White House Approves $9B for Spy Agencies to Catch Up on AI

The White House has approved a secret $9 billion funding request for U.S. intelligence agencies to acquire cutting-edge computer chips needed to power advanced artificial intelligence models, according to current and former U.S. officials as reported by the New York Times. The investment — far larger than the $1 billion initially indicated in earlier summaries — aims to address a critical shortage of AI computing capacity at agencies including the CIA, NSA, and others. The administration is also reprogramming $800 million for more rapid acquisition of computing capacity, though the full $9 billion still requires congressional approval.

The AI Chip Shortage Crisis

New AI models require staggering amounts of computing power — more than many technology experts anticipated even a year or two ago. That has fueled concerns in the White House and Congress that a chip shortage is causing intelligence agencies to fall behind in testing and deploying tools for top-secret espionage work. The funding is intended in part to boost infrastructure supporting Nvidia’s Grace Blackwell superchip, which requires data centers with enormous electrical energy and specialized liquid cooling systems.

The chip shortfall has hampered the CIA, the NSA, and other agencies working on classified cloud networks from testing or using the newest versions of AI models requiring Nvidia’s superchip. Even with immediate funding, classified cloud networks run by Amazon Web Services would face significant delays before data centers could be built with Grace Blackwell chips. Amazon announced a $50 billion effort last year to upgrade its government cloud computing services.

“Our intelligence community needs the frontier — the best AI chips, models, systems, talent — on a timeline that matches the threat,” said Vinh Nguyen, the former chief data scientist at the NSA and a senior fellow on AI at the Council on Foreign Relations.

The Anthropic-NSA Workaround

To work around the chip shortage, White House Chief of Staff Susie Wiles authorized the NSA to continue using an advanced model made by Anthropic, even though the Pentagon has designated the company a supply chain threat, U.S. officials said. Anthropic and the government are finalizing a classified contract that would allow the NSA to maintain access to Anthropic’s products, including its new model known as Mythos.

Mythos, released in April 2026, is said to be exceptionally capable at finding and weaponizing cybersecurity bugs. It was initially shared with only a small number of government agencies, banks, and firms in the U.S. and Britain, and has still not been made available for wider release. The model runs more efficiently on new Grace Blackwell chips but can also run on previous-generation chips — a key factor enabling the NSA workaround.

Earlier this year, the Defense Department demanded the authority to employ Anthropic’s technology for “any lawful use,” setting off a fight between the two sides. The new NSA contract does not include that language. Instead, the contract will include a carve-out to ensure the AI model is not used on Americans’ data, according to officials who added that the White House wants the deal to serve as a model for other companies.

Pentagon vs. AI Companies

The broader tension between the Defense Department and AI companies has become a defining feature of the Trump administration’s approach to AI governance. The Pentagon reached deals with eight major tech companies — including SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, AWS, and Oracle — that agreed to the “any lawful use” standard. But Anthropic held out, leading to its designation as a supply chain threat.

Rep. Jim Himes (D-Conn.), the top Democrat on the House Intelligence Committee, has been a vocal advocate for ensuring spy agencies have early access to advanced AI models. Speaking at Politico’s Security Summit on May 12, Himes argued it would be “insane” for intelligence agencies not to have access to the most capable hacking tools, as reported by Nextgov/FCW.

“Across the government, we should be looking at these capabilities. We ought to be cultivating — not damaging — our relationship with the producer of this remarkable new technology,” Himes said, in a nod to the ongoing tensions between Anthropic and the Pentagon.

Himes also warned of potential long-term damage: “If this drags out, if [Defense Secretary] Pete Hegseth gets a bee in his bonnet about this and just decides to target because his ego is damaged … that will be a massive liability for United States national security.”

The Canceled AI Executive Order

The story unfolds against a backdrop of broader AI governance debates within the Trump administration. On May 21, 2026, the White House abruptly canceled a signing ceremony for a new AI executive order hours before it was scheduled. President Trump told reporters he “didn’t like aspects of it.”

The order was reportedly seeking to formalize a process to share AI models with the government — including national security agencies — before public release, in recognition of the potentially expansive threat to cybersecurity the models now pose. The order faced opposition from Elon Musk, Mark Zuckerberg, and former AI adviser David Sacks. Trump has expressed concern that regulation could undermine America’s lead over China in AI, Bloomberg reported.

This follows Trump’s revocation of the Biden-era AI safety executive order (EO 14110) upon returning to office in January 2025. The administration has also been debating whether the Commerce Department or the intelligence community should oversee evaluations of AI models — a turf battle that remains unresolved.

The Classified Cloud Infrastructure Challenge

U.S. intelligence agencies face a unique challenge: their classified networks are designed to never connect to the public internet and rely on isolated computer infrastructure. The data centers hosting classified cloud networks are physically separate from unclassified networks with stricter security protocols. Companies cannot quickly upgrade commercial data centers to serve classified government work.

The agencies have potential workarounds, including working on certain problems on their unclassified networks, which run on the same internet and commercial data centers the general public uses. But doing so could risk exposing classified information, former intelligence officials said.

OpenAI’s contract with the Pentagon does not currently include the NSA. Intelligence officials hope the Anthropic contract will pave the way for an agreement with OpenAI to give the spy agency access to its technology.

Implications and What’s Next

The $9 billion figure — far larger than the $1 billion initially reported — signals that the U.S. government views AI infrastructure as a national security asset comparable to nuclear technology or satellite systems. The funding requires congressional approval, setting up potential debates over intelligence spending.

In a sharply worded statement, the White House declined to discuss the chip shortfall or its efforts to address it. “Sensitive national security deliberations are conducted with the seriousness they demand — not leaked to reporters and repackaged through selectively sourced, unverified claims designed to drive headlines rather than truth,” said Steven Cheung, a White House spokesperson. “The fact is the United States is leading the world in technology and is well prepared to deal with a variety of issues that may arise.”

Looking ahead, several key questions remain: Will Congress approve the full $9 billion? Can the classified cloud infrastructure be upgraded quickly enough to meet the threat timeline? And how will the broader debate over AI governance — between Commerce, the intelligence community, and the Pentagon — be resolved?

What is clear is that the AI race has become a central national security priority for the United States, and the gap between ambition and infrastructure remains a critical challenge.