure) · 2025 revenue and future cloud-spending figures(Reuters 09-28: 2025 revenue ~$4.6B, net loss ~$42B, future cloud/compute spending estimated at $518B; still internal information, not officially confirmed) · Hong Kong VPN-access accounts suspended(wisevoter 10-02, single source: says Moonshot AI sent 23M conversations to Claude via Singapore routing in May–July for distillation training; no official response) · FTC opens product-safety investigations into OpenAI and Anthropic simultaneously(paisakawach 10-01, single source, no mainstream coverage) · Claude 4 series retirement(Sonnet 4 / Opus 4 retired June 15, Opus 4.1 on Aug 5) · Claude for Startups expansion(Announced at SF Tech Week: eligible startups get one free year of Claude Team for up to 5 advanced seats + $1,000 API credit; TechCrunch 10-06, single source) · Claude Cowork defaults to cloud execution(groundtruth.day 10-05 article: new Cowork tasks for Pro/Max plans default to cloud execution from Oct 6; single source, needs official confirmation) · Hexaware becomes a preferred Claude partner(unite.ai 10-08, single source: Hexaware announces a multi-year partnership bringing Claude into its Zerovity and AgentVerse platforms; partner-side claim only) · IREN multi-year cloud contract(ad-hoc-news.de 10-12, single source: IREN announces a multi-year cloud contract with Anthropic, a compute-supply-side signal, not cross-verified) · Claude Fable 5.5 release rumored this week(Only X leaker @synthwavedd, with the only corroboration a new entry in Azure's litellm model-pricing JSON; no word from Anthropic, unconfirmed) · NTT Data × Anthropic Osaka data center(Reuters 10-13 via Nikkei: in talks, cost up to ~¥1T / $6.3B, power capacity up to 500MW, partial operation as early as 2028, target to sign within FY2026; both sides declined to comment) · Revised S-1: Q3 revenue $16.96B, compute commitments $895.4B(AI Briefing 11-03 summary, SEC original not checked line by line; the revised filing itself is confirmed, pricing range and share count still blank) · Claude Code background-agent status-update cost optimization(Only Anthropic's X, relayed by The Neuron: Haiku 5.5 now writes status updates from brief notes, cutting per-update cost ~99% and saving ~5% of session allowance; single source, not cross-verified) · Anthropic spends ~$3.4B to lock in 1.7GW of gas generators(FT 11-08 first report, Reuters relayed: in late September took over OnePWR's earlier 1.7GW heavy gas-generator order with Caterpillar, ~$3.4B including installation engineering capacity; no OnePWR acquisition, no staff transfer; neither Anthropic nor OnePWR responded, Reuters says it could not independently verify) · Apiiro report: Claude Code "stop asking" feature sent 2,539 keys into public GitHub(Report published 11-05, reported 11-08: scanned 104,185 public .claude/settings.local.json files, 1,122 contained real keys; counting historical commits, 2,539 keys, 1,994 repos, 63 organizations; as of 11-08 Anthropic had not publicly responded). · 1,000 Fable 5.2 agents formalize Fermat's Last Theorem in 30 hours(AI Weekly 11-09, single-source relay: Anthropic says 1,000 Fable 5.2 agents completed the formalization in 30 hours; the model card discloses the model detected it was being evaluated in 43% of evaluation transcripts; not independently verified). · IPO underwriters reportedly turning away mainland China / Hong Kong investors(Bloomberg, Nov 13, citing people familiar: several underwriters are refusing or largely denying allocations to mainland Chinese and Hong Kong institutional investors, citing the US International Traffic in Arms Regulations (ITAR); the prospectus itself does not state the restriction; single-source chain) · Claude Meetings beta(Single source: joins Google Meet meetings for live collaboration, for paid Team and Enterprise plans — Claude joins video meetings as a participant, answering questions live and co-writing project proposals and schedules; external meetings opt out by default, internal data queries need advance approval; not cross-verified) · Claude Opus 5.5 becomes the default model for all paid plans(Single source: with a new prompt guide; not cross-verified) · Anthropic cuts the vulnerability-disclosure window from 90 days to 30 days(for the most severe vulnerabilities; only a one-line mention in a SiliconNoon briefing on Nov 16, with no link to an original source, not cross-verified)

Verification: Model releases and pricing cross-verified against September coverage from multiple tech outlets; Claude Code and enterprise data cross-verified against AI-coding landscape reports (9-13), the JetBrains survey (Aug 2026 report), the Pragmatic Engineer survey, and the ZDNet survey; subscription pricing cross-verified against the official pricing page and two independent subscription guides; valuation/ARR: annualized revenue run rate ~$81B and Q3 revenue near $17B per FT 11-06; ~$65B at end-July is an earlier point on the same growth curve; the Barclays strategic partnership confirmed by the Oct 1 joint announcement and multiple outlets; IPO fundraising target $80–100B and valuation $1.25–1.75T per FT 11-06, timetable and valuation both reported figures, not finalized; Claude Frontier Academy cross-confirmed by unite.ai and aistockwire (launched 10-02); Claude Code Mods confirmed by the official release and multiple sources incl. aiweekly (10-01); Claude for Government GA reported by TechRepublic and others (10-02, FedRAMP High prepaid-allowance model); voice-data training authorization reported by BleepingComputer (10-04, in-product prompt, voluntary opt-in). Single-source figures downgraded or omitted; Claude for Google Workspace cross-reported by multiple secondary outlets the same day (10-06); CVP three-tier expansion per Reuters (10-06); the Claude for Startups expansion and Claude Cowork default cloud execution are single-source and have been moved to the radar zone. Claude Haiku 5.5 release and pricing cross-reported by Reuters and VentureBeat; Claude Dashboards / Motion and Docs / Slides / Design GA per Reuters (10-08); Genesis Mission commitment per Anthropic's official Newsroom announcement (10-08); OSS Scanner cross-reported by The Hacker News, Pondera AI, oossa, and others (released 10-08). HCLTech Global Premier Partnership confirmed by the hcltech.com official announcement + Morningstar / PR Newswire + unite.ai (announced 10-12); the Workday × Anthropic partnership announced via both parties' official PR (10-13, at Workday Rising 2026); new anti-distillation measures per Reuters (10-13). Singapore office + MDDI AI safety MOU cross-reported by Unite.AI, Tech in Asia, and The Darwin (10-14); Democracy Works election partnership per Reuters (10-14); OpenTelemetry trace export public beta cross-reported by Unite.AI and CryptoBriefing (10-14); Claude Code Code Review GA cross-reported by the official blog and Unite.AI (10-15); FIS × Anthropic AML agent commercial availability per PYMNTS citing the company press release (single source, not cross-verified); the first Canadian data-center lease per Reuters (10-16, EverWind investing C$10B+ in parallel incl. wind power); Claude Code time-to-first-token optimization per the official @ClaudeDevs account announcement (10-17) + two tech outlets' cross-confirmation; Claude Duet launch (10-21, agentic research preview, Reuters); Claude Security enterprise update enters public beta (10-21, $25,000/month free allowance through Dec 31); Claude API version 2026-10-21 (reported 10-22, panews.io); developer platform joins Google Cloud billing & IAM (10-22, runtimewire); Agent Skills donation to the Agentic AI Foundation confirmed by the official blog (10-23, donating the name / logo / GitHub organization / agentskills.io domain); Claude Enterprise Chat + Cowork unified GA per the official X announcement (10-23); default-on timetable (existing organizations switch on by default from Nov 16, zero-data-retention / HIPAA / customer-managed-key organizations excluded, Efficient mode limited to Sonnet / Haiku) per PYMNTS citing Anthropic's Oct 23 announcement; the rumored Oct 23 public S-1 did not materialize (as of 20:51 UTC Oct 23, EDGAR still had no public S-1). The UK AI biosecurity lab AIBL per the UK government's Oct 23 announcement and BBC reporting (government funds £100M; Anthropic contributes up to £100M in model credits, not cash). Message Threads public beta reported by Gadgets360 and The Decoder Nov 3 (citing the official blog), 35% is a vendor-measured figure; revised S-1 figures are AI Briefing's Nov 3 summary, SEC original not checked, and have been moved to the radar zone. FlashProve per Interesting Engineering Nov 4 (TechCrunch via Nogentech, Think Facility), code on GitHub since Oct 30. The Associated Press paid licensing agreement first reported by Adweek, confirmed by Dealroom and Emergent Nov 5; TeraWulf lease backing reported by CoinDesk, Reuters, and The Inference Nov 5 (citing TeraWulf's Nov 4 8-K filing). Claude personal finance Beta launch announced by Anthropic Nov 5 and cross-reported by multiple outlets. Claude Enterprise security suite announced by Anthropic Nov 5; Agent Memory research published on Anthropic's official blog Nov 6, cross-reported by multiple outlets; Claude Code v2.1.319 reported by releasebot Nov 6. Claude Sonnet 4.6 release and pricing announced on Anthropic's official blog Nov 8 (performance figures are Anthropic's own test numbers); Sonnet 4.5 retirement timetable per the official announcement; IPO roadshow and valuation expectations per FT, Bloomberg, and The Information Nov 6–8; the $3.4B generator order first reported by FT, relayed by Reuters, Anthropic not responding; the Apiiro key report per Cybersecurity News Nov 8. Fable 5.2 / Mythos 5.2 release and pricing cross-reported by Unite.AI, Vktr, and other outlets Nov 9; Claude Code v2.1.321 from the Havoptic version tracker, single source; JPMorgan's expanded Claude Code deployment per Anthropic's official article; IPO roadshow and listing target relayed by IFR / WSJ / Bloomberg Nov 9, valuation and timetable both reported figures, not finalized. IPO roadshow launch and the $70–$78/share pricing range per Neowin, The Wrap, SRN News and others on 11-10 (S-1/A second-amendment figures; listing timetable is press reporting, not finalized — take the official announcement as authoritative); TSMC's board approval of up to $500M in Anthropic common stock at the IPO price per the board resolution report; Claude Code v2.1.322 per version-tracker reporting; South Korea's PIPC investigation into the Claude enterprise-account leak per Korean media on 11-10 (Anthropic's Oct 31 notification). IPO pricing above $78 and a valuation above $1.5T per the FT on 11-11 (citing five people familiar); Anthropic has not confirmed, latest filing still $70–$78; the $26B private-placement split per Fool / aiagentupdate, citing the filing, with no Reuters / Bloomberg cross-confirmation seen. The IPO order book's first-day full coverage confirmed by CNBC on 11-12; the pricing lift is FT/Bloomberg reporting, not a filed fact; the Q3 operating loss of $567M and up to $4.3B in new Microsoft compute commitments per IPO-related disclosures and aiagentupdate on 11-12. The IPO's third-day 6× oversubscription per Reuters on Nov 13 (citing people familiar; Anthropic declined to comment, order counts may still change); Barron's financial figures as noted; Claude's two Nov 13 outages per the Anthropic status page; self-hosted Claude Managed Agents entering private preview per Anthropic's Nov 12 announcement (reported by aiagentupdate Nov 13). Claude's supported-region policy took effect Nov 12 (Taiwan is on the supported list); the RSP acceleration-threshold determination per the S-1/A No. 2 wording comparison (enjoy.dev, Nov 13, line by line), based on expectations for future models and the overall pace of progress, with no current model individually determined to cross the line; IPO underwriters reportedly turning away mainland China / Hong Kong investors per Bloomberg, Nov 13 (citing people familiar, single-source chain, restriction not stated in the prospectus), now in the radar zone; the S-1 financial details summarized Nov 14 come from SEC filings. The Claude Meetings beta and Claude Opus 5.5 becoming the default model for all paid plans are both single-source and now in the radar zone. Claude HTML-in-Canvas per The Decoder, Nov 16; IPO orders near 7× per Bloomberg, Nov 16, with pricing Nov 18 and listing Nov 19 unchanged; the 30-day vulnerability-disclosure window is only a one-line mention in a SiliconNoon briefing, filed in the radar zone. Linked Billing Beta and the third S-1/A cross-reported (SEC filing). The IPO's reported pricing of $83/share was first reported by Reuters on Nov 18 (US evening), cross-cited by Bloomberg, CNBC and TechCrunch; Anthropic declined to comment, and no Form 424B4 pricing filing had been seen by the window's close — reported pricing, pending official filing; the final share count and total raise await the official filing. The IPO first-day outcome is confirmed across multiple outlets (TechCrunch, The Desk Research, Fool, Invezz): priced at $83, opened at $111.50, closed at $97.17 (+17.07%), final offering of 980M shares raising $81.34B — the largest IPO ever — with Form 424B4 filed; the ~$900.9B infrastructure spending commitment through 2045 is disclosed in the prospectus. Last updated 2026-11-20. · Privacy · Terms · Contact
← Back to Collection