SEATTLE, March 10, 2026, 06:13 (UTC-07:00)
Amazon’s cloud division on Tuesday said it’s working with customers to shift defense work—specifically projects linked to Pentagon contracts—away from Anthropic’s Claude. Those jobs are getting migrated to other AI models available on AWS. According to an Amazon spokesperson, Claude remains an option for everything else that’s not Pentagon-related.
Washington’s clash over AI regulations is hitting Amazon’s cloud division. Last week, the Pentagon—referred to as the Department of War by the Trump administration—tagged Anthropic as a supply-chain risk, after the startup wouldn’t lift its safety controls on autonomous weapons or domestic surveillance. Now, contractors can’t use Anthropic’s tech for Pentagon projects. Anthropic responded with a lawsuit Monday to try to stop the ban.
That drops right in the thick of Amazon’s AI ambitions. AWS pulled in $128.7 billion in revenue and $45.6 billion in operating profit for 2025. Chief Executive Andy Jassy called out “seminal opportunities like AI” and pointed to plans for roughly $200 billion in companywide capital spending in 2026. Bedrock, Amazon’s toolkit for AI app builders, now includes over 20 managed models—sourced from Amazon itself, Anthropic, Google, OpenAI, and others—so users can swap between models without having to rewrite their code. Amazon
Amazon counts among Anthropic’s largest partners. Back in 2024, Anthropic revealed Amazon’s total investment would reach $8 billion, keeping AWS as Anthropic’s main cloud and training provider, and leaving Amazon as a minority shareholder.
Other companies didn’t waste time either. Microsoft’s legal team determined Claude could stay on its platforms for clients not tied to Pentagon contracts, even as federal agencies like State, Treasury, and Health and Human Services started shifting away from Anthropic in favor of competitors OpenAI and Google. OpenAI, meanwhile, confirmed it has its own Defense Department deal for deploying tech on classified government networks. CEO Sam Altman said that agreement would now clarify the AI won’t be “intentionally used for domestic surveillance” of Americans. Reuters
Anthropic CEO Dario Amodei maintains that the designation “has a narrow scope”—in his view, it’s limited to Claude when used specifically in Pentagon contracts, not broader government business with other customers. Reuters
But there’s risk for both Amazon and Anthropic as this dispute lingers. Dan Ives at Wedbush flagged a potential “ripple impact” for Anthropic’s enterprise business—he sees some customers possibly halting deployments, waiting for the courts to untangle things. In filings, Anthropic warned the court that the government’s move could shave off several billion dollars from its 2026 revenue, although the firm isn’t shutting the door on negotiations. The Pentagon, for its part, maintains that national defense calls can’t be dictated by a private firm. Reuters
Amazon shares edged higher by roughly 0.1% at 12:58 UTC, market data show, with little sign of a notable move.
