Anthropic's Amodei: open models just shift AI power to chip owners

Anthropic CEO Dario Amodei pushed back on the idea that open-weight AI models solve the industry's power-concentration problem, arguing in an exchange with investor Gavin Baker on X that AI centralizes power by nature and that open weights simply move that concentration around rather than eliminating it, The Decoder reported.

The exchange started after Baker publicly framed AI governance as a binary choice — concentrate the technology among a few entities, or distribute it as widely as possible — and argued that concentrated power has "rarely ended well throughout history." Former Meta researcher Yann LeCun weighed in on the same side, comparing open AI models to the printing press and the internet as tools that amplify human intelligence precisely because access to them is broad. Amodei rejected the binary framing directly, and made his position on open weights explicit:

“They just shift the concentration toward whoever has the most computing power and the most AI chips, which brings it right back to the big labs.”

— Dario Amodei, Anthropic CEO, in an X exchange with investor Gavin Baker

Amodei's case is that AI centralizes power "mainly because of the scaling laws and not because of regulation" — the returns to more compute keep going up, so whoever controls the most chips keeps compounding an advantage regardless of whether the resulting model's weights are open or closed. He pointed to Anthropic's own policy work as evidence the company isn't simply arguing for rules that entrench itself, citing California's SB53 as a proposal built to "slow down leading labs and favor smaller ones" rather than lock in whichever firm is currently ahead.

The exchange also cleared up a separate rumor that had been circulating on X: investor Gavin Baker had claimed Amodei told people internally that Anthropic could end up as the only private AI company left standing. Anthropic researcher Sholto Douglas called that characterization "totally false" before the broader debate over open weights and centralization took over the thread. Not everyone in the conversation was persuaded by Amodei's framing, either — former White House AI adviser David Sacks pushed back on a separate Amodei proposal for a federal agency to review frontier AI models, arguing it would introduce delays that put the US at a disadvantage against China, a criticism aimed less at the centralization argument itself than at what Amodei wants done about it.

The debate lands at an odd moment for the open-weights side of the argument: model releases like Moonshot AI's decision to fully open-source Kimi K3's weights show open models keep getting more capable and more widely available, even as Amodei's underlying point — that running a frontier-scale open model still requires enormous compute most people and companies simply don't have — remains largely unaddressed by open-weights advocates. It's also a reminder of how central this argument has become to Anthropic's own identity as a company: a safety-first lab whose CEO is now spending public arguments defending the idea that its own caution-driven regulatory proposals aren't just a moat dressed up as policy — a defense that only gets harder to make persuasively the larger Anthropic's own share of frontier compute becomes.

This article is for informational purposes only and does not constitute investment advice.