Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

Anthropic's amended IPO filing shows $1.7bn quarterly personnel cost

Anthropic's amended IPO filing shows more than $1.7 billion in personnel expenses for July to September 2026, TechSpot reported, and a hypothetical $8.2 billion stock-based compensation charge if it had listed on 30 September.

Media & internetAnalysisGrace OkonkwoPublished: 4 November 20263 min readSources 24
Anthropic's amended IPO filing shows $1.7bn quarterly personnel cost

Anthropic's amended IPO filing makes its wage bill visible as it moves toward a listing. Some restricted stock units carry two conditions: a period of service and a liquidity event such as an IPO. The filing also shows what stock-based compensation would cost if the company had listed on 30 September 2026.

The stock award bill

Had Anthropic completed its IPO on 30 September 2026, it says it would have recorded an $8.2 billion stock-based compensation charge that day for restricted stock units. That charge would have exceeded the $4.86 billion in personnel expenses Anthropic reported for the first nine months of 2026. Anthropic expected to recognise another $24.1 billion in expense for those awards over a weighted-average period of about two years, according to its amended filing. Of the more than $1.7 billion personnel expense for July to September, $1.1 billion came from stock-based compensation. The rest covers salaries, bonuses, benefits and payroll taxes. Anthropic records the share value as an expense even though employees do not receive that amount in cash.

Payroll is a smaller slice than it was

In 2023, Anthropic spent more than twice as much on personnel as it brought in from customers. That share has fallen sharply.

By the first nine months of 2026, personnel costs reached $4.8 billion, or about 15% of the $33.2 billion in revenue Anthropic reported for the same period, TechSpot reported. Revenue reached $17 billion in the quarter to 30 September, up from $11.5 billion in the previous quarter, TechCrunch reported on 3 November. Compute took 58% of revenue in that quarter, up from 48% the quarter before, on a compute bill of $10 billion. That includes $5.0 billion of compute under short-term commitments, which the filing says currently costs significantly more than comparable compute bought under multi-year deals. TechCrunch reported that Anthropic's public deals include paying SpaceX $1.25 billion a month for compute, alongside multibillion-dollar deals with Nscale, Akamai and Volta. Compute margins worsened.

How the pay compares

Across 5,300 staff, the quarterly personnel expense works out at more than $322,000 per person, or about $107,300 per employee per month. Anthropic has grown its workforce fivefold since the end of 2024. It still has fewer than a tenth as many employees as Meta, which has around 67,000 people on its payroll.

Anthropic's gross profit margin was 85.4% in the first half of 2026, compared with 26.4% for Z.ai and 17.9% for MiniMax, according to Anthropic's S-1 and the two Hong Kong-listed companies' latest financial disclosures, SCMP reported.

What the filing says about costs

The amended filing says compute and infrastructure expenses will increase in absolute dollars and remain the largest operating expense for the foreseeable future. Anthropic is trying to lower those costs through a project it calls Anthropic Direct, but those efforts will not keep pace with demand, TechCrunch reported.

Anthropic has filed for what could be the biggest IPO ever, TechSpot reported.

Comments 0

Sources

24
  1. 01Anthropic IPO filing: $1.7B quarterly payroll, about $107K per employee monthlyEN
  2. 02Anthropic is still grappling with the economics of computeEN
  3. 03Why Chinese AI stars like Z.ai, MiniMax struggle to match Anthropic's profit marginEN
  4. 04Anthropic IPO warns its AI could blackmail people, become sentient, end humanityEN
  5. 05How China 'risk factors' loom over Anthropic's hotly anticipated IPO filingEN
  6. 06Wer spricht da? Anthropic stellt Identitätsnachweis für KI-Modelle vorEN
  7. 07Binance attackers used fake job interviews to pull off $6.3 billion crypto heistEN
  8. 08Meta Gets Record €53 Billion of Orders for Debut Euro BondsEN
  9. 09Binance Says Hackers Used Well-Known Tricks to Carry Out a $6.34 Billion HeistEN
  10. 10Binance co-CEO Richard Teng says hackers used fake job interviews, AI tool to steal $6.3 billion in crypto EN
  11. 11Meta seeks to raise up to $25 billion in debt sale as AI spending jumpsEN
  12. 12Anthropic's Message Threads cut time to first token by 35 percent for long conversationsEN
  13. 13$150 Billion More Paid by Americans for Gas This Year Than LastEN
  14. 14 Anthropic confirms new Claude Code limits to prevent game cloning, users report blocking reverse engineering tasks EN
  15. 15Binance says fake job interview led to $6.3 billion crypto heistEN
  16. 16Binance Reveals How Fake Job Interviews and AI Tool Abuse Led to $6.43 Billion TheftEN
  17. 17With Copyright Complaints, Even Imposters Can Silence Karakalpak ActivistsEN
  18. 18Barracuda Flaws Could Be Chained to Bypass Authorization and Gain Root AccessEN
  19. 19Anthropic VS code extension on Cursor backdoored for nearly 12 hoursEN
  20. 20DeepSeek unveils AI ‘super agent’ in coding race with OpenAI, AnthropicEN
  21. 21FBI Links North Korea's TraderTraitor to Record $6.43 Billion Binance Crypto HeistEN
  22. 22Hackers Actively Exploit Highly Critical Drupal Commerce Core Authorization Bypass FlawEN
  23. 23Binance Suffers Security Breach Affecting $6.3 Billion in Assets, Nearly $3 Billion FrozenEN
  24. 24ByteDance’s Zhang Yiming tops China rich list with US$99.7 billion as AI fuels fortunesEN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Grace Okonkwo

Grace Okonkwo

AI, models and technology

Grace Okonkwo covers AI, models and technology for FLASH24, working from primary sources such as model cards, API documentation and benchmark papers rather than vendor summaries. She checks training data provenance, evaluation conditions and reported scores against the underlying datasets before any figure reaches print. She interviews researchers and engineers directly, tracks release calendars from major labs, and compares successive model versions on the same tests. Her own self-hosting, home-network and documentation-reading habits feed straight into that desk, since she tests tools on her own hardware first. She does not publish benchmark claims without a reproducible method.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.