Meta’s artificial intelligence ambitions are running headlong into a problem that money alone can’t solve: nobody inside the company seems to agree on what the plan actually is.

Executives and employees alike are grappling with a chaotic AI strategy, according to sources and internal discussions reviewed by WIRED. The confusion spans everything from workforce restructuring to a fundamental philosophical question about whether Meta’s AI models should remain open-source or shift toward proprietary, closed systems.

The great reshuffling

Meta is in the middle of moving roughly 7,000 employees into AI-focused roles as part of an aggressive restructuring effort. The reallocation is happening alongside planned management cuts and layoffs. The company that once bet its future on the metaverse, even renaming itself in the process, is now redirecting that energy toward AI.

Part of the restructuring involves Meta’s superintelligence lab, which recently produced its first model called Muse Spark. That model is slated for deployment across Instagram and WhatsApp, suggesting Meta wants its AI capabilities embedded directly into products that billions of people already use daily.

Open source or closed doors

The most consequential debate inside Meta right now centers on the future of its Llama model series. Following the release of Llama 3 in 2024, internal discussions about transitioning away from open-source models toward closed systems have intensified considerably.

This matters far beyond Meta’s campus. The company’s open-source approach to AI has been one of its strongest differentiators against competitors like OpenAI and Google, both of which keep their most powerful models behind API paywalls. Llama became a cornerstone of the broader open-source AI ecosystem, with developers, startups, and researchers building on top of it.

Meanwhile, Meta’s AI capital expenditures are projected to keep climbing, with spending expected to exceed 2025 levels on a year-over-year basis.

Why crypto should be paying attention

Meta’s potential move away from open-source AI development feeds directly into one of crypto’s most compelling narratives: the case for decentralized alternatives to corporate-controlled technology. If the largest open-source AI contributor in Big Tech starts closing its doors, the argument for decentralized compute networks becomes significantly more concrete.

Projects like Bittensor, whose TAO token underpins a decentralized machine learning network, and Fetch.ai, which trades under the FET ticker and focuses on autonomous AI agents, sit at the intersection of these two trends.

Meta also has a complicated history with crypto itself. The company previously explored stablecoin implementation through its Libra project, later renamed Diem, before regulatory pressure forced it to abandon the effort entirely.