Meta Won't Rely on Open-Source AI; Building Own Frontier Models

During Meta's second-quarter earnings call, CEO Mark Zuckerberg made explicit what had been implicit in the company's AI trajectory: Meta will not bet its future on open-source AI models alone. When asked whether the company could rely on third-party open-source alternatives, Zuckerberg was direct: "Right now, the open-source models are not as strong as the frontier models, so no is the basic answer." He elaborated that Meta believes it can develop superior models in-house and that depending on external technology creates unacceptable long-term strategic risk. "I think we're going to be able to do better work, and we think that there's some risk in that reliance," he said.

This position does not signal an abandonment of open-source work entirely. Zuckerberg said Meta has no plans to abandon its open-source approach and expects to release open-source models again "at some point soon," adding, "Like we've always said, we're not dogmatic about this." The nuance matters: Meta will continue participating in open-source development, but it will not subordinate its core AI strategy to it. This aligns with a broader industry consensus. Nvidia CEO Jensen Huang recently shared an industry letter signed by more than 20 companies and organizations, including Meta, backing open models as drivers of innovation, cybersecurity, and AI sovereignty — a position consistent with Zuckerberg's stance that open-source work has value, just not as a substitute for frontier capability.

The financial picture reinforces the commitment. Meta posted second-quarter revenue of $60.80 billion（約9.7兆円）, topping analyst estimates of $59.50 billion（約9.5兆円）, but adjusted earnings came in at $6.18 per share, below expectations of $7.13 per share. More tellingly, the company increased the lower end of its full-year capital expenditure guidance to $130 billion（約21兆円）–$145 billion（約23兆円）, up from $125 billion（約20兆円）–$145 billion（約23兆円）. For the third quarter, Meta expects revenue between $61 billion（約9.8兆円） and $64 billion（約10兆円）, below analyst estimates of $62.68 billion（約10兆円）. The earnings miss and lowered forward revenue guidance suggest margin pressure, yet Meta is not cutting AI spending — it is raising it. That trade-off signals how seriously the company takes the race for frontier models.

Zuckerberg's statement represents a recalibration of Meta's AI strategy, not an outright rejection of open-source models. While Meta has signed industry letters backing open models as drivers of innovation and cybersecurity — a position the CEO reiterated — the company's near-term focus is on proprietary frontier models. This stance mirrors Nvidia CEO Jensen Huang's broader industry view that the world needs "both frontier closed models and frontier open models." The tension is real: Meta wants the innovation and sovereignty benefits of open-source work but believes that for competitive and strategic reasons, it cannot afford to rely solely on alternatives built by others.

The company's raised capital expenditure guidance to $130 billion（約21兆円）–$145 billion（約23兆円） underscores this commitment. That spending increase comes against a backdrop of earnings that beat revenue expectations ($60.80 billion（約9.7兆円） versus $59.50 billion（約9.5兆円） estimate) but missed on profitability (adjusted earnings of $6.18 per share versus $7.13 expected). For investors and competitors, the message is clear: Meta is willing to prioritize AI capability investment over near-term margin expansion.

No comments yet. Be the first to share your thoughts!