Due to a severe shortage of modern chips, the US National Security Agency (NSA) continues to use Anthropic models despite a Pentagon ban. There is currently no alternative for handling classified information: the capacity of existing data centers does not allow running advanced models on private hardware. This is reported by Ixbt.com reports .
The New York Times reports that the White House has urgently approved $9 billion for the construction of specialized data centers for intelligence and military agencies. These centers will be equipped with NVIDIA Grace Blackwell superchips, which require specialized power supply and liquid cooling systems. Until the infrastructure is ready, the NSA is forced to use Anthropic models despite the official ban.
Modern artificial intelligence systems are essential for analyzing massive amounts of data — satellite imagery, intercepted messages, and threats invisible to the human eye. The chip shortage is considered a threat to national security, and the US government is urgently reallocating $800 million from other budgets to purchase computing power.
The paradox of the situation is that Anthropic, due to its ownership structure and foreign investments, has been deemed a supply chain risk, yet it remains irreplaceable. The company's Claude model is considered one of the most advanced for threat analysis. For example, the Glasswing project developed by Anthropic identified over 10,000 critical vulnerabilities for Western intelligence services in one month.
This problem affects not only intelligence but the entire global market. Memory chip manufacturers like Samsung, SK Hynix, and Micron are shifting resources from consumer electronics to artificial intelligence. Anthropic's revenue growth from $9 billion to $30 billion in a year and its expected $800 billion IPO confirm the company's unique position: it is simultaneously under a ban and critically necessary for government structures.
Read “Zamin” on Telegram!
