US Intelligence Agencies Secure $9 Billion for AI Chips and Access to Anthropic’s Mythos Model

Washington — In a significant boost to America’s espionage capabilities, the White House has approved a secret $9 billion funding request for the National Security Agency (NSA) and Central Intelligence Agency (CIA) to acquire advanced computing hardware and infrastructure. The investment aims to close a critical gap in artificial intelligence deployment caused by a severe shortage of cutting-edge chips.

The funding, which still requires congressional approval, will primarily support the purchase of Nvidia’s Grace Blackwell superchips. These powerful accelerators demand specialized data centers with massive electrical capacity and advanced liquid cooling systems — infrastructure that current classified government facilities largely lack.

Building Isolated AI Infrastructure

US Intelligence Agencies Secure Billion for AI Chips and Access to Anthropic’s Mythos Model Intelligence officials have expressed growing alarm that the chip shortage is hampering the agencies’ ability to fully test and deploy the latest generative AI models for top-secret operations.

AI tools have already proven invaluable for sifting through vast troves of intelligence data, identifying overlooked communications, and enhancing cybersecurity efforts.

The $9 billion allocation represents a dedicated push to create isolated, high-security AI infrastructure for the spy agencies, bypassing some of the slower procurement channels traditionally managed through the Pentagon.
Administration officials are also reprogramming an additional $800 million for faster acquisition of computing capacity, with experts warning that even larger investments will be necessary in the coming years.

Access to Anthropic’s Unreleased Mythos Model

US Intelligence Agencies Secure Billion for AI Chips and Access to Anthropic’s Mythos Model To address immediate needs amid the hardware constraints, the White House has authorized the NSA to continue using Anthropic’s advanced AI model, known as Mythos. This comes despite the Pentagon’s earlier designation of Anthropic as a supply chain risk.

Mythos, a cybersecurity-focused model capable of identifying thousands of vulnerabilities in operating systems, browsers, and software libraries, offers unique advantages. Its technical requirements allow it to run efficiently on previous-generation servers, providing a bridge until new Grace Blackwell-compatible data centers are completed.

Negotiations between the government and Anthropic have been contentious. The Department of Defense had previously pushed for broad rights to use the company’s algorithms for any lawful purpose. Anthropic resisted, citing concerns over potential misuse. In the finalized classified contract, that expansive clause was removed. Instead, the agreement includes explicit prohibitions against using Mythos for mass data analysis or surveillance of U.S. citizens.

This compromise is expected to serve as a legal template for future contracts between U.S. government agencies and leading AI developers, balancing national security needs with ethical and legal safeguards.

US Intelligence Agencies Secure Billion for AI Chips and Access to Anthropic’s Mythos Model Also read:

Strategic Implications

US Intelligence Agencies Secure Billion for AI Chips and Access to Anthropic’s Mythos Model The moves underscore the intensifying global AI arms race. U.S. intelligence agencies are racing to match the computational power available to commercial leaders and potential adversaries. By securing both hardware and privileged access to frontier models like Mythos, the NSA and CIA aim to maintain a technological edge in intelligence analysis, cyber defense, and offensive capabilities — all while operating within tightly controlled, classified environments.

As AI becomes central to modern espionage, this $9 billion investment signals a long-term commitment to treating advanced computing infrastructure as critical national security architecture. Congress is now expected to review the request as part of broader budget deliberations.

This article draws on reporting from The New York Times and related sources.