5 funding signals hidden in Zuckerberg’s ‘superintelligence for everyone’ manifesto

Mark Zuckerberg’s August 10 manifesto sounds philosophical, but it also reads like a spending map. The essay leans on three ideas — individual empowerment, invention over automation, and power balance as a safety rule — and those ideas line up with five places Meta is already putting capital: compute, energy, agents, open source, and tools for new businesses.

The clearest product bet is a personal AI agent that keeps working in the background on one person’s life. Zuckerberg imagines it handling health, money, careers, relationships and hobbies, with smart glasses as the natural way to talk to it. He even uses his own home as an example, including an agent that tracks his fitness and another that helps his daughter bake by assembling recipes and ordering ingredients.

That vision needs a lot more than a chatbot. Meta guided 2026 capital spending to as much as $145 billion, up from $72.2 billion in 2025, and said it had $279 billion in future AI-related data-centre lease commitments in its Q2 filing, a 53% jump in one quarter. Some of the build-out runs through joint ventures, including a Blue Owl-backed vehicle tied to most of the cost of Hyperion in Louisiana. The money trail is already pointing in the same direction as the manifesto.

The bottlenecks are physical, not abstract. Verse’s $54 million Series B, with NVIDIA participating and Bessemer Venture Partners leading, is one example of startups attacking grid delays by using on-site battery storage to get AI data centres online up to three years faster. Meta is pushing on the same front, saying it is building its own energy-generating capacity for some projects and designing new facilities to be among the most water-efficient in the world.

Then there’s the open-source piece. Zuckerberg said Meta will keep releasing some open models, and the company’s independent board now gets to approve the safety criteria used for releases. He also wants closer, proactive collaboration with government rather than fixed review timelines. Add that to funding for infrastructure around open-weight models, like Featherless.ai’s $20 million Series A backed by AMD and Airbus, and the pattern is obvious: Meta wants AI to spread widely, but on terms it helps set.