Meta’s Zuckerberg Makes the Case for Distributed Superintelligence — and Why the Argument Is Also a Strategy

In a Wall Street Journal op-ed published July 28, Mark Zuckerberg argues that who controls access to superintelligence is the defining question of the age — a genuine philosophical case that happens to align precisely with where Meta holds its strongest competitive position.

What Happened

Writing in the Wall Street Journal on July 28, 2026, Meta CEO Mark Zuckerberg published an op-ed titled “The AI Future Is for Everyone,” organized around a single contested premise: the central question of this technological moment is not whether superintelligence will exist, but who will have access to it. His answer is that a superintelligence restricted to a handful of institutions poses a greater danger than one widely distributed — and he structures the argument around three principles he describes as the foundation of Meta’s AI work.

The three principles are individual empowerment as the engine of prosperity, “invention, not automation” as the primary purpose of superintelligence, and balance of power as the true basis of safety. On each, Zuckerberg makes specific — and contestable — claims. On safety, he draws an analogy to open-source software, arguing that distributing access to powerful systems broadly tends to improve security over time, while explicitly carving out biological risk as a domain requiring additional government coordination. On fairness, he offers a thought experiment: a world where only one person has access to a superintelligent lawyer is a world with unfair courts; a world where everyone does allows people to check institutions and each other. On jobs, he inverts the automation-doom narrative, arguing that wide distribution of superintelligence would lower the capital required to start companies, create more jobs, and shift economic weight toward small businesses.

He pushes back directly on what he frames as doom-laden concentration arguments — the position that AI is so dangerous the only safe path is extreme centralization of control — calling that logic itself dangerous and citing history’s poor record with absolute power exercised benevolently. Meta, he writes, is building toward empowerment, invention, and balance of power. These are his arguments, presented in an opinion context; they are not established facts, and several — particularly the “open is safer” claim — remain genuinely disputed among security researchers.

The Structural Read

Read in isolation, Zuckerberg’s op-ed is a coherent political philosophy. Read in context — against the same week’s Anthropic open-weights position and the OpenAI/Anthropic employee petition for deliberate pacing — it is also the clearest statement yet of a deepening fault line in AI governance. Two camps have now formed with relatively hard edges: one argues that safety requires controlling the pace and concentration of frontier capability; the other argues that safety requires distributing that capability so broadly that no single actor can abuse it. This week, both camps published.

The fair strategic note is that Zuckerberg’s philosophical position and Meta’s competitive interest are difficult to disentangle. Meta is the open-weight leader through Llama. “Distribute superintelligence to everyone” is also, structurally, a move to commoditize the base model layer that OpenAI and Anthropic currently monetize — pushing value toward distribution endpoints, where Meta’s roughly three billion users live. This is the logic the Open-Weight Alliance framework describes: open the layer rivals monetize; capture value at the layer you already control. Zuckerberg is stating that strategy as political philosophy. Both can be true simultaneously — a genuine belief and a self-serving position are not mutually exclusive.

The value-cascade framing from Beyond NVIDIA’s Moat makes the underlying map legible. Durable value in the AI stack sits at three points: the physical compute floor, the routing and permission junction, and the distribution endpoint. The base model layer — the one Zuckerberg wants fully commoditized, and which Kimi K3 and Llama are already commoditizing — is not one of them. Anthropic and OpenAI appear to concede this too: their position is not “closed models forever” but rather “control the compute floor and the distillation channel.” Same diagnosis of where base models are heading; opposite prescription for what to do about it; each prescription aligned with where the author’s assets sit.

Where Each Camp’s Assets Sit — and What They’re Arguing For

Three Implications