In an episode published on November 2, 2026, The Register’s Kettle podcast examined the 614-page initial public offering filing that Anthropic has made public (The Register). Systems editor Tobias Mann told host Brandon Vigliarolo that Anthropic’s operating losses were about 700 million dollars (The Register). The newsroom’s summary put the company’s compute commitments at nearly a trillion dollars (The Register).
Key Facts
| Item | Detail | Source |
|---|---|---|
| Filing length | 614 pages | The Register |
| Net revenues | 16.3 billion dollars, as the host recalled the reported figure | The Register |
| Operating losses | About 700 million dollars; the host cited 693 million dollars | The Register |
| Total losses | 80.7 billion dollars, described as an accounting adjustment | The Register |
| Compute commitments | Upwards of 800 billion dollars and counting, over about two decades | The Register |
Timeline
- 2026-11-02: The Register published its Kettle podcast episode on Anthropic’s IPO prospectus (The Register).
- 2026-11-02: The same episode was published on Apple Podcasts (Apple).
Details (as Reported)
Mann described the filing as a mix of risk mitigation that reads like a doomsday prophecy, as most S-1 filings do (The Register). He noted that, for the first time in the company’s history, revenues outpaced compute expenses two to one (The Register). Vigliarolo observed that day-to-day business losses of 693 million dollars were down from 1.2 billion dollars in an earlier comparison period (The Register).
On the 80.7 billion dollars in total losses, Mann believes a large part is the write-off of equity the company will have to pay off when it goes public (The Register). He explained that Anthropic’s business is cash heavy and that partners provided capital for compute either as debt or as repayment in equity (The Register).
The newsroom’s summary also states that much of the company’s computing power relies on customers who are also competitors (The Register). Senior reporter Tom Claburn pointed to inference costs, noting that inference has become cheaper per token over the past year (The Register).
What Is Not Yet Confirmed
- Mann said the filing does not specifically explain the 80.7 billion dollars in total losses (The Register).
- Claburn asked whether the company is aiming for a two trillion dollar valuation, and Mann replied that the number depends on whether the company can sustain its results (The Register).
- Claburn said the company has not demonstrated that the use case for AI is as wide at volume as it claims (The Register).