The cost of scaling Claude 4

Anthropic faces immense pressure to sustain its Claude 4 rollout because training requires massive capital and compute resources. The company committed more than $100 billion to AWS technologies over the next 10 years. The updated Amazon deal provides $5 billion now, with up to $20 billion more tied to "certain commercial milestones." Amazon previously provided $8 billion in investments. Anthropic reported an operating loss of $8.06 billion in 2025 and a net loss of $42 billion for the same period. This net loss included a $34 billion non-cash charge tied to financing that can turn into shares. The company earns approximately $115 million per month. To increase this, Anthropic signed a five-year, $100 million deal to sell AI services to Databricks customers. Anthropic expects to grow revenue to $34.5 billion by 2027. This goal follows a $65 billion Series H round in May 2026 that valued the company at $965 billion. Strategic investors in that round included memory chipmakers Micron, Samsung, and SK hynix. Anthropic also uses a $15 billion revolving credit facility to manage expenses. Anthropic runs its large language models on AWS Trainium and Inferentia chips to build and deploy foundation models. This partnership includes adding five gigawatts of capacity to develop its models. While Anthropic seeks $34.5 billion in revenue, its competitor OpenAI expects to pull in $12.7 billion this year. OpenAI is also finalizing a $40 billion funding round led by SoftBank. OpenAI expects to burn $25 billion in cash in 2026 against a revenue target of $30 billion. Anthropic’s revenue is growing quickly, and the company has nearly closed the gap with OpenAI despite being founded five years later.

Regulatory scrutiny and market control

The UK Competition and Markets Authority cleared the $4 billion Amazon partnership after deciding it lacked jurisdiction because Anthropic earns less than £70 million in the UK and holds less than a quarter of the market. The regulator investigated whether the agreement resulted in Amazon having material influence over Anthropic. However, the agency still examines the partnership between Anthropic and Google. The US Federal Trade Commission also investigates whether large investments help tech giants gain an unfair advantage. This scrutiny follows a pattern where regulators examine how big tech firms use partnerships to consolidate power. Amazon claims its investment helps spur competition, yet the agency continues to monitor the interconnected web of over 90 partnerships in the sector. The agency previously investigated Microsoft’s $13 billion investment in OpenAI. The regulatory focus centers on the "GAMMA" firms, including Google, Amazon, Microsoft, Meta, and Apple, which control the foundational model value chain. Concerns exist that these large investments could see companies exert control over the direction of AI firms without attracting the regulatory scrutiny that a full acquisition might. The agency is also considering whether to deepen its probe into the partnership between Anthropic and Alphabet. Anthropic stated that Amazon does not have a seat on its board and carries no board observer rights.

Internal integrity and safety debates

Internal tensions regarding AI safety and company values create operational risks. A researcher resigned on February 9th, stating the organization faces constant pressure to set aside what matters most. This former employee worked on defenses against AI-assisted bioterrorism and studied how AI assistants could make humans less human. The researcher also worked on understanding AI sycophancy and creating internal transparency mechanisms. George Arison, the chief executive of Grindr, suggested that Anthropic makes safety statements to justify its high valuations. Clement Delangue, the chief executive of Hugging Face, dismissed the extinction risk concerns as being out of perspective. Jensen Huang, the boss of Nvidia, also dismissed these concerns during a conference. You should consider how these internal struggles impact models like Claude Opus 5.5, which leads the Arena creative-writing board at 1519.1, or Claude Fable 5.1, which holds the writing crown. Claude Fable 5.1 also tops the LiveBench Language board with a score of 90.7. Claude Sonnet 5.5 remains the free-tier value pick. The researcher also wrote one of the first AI safety cases and arrived in San Francisco two years ago after completing a PhD. Will the organization keep its focus on reliability while competing for market share?