Meta Internally Projected Ten Billion Dollar Annual Spend on Rival AI Firm Anthropic

Meta planned to spend a huge amount of money on software from one of its biggest rivals. Mark Zuckerberg recently published a long essay criticizing closed artificial intelligence laboratories. He accused them of trying to concentrate power within a few large institutions. Despite this public attack, his engineers at Facebook and Instagram were quietly burning through software tokens provided by Anthropic.

Internal financial projections revealed that the social network expected to spend up to ten billion dollars in a single year on these outside services. To put that number into perspective, Anthropic estimated its total yearly revenue would reach 65 billion dollars. A ten billion dollar contract means a single client would account for a large portion of their total business.

The Reliance on Competitors

Engineers inside the company depend heavily on Claude Code. This programming assistant helps them write and debug software. Usage spiked so quickly early this year that employees started competing on internal leaderboards to see who could use the most tokens. Management eventually had to shut down the competition in June because costs were climbing too fast. They warned staff that software expenses would easily reach the billions.

This reliance goes beyond simple coding assistants. The company is currently testing an upcoming personal assistant product internally named Hatch. While developers test and refine Hatch, they are powering it with models from Anthropic. The plan is to swap those third-party models for proprietary ones before the official public release. Currently, developers are building a new internal model named Watermelon. They paused the training phase in July but recently restarted operations, pushing the expected launch past October.

The Silicon Valley Dynamic

This situation highlights a common trend across the technology sector. Companies are competing fiercely while simultaneously paying each other billions of dollars. Google and Amazon invested large sums into Anthropic while building their own proprietary systems. Microsoft invested heavily in OpenAI but now stresses that both entities operate independently. You can read more about how these corporate relationships shape product development as the technology sector matures.

Zuckerberg tried to pull talent away from his competitors over the summer. The company offered huge compensation packages to researchers at rival labs. Reports indicate those hiring efforts largely failed.

In response to the rising costs, internal teams launched a proprietary coding tool called Muse Code. Adoption among employees is climbing, which might eventually replace the need for outside vendors. As companies try to cut costs, these intertwined relationships will likely continue shifting.