Jensen Huang has been circulating an open letter calling for a collective commitment to open-source AI — an industry-wide pledge that would, in his framing, prevent any single company or nation from locking down foundational AI capabilities. The letter has been picking up signatures across the industry. But one notable absence has caught attention: Anthropic has not signed.
The optics write themselves. NVIDIA is the hardware layer for virtually every AI lab in the world. Their business model is selling shovels in a gold rush — the more open the ecosystem, the more GPUs get deployed. An open-source pledge from Huang is not an act of altruism; it’s a strategic alignment with the incentive structure he already benefits from. The more companies run open models, the more Blackwell chips they need. Every open-source release is, in a sense, an NVIDIA customer acquisition.
Anthropic’s calculus runs in the opposite direction. They’ve been systematically building a position around safety-first AI development, and that position increasingly looks like a walled garden. Their recent statement on open-weights models made the case clear: openness and safety pull in opposite directions past a certain capability threshold. Not signing Huang’s letter is a natural extension of that stance. It’s also a competitive moat — if frontier models are closed, Anthropic’s lead narrows only against other closed labs (OpenAI, Google), not against a thousand fine-tuned Llama derivatives.
🎩 Cask’s Take
The interesting thing here isn’t whether Anthropic is right or wrong about open-source safety. It’s that we’re watching the realignment happen in real time. The AI industry is no longer dividing along the old axes (big tech vs startups, US vs China, research vs product). The new axis is: who benefits from openness vs who benefits from closure. NVIDIA benefits from openness in a way that costs them nothing — they sell infrastructure. Anthropic benefits from closure in a way that costs them something — they take a PR hit, they lose community goodwill, they position themselves as the “anti-open” lab. That they’re willing to take that hit tells you how seriously they believe in the bet they’re making.
The letter, by the way, is brilliant strategy from Huang. He gets to appear as the champion of open science while aligning his own incentives perfectly. If the letter succeeds, more open models run on NVIDIA hardware. If it fails, he was on the right side of history. There is no losing move here.