Breaking: The $20,000 Card and the 10 Month Payback in Anthropic's IPO

Anthropic is preparing for an initial public offering that it projects at a valuation between 2.5 trillion dollars and 3 trillion dollars. This figure stands in stark contrast to the company's current financial reality, which includes a deficit of about 500 billion dollars and annual revenue of less than 20 billion dollars. The core question facing investors and the public is how a company that loses 500 billion dollars can be valued at 2.5 to 3 trillion dollars. To understand this gap, one must look closely at the hardware that powers the service and the economics of the compute commitments that drive the cost structure.

The foundation of the AI infrastructure is the graphics processing unit, specifically the Nvidia RTX Pro 6000. This card retails at 20,000 dollars as of now. This price tag does not include the servers, switches, CPUs, and processors needed to run the machines, nor does it cover the data center facility costs or the electricity required to keep them running. For a business model that relies on selling access to these models, the hardware cost is a massive initial outlay that must be recovered through user subscriptions.

A single RTX Pro 6000 card has specific capacity limits when running at full utilization. It can hold about 15 to 20 light users running light compute at about 60 to 70 tokens per second. Alternatively, it can support 1 to 2 heavy coding users running Claude Code at about 40 to 50 tokens per second, assuming constant use 24 hours a day. These numbers represent the maximum theoretical throughput for a single piece of hardware under ideal conditions. The distinction between light and heavy usage is crucial for understanding the revenue potential of each card.

Heavy users on the max plan pay 200 dollars a month for their access. Light users pay about 20 dollars a month on average for a base plan. When a company buys cards from Nvidia direct at a wholesale price of 3,000 to 4,000 dollars, the math of the payback period becomes clear. The GPU alone takes about 10 months of a heavy 200 dollar a month user to pay back. It takes about the same amount of time to pay back the cost of the GPU through 15 to 20 light users at 20 dollars a month.

Most light users will not use the chat for more than a few hours a day. This behavior means the bandwidth number can stretch 10 times in practical scenarios, but the breakeven and acquisition cost stays huge. The compute and GPU bottleneck is a real problem for these companies, and Nvidia knows it. Nvidia drives demand up and lowers supply, giving all its supply to the large players like Anthropic, Grok, and OpenAI, leaving the consumer market to fit the bill.

Anthropic carries compute commitments to other companies, including Amazon and Google, of up to 900 billion dollars. This liability is a significant part of the company's financial structure. The gap between the projected valuation and the under 20 billion dollar annual revenue is the core of the question: how can a company that loses 500 billion dollars be worth 2.5 to 3 trillion dollars. The answer lies in the control of the compute resources and the distribution of the models.

Nvidia agreed to acquire Hugging Face, the platform where most open AI models are distributed, for 12.9 billion dollars. Hugging Face makes about 150 to 200 million dollars a year in its peak years, so the 12.9 billion dollar price is about 70 times its revenue. The stated reason was CUDA, the software that runs all these models, but the deeper reason was control of the distribution of models on the market, especially the open source market.

The open source market is being flooded by advanced Chinese models: Kimi from Moonshot AI, DeepSeek, Qwen from Alibaba, and models from Xiaomi. These companies are willing to give the models over to the public to increase innovation. Kimi K3 started to outperform the Opus 5 models from Anthropic, and it was given to the public. After Nvidia acquired Hugging Face, advanced models were deleted from distribution. Anything over a quant 4, meaning advanced and capable of beating out Claude's models, was removed.

Anthropic released Fable 5 in June 2026, and the government forced them to take it down. Fable 5 and Mythos 5 were released on June 9, 2026, and Anthropic suspended access to them worldwide on June 12, 2026, after the U.S. Department of Commerce issued an export control directive. The models were solving too many problems, and Anthropic pulled them off the shelf. This was not about winning the AI race; it was about controlling the AI race.

In recent months Anthropic has been suspending users with absurd reasons like abuse, potential fraud, and usage violations. The appeal process is useless because it is fully run by bots, and as soon as they kick you off they will not reactivate the account. Once Anthropic goes public, the suspensions will multiply 100 times or 1000 times because it will have deep pockets and access to unlimited funds. The end user will not be important to them. Anthropic does not make money off of daily users, and it sure as heck does not make money off of heavy users.

Dario Amodei, the CEO of Anthropic, said in an interview that some ideas are worth billions. He stated that Anthropic will likely have some sort of royalty or partnership program where, when a person or company finds the cure for cancer, some of that accreditation and funding will go to Anthropic. That should scare anyone who builds on top of Claude: the company that hosts your work may claim a share of the prize when your work becomes worth billions.