Elon Musk downplayed suggestions that he might sever Anthropic’s access to SpaceX servers, despite a reported compute agreement valued at approximately $40 billion that ties xAI’s infrastructure closely to his operations. This arrangement, extending through 2029, emphasizes the growing influence of AI infrastructure providers over the AI model developers themselves.
In recent months, Anthropic launched two advanced AI models, Claude Fable 5 and Claude Mythos 5, both unveiled in early June. Soon after, the U.S. government imposed an export-control order that temporarily restricted access for foreign nationals, citing national security, cybersecurity, and jailbreak risks. These controls were later lifted, allowing Fable 5 to become available globally again by July. Anthropic describes these models as sharing the same fundamental design, with Fable 5 incorporating enhanced safeguards for wider usage, while Mythos 5 was limited initially to select cybersecurity partners.
The economic stakes in AI's computational resources have surged sharply as Anthropic committed to purchasing 300 megawatts of compute capacity from xAI’s Colossus 1 data center in Tennessee. This commitment alone is valued at roughly $1.25 billion per month through mid-2029. Google has signed a comparable infrastructure rental agreement worth $920 million monthly, underscoring the concentrated and costly nature of advanced compute markets.
Anthropic is positioning its safety measures as a key component in addressing the risks inherent in AI deployment. The company collaborates with major cloud providers like AWS, Google Cloud, and Microsoft Foundry to offer its models, priced with transparent rates for input and output token processing. It also spearheads industry efforts for an AI jailbreak severity framework alongside Amazon, Microsoft, and Google, aiming for standardized safety protocols.
This dynamic unfolds against a backdrop where infrastructure providers wield critical control, with model developers dependent on these platforms that can enforce access restrictions due to regulatory, security, or commercial factors. The emerging landscape reveals infrastructure as a central battleground shaping AI’s future direction and accessibility.