Meta is negotiating a deal to lease its AI data center computing power to Anthropic for as much as $10 billion over two years, a move that would turn one of Silicon Valley's biggest censors into a landlord for the AI models that will shape how Americans work, search, and speak online.

Anthropic approached Meta in June with the proposal, which would involve monthly payments and allow either party to exit early, according to International Business Times and The Verge, citing sources familiar with the discussions. Neither company would comment. The deal would mark a strategic shift for Meta, which has historically kept its infrastructure in-house.

The stakes are straightforward. The same platform that throttles conservative voices and deboosts dissenting content is now positioning itself as a utility provider for the next generation of AI. When Meta controls the pipes, it controls what flows through them. Working Americans who already face algorithmic suppression on Facebook and Instagram would now have their economic tools — search, productivity software, business analytics — running on infrastructure owned and operated by a company with a documented track record of political censorship.

The scale of the buildout is staggering. Meta CEO Mark Zuckerberg has said the company expects to spend up to $145 billion this year on AI and infrastructure, more than double the approximately $72 billion Meta spent last year, International Business Times reported. Anthropic plans to invest $50 billion building its own data centers, according to The Verge, but it is racing to lock up computing capacity wherever it can find it.

Anthropic already signed a three-year deal with SpaceX valued at roughly $45 billion — about $1.25 billion per month — for computing capacity, International Business Times reported. Separately, SpaceX is in talks to provide the Pentagon with billions of dollars in data center access for AI models, Reuters reported, putting Elon Musk's company in direct competition with neocloud firms like CoreWeave. Unlike Meta's censorship infrastructure, SpaceX's Pentagon discussions represent an American company providing critical capability to the U.S. military — a genuine national interest.

The pattern across Big Tech is clear: a handful of companies are buying up the computing infrastructure that will power the American economy. Meta, Microsoft, Google, and Amazon have collectively committed hundreds of billions of dollars to data centers and advanced semiconductors. Amazon and Google have both invested directly in Anthropic, meaning they hold equity stakes in a company they also supply with infrastructure. That is not a competitive market. That is vertical integration with a velvet rope.

Anthropic presents itself as the safety-conscious alternative to OpenAI. But its funding comes from the same Big Tech oligopoly that has spent years colluding with government agencies to shape online speech. When the companies that build the AI, fund the AI, and host the AI are all reading from the same playbook, the question is not whether the technology works — it is who gets locked out.

The Meta-Anthropic talks are early and could collapse. But the direction is unmistakable: AI infrastructure is being consolidated into fewer hands, and the people who already control the digital public square are making sure they will control the digital economy too. The open question is whether anyone in Washington will notice — or whether bipartisan silence will once again greet a consolidation of power that leaves working Americans paying the bill.