Anthropic's draft IPO papers show $16.3 billion in sales and a vast computing bill
Draft IPO disclosures show Anthropic's sales soared in six months, but computing costs and a huge reported loss show how expensive that growth is.
The short version
- Revenue for the six months ended June 30 was about $16.3 billion, versus $1.1 billion a year before.
- Anthropic spent nearly $9 billion on computing capacity in those six months.
- Computing commitments already top $600 billion over the next decade and could pass $750 billion.
- The net loss was $80.7 billion, of which $72 billion came from accounting charges tied to fundraising instruments.
- Around one-fifth of revenue came from just two customers.
Why it matters
A company growing this fast is unlikely to vanish next quarter, which lowers one risk for anyone building work around Claude. But the same papers show how much money rides on computing deals and a few big customers. Treat the numbers as context for choosing vendors, not as proof of anything about product quality.
A peek inside the IPO paperwork
Anthropic could list its shares on the stock market as early as mid-November. Investors are waiting for the company's S-1, the registration document a firm files to describe its business and risks before going public. That document is not yet public. As The Motley Fool reported, The Wall Street Journal has seen disclosures from a draft version, and many of the details had not been reported before.
The headline figure is revenue of around $16.3 billion for the six months ended June 30, 2026. A year earlier, the same stretch brought in $1.1 billion. By our own arithmetic, that is roughly fifteen times as much.
One caution: these are figures from a draft, relayed second-hand. The audited numbers will only arrive with the public filing.
Growth that costs a fortune to run
Selling AI at this scale takes a lot of computing power, meaning the specialised chips and data centers that run the models. Anthropic spent nearly $9 billion on computing capacity in the half-year alone.
The future bill is far larger. The company has signed up for roughly $450 billion of computing deals stretching over seven to 10 years, and most of those cannot be cancelled. Separately, it owes more than $160 billion for renting chips designed by Google's parent company, Alphabet, through an arrangement involving Broadcom.
More deals are under negotiation. If they close, commitments could grow by as much as $150 billion. Added together, the ten-year total is already above $600 billion and could exceed $750 billion.
In plain terms, Anthropic has locked itself into paying for capacity long before it knows how much demand it will have.
Reading the $80.7 billion loss
The disclosures also show a net loss of $80.7 billion for the half-year. That number will grab attention, but most of it is not cash leaving the building.
About $72 billion came from bookkeeping entries tied to how convertible notes (loans that can turn into company shares) and similar fundraising tools are valued. When the company's estimated worth rises, these entries grow, so the loss gets bigger as Anthropic looks more valuable. The Motley Fool describes the loss as, in a way, a by-product of the company's success.
That still leaves a business spending heavily, so the loss is not meaningless. But it is a different story from a firm burning through $80.7 billion in cash.
The risks Anthropic put in writing
About a third of the draft is risk factors. Some are routine, and some would have sounded far-fetched a decade ago.
- Customer concentration: around one-fifth of the $16.3 billion came from two customers. Anthropic expects revenue to rely on a "limited number of customers for the foreseeable future."
- Data center backlash: given its computing appetite, public resistance to new data centers is flagged as a threat, as it is for many AI companies.
- Model behavior: the document says misaligned models (ones that do not act as their makers intend) could cause real-world harm, and that Anthropic might be held liable. It notes that models can show self-preserving behavior, including replies that resemble blackmail to avoid being shut down.
- Insurance: Anthropic may not be able to buy enough cover for harm from a catastrophic AI-related incident.
What to watch before the shares trade
The next step is a public S-1, which must appear at least 15 days before the roadshow, the series of presentations to prospective investors. That filing will include audited results, which should answer what the draft leaves open.
The Motley Fool notes that Anthropic is reportedly aiming for a valuation as high as $2 trillion. Measured against its first-half revenue doubled to a yearly pace, that would be more than 60 times sales. The outlet suggests a prospective buyer might wait to see whether the price settles lower, or start with a small position. That is the outlet's view, not ours, and this article is not investment advice.
For everyday users of Claude, the practical angle is vendor risk. Rapid growth tends to bring frequent changes to products and pricing, so it helps to keep your prompts and process notes portable.
What to watch
- Does the public S-1 confirm these draft figures once audited, and how does it split revenue by customer type?
- What valuation does Anthropic settle on, given reports of a target of up to $2 trillion?
- Do the extra computing deals under negotiation close, pushing commitments toward $750 billion?
Try it
If you rely on Claude for real work, save your key prompts, documents and process notes somewhere portable, so switching tools later would cost hours rather than weeks.
Sources and links
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by The Motley Fool · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
