Breaking: Nvidia Controls AI Supply While Anthropic Preps Trillion Dollar IPO

Anthropic is preparing to take its Claude models public in an initial public offering that projects 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 for investors and independent observers is how a company losing that much money can command such a high valuation. The answer lies heavily in the compute commitments Anthropic has secured from partners like Amazon and Google, which total up to 900 billion dollars. These commitments drive the narrative of future dominance, but they also highlight the massive infrastructure costs required to sustain the AI race.

The economics of the graphics processing units that power these models reveal a tight grip by Nvidia on the supply chain. A single Nvidia RTX Pro 6000 card, which retails at 20,000 dollars, can support specific user loads under constant use. The hardware can handle about 15 to 20 light users running at 60 to 70 tokens per second, or 1 to 2 heavy coding users running Claude Code at 40 to 50 tokens per second. While buying these cards directly from Nvidia at a wholesale price of 3,000 to 4,000 dollars reduces the acquisition cost, the breakeven point remains steep. For a heavy user paying 200 dollars a month, the GPU alone takes about 10 months to pay back. For light users paying about 20 dollars a month, the timeline is similar. This calculation does not even include the servers, switches, CPUs, data center facility costs, or electricity.

Nvidia leverages this bottleneck to drive demand up and lower supply, effectively giving the cards to the large players like Anthropic, Grok, and OpenAI. The consumer market is left to fit the bill, often facing higher prices and limited access to the latest hardware. This strategy ensures that the major AI labs maintain their lead while smaller entities struggle to compete. The GPU economics are not just a technical hurdle but a strategic tool that shapes the entire market landscape. By controlling the flow of compute resources, Nvidia influences which models can be trained and deployed at scale, leaving the broader ecosystem dependent on the decisions of a few key players.

This control extends to the distribution of models, as seen in Nvidia's agreement to acquire Hugging Face for 12.9 billion dollars. Hugging Face is the platform where most open AI models are distributed, and its peak annual revenue of 150 to 200 million dollars makes the acquisition price about 70 times its revenue. The stated reason for the deal was CUDA, the software that runs these models, but the deeper motive was control over the open source market. For Nvidia and its CEO Jensen Huang, the cost was pocket change, but the strategic gain was significant. This move allowed Nvidia to influence which models reached developers and users, particularly in a market flooded with advanced Chinese models.

The open source landscape has been shaped by the rise of Chinese models like Kimi from Moonshot AI, DeepSeek, Qwen from Alibaba, and models from Xiaomi. These companies are willing to give their models to the public to increase innovation. Kimi K3 started to outperform the Opus 5 models from Anthropic and was made available to the public. The internet community quantized these large language models, making them usable on smaller machines like Mac Studios, A100 cards, and RTX Nvidia models. This democratization of power threatened the proprietary models of major labs, leading to a fight over who controls the distribution and evolution of AI technology.

After Nvidia acquired Hugging Face, the platform began removing advanced models from distribution. Anything over a quant 4, which meant models capable of beating Claude's offerings, was taken down. The 4 bit, 6 bit, 8 bit, and 16 bit full sized models were banned from the Hugging Face platform. Developers on Discord rallied against what they saw as censorship and corporate overreach. They opened Pirate Face, a new place to upload and download all models, but not before the large 400 gigabyte models at quant 3 and higher were scrubbed from the network. This episode highlighted the tension between open innovation and corporate control in the AI sector.

The fight over control has also reached the halls of Congress. Anthropic demanded that the government ban Chinese models, citing national security, though the real goal was to control the market. The company recently engaged in a publicity stunt, releasing media about AI taking over the world when disgruntled employee Jacob Coxon quit amidst his disappointment in the technology. This panic helped push Congress to pass a bill banning any government and its affiliates from using any Chinese models. Anthropic is now trying to pass additional bills to stop other companies like Verizon from launching their own AI models in house using advanced open source models. The legislative push underscores the broader battle for dominance in the AI space.

The saga of Fable 5 illustrates the extent of this control. Anthropic released Fable 5 and Mythos 5 on June 9, 2026. Just three days later, on June 12, 2026, the U.S. Department of Commerce issued an export control directive that forced Anthropic to suspend access to these models worldwide. The models were solving too many problems, leading Anthropic to pull them off the shelf. This action was less about winning the AI race and more about controlling it. The government's intervention showed that even a leading private company must navigate complex regulatory landscapes when its technology advances rapidly.