Mark Zuckerberg has dropped Silicon Valley’s biggest AI bombshell of the year: a $14 billion investment to develop artificial superintelligence—what he calls “Meta’s defining project of the next decade.” With this announcement, Zuckerberg doesn’t just challenge OpenAI and Google DeepMind—he escalates a global race for dominance over digital cognition.
According to The New York Times, details the creation of a new “Superintelligence Lab” within Meta’s Menlo Park headquarters, independent from Facebook, Instagram, and WhatsApp operations. The lab will combine high-performance compute clusters—built with more than 350,000 Nvidia H100 GPUs—and a new generation of Meta’s open-source LLaMA models, aiming to develop Artificial General Intelligence (AGI) by late 2026.
“We’re building intelligence that can reason, plan, and even reflect—tools that will outthink humans,” Zuckerberg said during a livestreamed internal town hall on Tuesday, according to the, according to CNN.
From LLaMA to Leviathan: Meta’s leap toward general intelligence
According to CNBC, the new funding will be divided across data center expansions, custom chip development, and elite talent recruitment. Zuckerberg is personally overseeing the lab, which he says will be “responsible for building safe and scalable AGI for global benefit.”
Meta’s flagship AI model, LLaMA-3, has already disrupted the open-source space. But its successor, LLaMA-4, will reportedly be multi-modal, capable of vision, speech, and autonomous decision-making. Internal Meta memos reviewed by The Times reveal that this model is being trained on real-time global web data—triggering concerns among European regulators over data privacy compliance.
“We believe open AI is safe AI. If only one company controls intelligence, we all lose,” Yann LeCun, Meta’s Chief AI Scientist added.
Silicon Valley’s Faustian bargain?
Zuckerberg’s “superintelligence for all” mantra is already drawing backlash. Experts warn that by pushing to open-source AGI, Meta could risk enabling authoritarian regimes or extremist groups to exploit cognitive-level AI.
Former Google AI ethicist Timnit Gebru criticized the plan, writing on LinkedIn,
“Meta isn’t building AGI for humanity—they’re releasing tools too powerful, too fast. We need safeguards, not slogans.”
Others see the move as Zuckerberg’s bid for geopolitical influence. “This is about who controls the future of thinking,” said Dr. Rajat Malhotra, AI Policy Fellow at the Brookings Institution. “And Zuckerberg wants Meta to be that controller—before China, before Washington, before anyone.”
Global reactions: admiration, alarm, and calls for control
According to NYtimes, that Zuckerberg’s move comes at a time when governments are waking up to the power of AGI. The European Parliament called an emergency session in Brussels hours after Meta’s announcement, warning that “unilateral AGI development without state oversight poses existential risks.”
China, meanwhile, praised Meta’s announcement. In a rare statement of support, Beijing’s Ministry of Science and Technology called it “a healthy competition that demonstrates private-sector dynamism,” according to Global Times. This comes as China continues to ramp up its own AI development, including a sovereign LLM expected to debut later this year.
Back in the US, lawmakers across party lines expressed concern. Senator Elizabeth Warren (D-MA) posted on X:
“Zuckerberg is building the brain of tomorrow with no adult in the room. Congress must intervene now.”
Meanwhile, Donald Trump, speaking at a rally in Florida, mocked Zuckerberg’s AI obsession. “He can’t even fix Facebook. Now he wants to make robots smarter than people? Give me a break.”
Open-source AGI: Meta’s moonshot—or civilization’s cliff?
Perhaps the most controversial element of Meta’s strategy is its insistence on making AGI open source. Unlike OpenAI and Google, which closely guard their most advanced models, Meta says it will release LLaMA-4 and its successors publicly—with limited gating.
In an internal white paper reviewed by CNBC, Meta claims that decentralized AI models reduce single-point failure risk and foster innovation. But critics say the company is underestimating the dangers.
“Open-sourcing superintelligence is like handing out nuclear blueprints and hoping no one builds bombs,” said Dr. Evelyn Chow of MIT’s AI Risk Institute.
Despite this, Meta has already committed over $3 billion to AI alignment research—meant to ensure the models operate safely in uncertain environments. That includes hiring ex-OpenAI and Anthropic researchers who previously worked on interpretability and long-term safety.
Zuckerberg’s endgame: empire of the mind?
With this move, Zuckerberg may have completed his transformation from tech CEO to architect of post-human cognition. Meta’s AI roadmap spans healthcare, education, automation, and even diplomacy, with new tools in development for “machine-mediated negotiation” and AI-guided governance simulation.
As the Times reported, Zuckerberg’s pitch to investors is no longer about ad revenue—it’s about civilization:
“We are building the mind of the future—and it must serve everyone.”
Whether that mind is controllable, or whose values it will reflect, remains the defining question of 2025.
