Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
IndustryAPI / Product · ActionableHabr AI→ original

OpenAI Raises $11B at a $1.25T Valuation Against a $30B Target at $1.4T

A Habr columnist is convinced that cheap money for AI labs is running out. OpenAI aimed to raise $30 billion at a $1.4 trillion company valuation. According to the WSJ, it raised $11 billion at $1.25 trillion after lukewarm investor interest. The takeaway for developers is simple: if your product depends on one lab's API, prices and terms can change, so you need a backup provider.

AI-processed from Habr AI; edited by Hamidun News
OpenAI Raises $11B at a $1.25T Valuation Against a $30B Target at $1.4T
Source: Habr AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceHabr AI ↗
Editor-in-Chief
Key metric #1In March OpenAI was valued at $852 billion; now it is $1.25 trillion
Anthropicquarterly revenue +46.9%, compute costs +77.4%
Anthropic's commitmentsat least $319.9 billion through 2030 and another $451.2 billion

Imagine your product runs entirely on the OpenAI or Anthropic API. If the vendor cannot raise money comfortably, sooner or later it will raise token prices, cut rate limits or change its terms, and your unit economics will start to slip. That is the subject of a column translated on Habr: the author argues that the era of cheap venture capital for AI labs is ending.

At its center is OpenAI's round. The company expected to raise $30 billion at a $1.4 trillion valuation, that is, at the stated price of the whole company.

According to the WSJ, it ended up raising $11 billion at $1.25 trillion "after lukewarm interest from large investors." The lower the valuation and the smaller the sum raised, the weaker the demand for the shares.

This is the author's opinion, not an established fact, but he takes his numbers from the WSJ, Bloomberg, The Information and Anthropic's public IPO filing.

OpenAI: the round came in smaller and cheaper than planned

As the author recounts it, on September 16 the New York Times wrote about a possible round at a $1.5 trillion valuation. Two weeks later Bloomberg reported a target of $30 billion at $1.4 trillion, and on October 5 talks with BlackRock and UAE funds for roughly $10 billion became public. The result was $11 billion at $1.25 trillion. OpenAI also wants to raise up to another $20 billion from family-owned businesses and wealthy private investors. The author concludes that sovereign wealth funds, venture investors and private credit investors have already given all they could.

For comparison, the valuation in March was $852 billion. According to The Information, the company spent $12 billion on research and development in the first quarter of 2026. Judging by Microsoft's earnings, it spent $10 billion on compute and payments in the last fiscal quarter.

In other words, an $11 billion round does not even cover a quarter of compute costs. Yet as of March 31 OpenAI held more than $73 billion in cash and securities. Revenue grew from $5.

7 billion in the first quarter to $6.7 billion in the second, and Amazon and SoftBank transferred a further $35 billion and $30 billion. The author does not understand why the company needed another $11 billion with that much money on hand: it is either desperation or an attempt to keep pace with Anthropic's IPO.

He also sees a problem in the fact that an IPO would have to come at a $1.5–2 trillion valuation for recent investors to make any money at all, while the company has $750 billion in cloud compute spending planned through 2030.

Anthropic: what the amended filing showed

Anthropic filed an amended S-1/A that already includes third-quarter data. The author is convinced the figures were deliberately left out of the first version. Quarterly revenue grew 46.

9% (from $11.5 billion to $17 billion), while compute costs nearly doubled, up 77.4% (from $5.

6 billion to $9.9 billion). Part of the spending is tied to its main customer, SpaceX: according to Bloomberg, it accounted for about 34.

5% of costs, or $5.028 billion. The quarterly loss was $567 million, with an operating margin of about −3%.

The author notes that such costs fit poorly with what OpenAI and other companies spend on renting GPUs and TPUs. He allows that Anthropic may have bought capacity in advance at discounts that are now expiring.

The main complaint concerns commitments. The filing lists minimum payments of $319.9 billion through the end of 2030, and that is not all: another $451.2 billion or so is tied to signed leases that have not yet started. They begin between 2026 and 2028 and run from 1 to 20 years, averaging about 10 years. Hence the author's headline about $895 billion in debt obligations. For comparison, the first version of the filing showed $430 billion in non-cancellable spending, including a $161 billion deal with Broadcom.

There is also credit risk. According to a Financial Times report, Blackstone warned investors that Anthropic could receive a junk credit rating right after the IPO, and pension and insurance funds cannot hold such paper. A $35 billion TPU leasing deal backed by Broadcom has already received a BB+ rating from KBRA, one notch below investment grade, even though $30 billion of it is guaranteed by a highly rated company. Anthropic's expected IPO valuation is now $1.4–1.5 trillion, down from $2 trillion a month ago, and according to the WSJ institutional investors are pushing it down to $1.5–1.6 trillion.

What to do in practice

Do not treat current token prices as a constant: build a price-increase scenario into your calculations. Put an abstraction layer over the provider and keep a set of regression tests, so you can switch to a backup model in days rather than months. Track token spending the way you track a cloud bill: per-team limits, alerts, caching. Keep the caveats in mind too: this is opinion journalism with a firm stance, and the forecast may not come true if the capital market opens up again.

Story

Key Questions Answered

What are the key details on OpenAI: the round came in smaller and cheaper than planned?

As the author recounts it, on September 16 the New York Times wrote about a possible round at a $1.5 trillion valuation. Two weeks later Bloomberg reported a target of $30 billion at $1.4 trillion, and on October 5 talks with BlackRock and UAE funds for roughly $10 billion became public. The result was $11 billion at $1.25 trillion. OpenAI also wants to raise up to another $20 billion from family-owned businesses and wealthy private investors.

What are the key details on Anthropic: what the amended filing showed?

Anthropic filed an amended S-1/A that already includes third-quarter data. The author is convinced the figures were deliberately left out of the first version. Quarterly revenue grew 46.9% (from $11.5 billion to $17 billion), while compute costs nearly doubled, up 77.4% (from $5.6 billion to $9.9 billion). Part of the spending is tied to its main customer, SpaceX: according to Bloomberg, it accounted for about 34.5% of costs, or $5.028 billion.

What to do in practice?

Do not treat current token prices as a constant: build a price-increase scenario into your calculations. Put an abstraction layer over the provider and keep a set of regression tests, so you can switch to a backup model in days rather than months. Track token spending the way you track a cloud bill: per-team limits, alerts, caching. Keep the caveats in mind too: this is opinion journalism with a firm stance, and the forecast may not come true if the capital market opens up again.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…