Zuckerberg: Superintelligence Shouldn't Belong to a Few

Zuckerberg is back on his personal superintelligence pitch, and the framing is sharper this time: if a small group controls superintelligence, that group ends up steering economics, science and politics for everyone else. His answer isn't regulation — it's distribution. Put a superintelligence in every pocket and no single lab gets to be the chokepoint.

It's the same line he drew in Meta's July 2025 memo, where he explicitly split the field: Meta wants AI aimed at personal empowerment, versus rivals who believe superintelligence "should be directed centrally towards automating all valuable work." The convenient part is that this vision runs through Meta's own surfaces — Instagram, WhatsApp, Ray-Ban glasses — so "everyone's superintelligence" is still Meta's superintelligence.

And the spending says he means the infrastructure, not just the slogan. Meta's 2026 capex is guided at up to $135B, nearly double 2025's ~$72B, with Meta Compute targeting tens of gigawatts this decade. Worth remembering he also quietly walked back full open-sourcing of the most powerful models, citing superintelligence risk — which makes the anti-concentration argument harder to fully buy from the guy holding the weights.

Why it matters: "don't let a few control it" is becoming the standard defence for spending $135B to be one of the few.