A look at Meta's AI strategy, which is a bet that open sourcing the tech will drive down competitors' prices and spread Meta's version of AI more broadly
Wall Street Journal : Threads: @jjackyliang . Bluesky: @marypcbuk.bsky.social . X: @hoofnagle , @canadakaz , @ylecun , and @ylecun Threads: Jacky Liang / @jjackyliang : no tech company ever “open sources” out of altruism, i'll tell you that Bluesky: Mary Branscombe / @marypcbuk.bsky.social : It's a interesting strategy but there are more ‘open source’ models than Meta' s and ones that are closer to actually being open source [embedded post] X: Chris Hoofnagle / @hoofnagle : “Behind the contrarian strategy is a bet that making Silicon Valley's hottest new technology free will drive down competitors' prices and spread Meta's version of AI more broadly, giving Zuckerberg more control over the way people interact with machines in the future.” That's a Kaz Nejatian / @canadakaz : Meta's commitment to open source projects needs to be applauded. Much of tech industry is standing on Meta's shoulders. Yann LeCun / @ylecun : Unsurprisingly, majority of tweeps want “open source AGI to win” (whatever you mean by “open source”, “AGI”, and “win"). Meta is a key contributor to Team Open Source. Yann LeCun / @ylecun : @spectate_or Apple is famously secretive, despite recent forays into publications and open sourcing. Amazon has not been very open. Google (G-Brain) opened up considerably after the creation of FAIR (they would have lost many top people otherwise) but has been clamming up recently. The
Context & Ripple Effects
Meta’s AI approach follows its unusual decision to release LLaMA rather than keep it solely proprietary, a move framed as using wider adoption to extend the company’s influence over AI development. The strategy was reinforced by work to make a subsequent LLaMA version available for commercial use.
The bet matters because Meta is treating model availability as a competitive lever: lowering the cost of access could make its technology a common base layer while supporting its broader AI-driven product goals, including the earlier effort to improve Reels and ad targeting with AI.
First-order effects
- Meta makes broad model distribution central to its AI positioning, seeking adoption and developer influence rather than relying only on restricted access to its technology.
- Developers and companies evaluating AI models gain a lower-cost alternative tied to Meta’s ecosystem, while closed-model providers face added price pressure if adoption grows.
Second-order effects
- Competitors may have to differentiate through model performance, hosted services, safety controls, or distribution rather than treating access to a model alone as the product.
- A larger installed base around Meta’s models can pull more tools, applications, and technical know-how toward its approach, building on the earlier release of LLaMA as an influence strategy.
Third-order effects
- If open model releases remain competitive, AI value may shift away from charging for basic model access and toward distribution, infrastructure, and product integration.
- The later debate over whether closed models create lock-in shows that model-access choices could become a durable fault line in AI competition, though broad adoption does not by itself guarantee a lasting open-model advantage.
The trend: AI companies are increasingly using model-access policy—open releases versus controlled access—as a strategic choice over where they capture value in the stack.