🔥 Trending on HN

Anthropic's sales kept rising. Why did its growth look flatter?

2 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
quarter

A three-month period used to group business numbers.

annualized revenue

A one-year estimate based on the current pace.

Hacker News

A technology news site where readers score and discuss stories.

Anthropic, the AI company behind Claude, shared high-level third-quarter figures with investors. A Regenerator analysis examined reports about those figures. It found a striking split: revenue kept rising, but the pace of growth may have slowed sharply during the summer. That conclusion is an analysis, not proof that demand collapsed.

What happened

The reported figures were $4.8 billion in Q1, $11.6 billion in Q2, and $16.8 billion in Q3 2026. Q2 added $6.8 billion, or 141%, from the prior quarter. Q3 added $5.2 billion, or 45%. Revenue still rose. The new revenue added each quarter became smaller. That difference matters more than the huge headline number.

Why the monthly view changes the story

A quarter covers three months. Quarterly totals can hide when growth happened. The article used a separate report saying Anthropic had $65 billion in annualized revenue in July. That equals about $5.4 billion for July itself. If Q3 revenue reached $16.8 billion, August and September together produced about $11.4 billion. Their average was about $5.7 billion per month. That looks only slightly higher than July.

The article also estimated average monthly growth. It fell from about $760 million in Q2 to about $580 million in Q3. These monthly figures are back-of-the-envelope estimates. They are not official monthly accounts from Anthropic.

Why this matters

Investors care about future growth, not only current revenue. Future growth can affect how investors value a planned initial public offering. A short slowdown might not mean much. A lasting slowdown would mean something different.

The analysis also connects Anthropic's numbers with specialized cloud providers that rent AI computing capacity. Some of those companies borrowed heavily because they expected AI demand to keep rising. If demand grows more slowly, their equipment and debt become harder to support. The article presents this as a possible industry effect, not a proven outcome.

What is actually established

The source reports high-level figures shared with investors. The arithmetic comparing quarters is easy to check. The monthly pattern is an estimate based on reported numbers and a July annualized figure. The source itself says it uses Regenerator estimates and news reports.

The story received 210 points and 243 comments on Hacker News. That shows community attention. It does not prove the article is correct. It also does not reveal whether readers agreed.

What is still unknown

We do not know Anthropic's actual August revenue or September revenue from this article. We also do not know why growth may have slowed. The article lists several possibilities: August seasonality, a shortage of available computing capacity, slower AI spending, or cheaper models from China. It does not settle among them.

What to watch next

The next useful evidence would be official monthly figures or a later quarterly report. A September annualized revenue figure would help test the article's estimate. The next quarter would show whether the added revenue becomes larger again. Investors will also watch AI infrastructure spending and the debt carried by companies supporting that buildout. The main lesson is simple: enormous revenue growth and a possibly slower growth rate can exist at the same time.

💬 What the Hacker News comments say about Anthropic’s slowing growth

The thread does not reach one conclusion. Commenters explain the slowdown through release timing, improved model efficiency, cheaper competitors, and the assumptions behind paying back AI infrastructure investment. Financial, performance, and usage figures are commenter interpretations or user reports, not independently verified here.

  • One commenter argues that Opus 5.5 arrived on September 22, near the end of the quarter, so the article’s figures may reflect little of the usage increase that followed. This is a timing hypothesis, not proof.
  • A user report says an engineering team’s bill fell to one-third of its previous level after moving to Opus 5.5. If better models require fewer resources for the same work, usage and revenue may stop moving together.
  • In one commenter’s task-cost comparison, Opus 5.5 was about eight times more expensive than Sol 6.1. The commenter also argued that cheaper, self-hostable models such as DeepSeek 4.1 Flash may be good enough, so a coding lead alone may not guarantee long-term growth. The comparison conditions were not provided.
  • Another user report says DeepSeek 4.1 Flash can fail to recheck important assumptions, stop before finishing, and do too little independent exploration. The same report says it can perform those behaviors when explicitly prompted. This is an experience report, not a benchmark.
  • A commenter who said they read the S-1 reported $33.2 billion in revenue and $18.8 billion in compute costs for the first three quarters of 2026, concluding that inference is highly profitable and that the main need is more scale. Those figures and that interpretation come from the commenter.
  • The investment debate depends heavily on assumptions. One calculation says $900 billion of capex over a decade would require $360 billion in steady annual revenue at a 25% margin. A counter-calculation says that at a 2% margin with eight years remaining, the requirement would be $4.5 trillion annually.
  • A further rebuttal says market value cannot be compared directly with liabilities to determine solvency, citing Walmart and JPMorgan as examples. Large obligations alone do not prove insolvency, though this does not prove that Anthropic’s plan is feasible.

initial digest at 243 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

🔥 Trending on HN

Anthropic earned more, but its growth may be slowing

📰 Full story: Anthropic's sales kept rising. Why did its growth look flatter?

Anthropic's revenue is rising. The monthly pattern suggests its growth speed may have weakened.

1 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
quarter

A three-month period used to group business numbers.

annualized revenue

A one-year estimate based on the current pace.

Hacker News

A technology news site where readers score and discuss stories.

💡 The gist

  • Anthropic, the company that makes Claude, earned more money.
  • Its revenue increased more slowly from Q2 to Q3.
  • The monthly numbers are estimates, not confirmed reports.

Anthropic's revenue rose during the first three quarters of 2026. Q1 brought $4.8 billion. Q2 brought $11.6 billion. Q3 brought $16.8 billion.

From Q1 to Q2, revenue added $6.8 billion. That was a 141% increase. From Q2 to Q3, it added $5.2 billion. That was a 45% increase. Revenue grew each time. However, the added amount became smaller.

A quarter means three months. Quarterly numbers can hide changes inside those months. The article says Anthropic had $65 billion in annualized revenue during July. Annualized revenue means a one-year guess based on one month's pace. $65 billion divided by 12 months equals about $5.4 billion for July.

Q3 revenue was $16.8 billion. Subtracting July gives $11.4 billion for August and September together. Dividing by two gives an average of $5.7 billion. That suggests only modest growth after July. The article also estimates average monthly growth fell from $760 million in Q2 to $580 million in Q3.

These monthly figures are guesses. Anthropic did not publish confirmed monthly results in the article. The reason for the slowdown is also unclear. August may be seasonal. Anthropic may have lacked enough computing capacity. AI spending may also have become slower.

The story received 210 points and 243 comments on Hacker News. That measures attention, not truth. The next quarter should show whether growth speeds up again. The estimates come from the original analysis.

💬 Why Anthropic’s growth may look flat

Commenters disagree about the cause. Some point to the timing of the newest model. Others say better efficiency reduces customers’ bills, while cheaper rivals and huge investment obligations create pressure. The numbers and performance claims below come from commenters or users.

  • Opus 5.5 launched on September 22, near the end of a quarter. One commenter thinks the article may therefore miss much of its later usage growth.
  • A user report says an engineering bill fell to one-third after switching to Opus 5.5. A better model can make people more productive while making each task cheaper, so more use does not always mean more revenue.
  • One comparison claimed Opus 5.5 costs about eight times as much per task as Sol 6.1. Cheaper DeepSeek 4.1 Flash may be good enough for some jobs, but another user reported that it sometimes skips important checks or stops early.
  • A commenter’s reading of the S-1 gave $33.2 billion of revenue and $18.8 billion of compute costs for the first three quarters of 2026, and called inference profitable. Other commenters dispute how easily future costs can be covered.
  • The payback math changes with the assumptions: $900 billion of capex could require $360 billion of annual revenue at a 25% margin, but $4.5 trillion at a 2% margin with eight years left. Another commenter says liabilities versus market value alone do not prove insolvency.

initial digest at 243 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

🔥 Trending on HN

Anthropic sold more, but growth may be slowing

📰 Full story: Anthropic's sales kept rising. Why did its growth look flatter?

Anthropic's sales grew. The extra money may be getting smaller each time.

1 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
Anthropic

The company that makes Claude.

annualized revenue

A guess for one year using today's pace.

Hacker News

A website where readers score technology stories.

Anthropic, the company that makes Claude, sells AI help.

Sales means money from selling something. The first three-month period brought $4.8 billion. The second brought $11.6 billion. The third brought $16.8 billion.

Sales grew each time. But the extra money became smaller. $11.6 billion minus $4.8 billion equals $6.8 billion. $16.8 billion minus $11.6 billion equals $5.2 billion.

An article guessed the monthly pattern. July's annualized revenue was $65 billion. That means a one-year guess using July's pace. $65 billion divided by 12 equals about $5.4 billion monthly. $16.8 billion minus $5.4 billion equals $11.4 billion. $11.4 billion divided by two months equals $5.7 billion.

This guess makes later months look only a little bigger. It does not prove sales stopped growing. Maybe August was slow. Maybe Anthropic lacked enough computer machines. Maybe people chose cheaper AI.

Hacker News is a technology news website. The story got 210 points and 243 comments there. That shows many people noticed the story. It does not prove the story is true.

💬 The simple Anthropic argument

People in the comments disagree about whether Anthropic’s growth really stopped.

  • The new Opus 5.5 came out at the end of a quarter, so its popularity may not be in the numbers yet. But one user said their work bill fell to one-third because the new model was more efficient.
  • Cheap AI models are competing too. One commenter said Opus 5.5 was about eight times more expensive than Sol 6.1. Another user said a cheap DeepSeek model sometimes forgot important checks or stopped too soon.
  • The money math is also disputed. A commenter reported $33.2 billion in revenue and $18.8 billion in compute costs. Depending on whether the profit margin is 25% or 2%, the needed yearly revenue could be $360 billion or $4.5 trillion.

initial digest at 243 comments (revision 1). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

Sources