The Weights and Means of Production: On Open Source And Securing American Prosperity

Moonshot AI, a Chinese lab, released Kimi K3, an absolute banger of an open-weight model that handles text and images natively; boasting 2.8 trillion parameters and a million-token context window (tldr: big numbers means model good). Moonshot subsequently released the full weights for the model online, free for anyone to use. Demand for the hosted version hit such a feverish pitch that Moonshot had to pause new paid subscriptions. It might not top every leaderboard but, at $3 dollars per million input tokens vs Anthropic’s Fable 5 $10, K3 heavily undercuts Anthropic’s Fable 5, providing near-frontier reasoning at a third of the cost.

Naturally, US companies are paying attention. On OpenRouter, US models’ share of routed tokens has collapsed from 70 percent to 30 percent in a year. Coinbase cut its compute bill in half after switching to Kimi and GLM, and Cursor built its top-tier coding assistant on a Kimi backbone. The industry is moving from token-maxxing to harness-maxxing as firms start to focus less on the model type and more on what each finished task costs.

While Washington cycled through the stages of grief, alleging theft and threatening sanctions, Xi Jinping opened the 2026 World AI Conference with this: “We should adhere to the principle of openness and win-win and boost innovation-driven development. As a new engine of world economic growth and an accelerator for the shift of growth drivers, AI is moving from the digital world into the physical world. We should seize this rare, historic opportunity to encourage open source, openness, collaboration and sharing.”

It’s imperative that leaders and politicians understand what China is actually doing in the AI arena. Joel Spolsky described this quarter-century ago: commoditize your complement. Attack the price of complementary products or services, and demand for your product increases. If intelligence is the complement, what’s the product? Manufacturing. Energy.

Whoever is the leader in the means of production succeeds in the world where the marginal cost of intelligence approaches zero. That’s the arena where America is desperately behind. Our binding constraints are energy and manufacturing: a grid a third the size of China’s, adding a tenth as much capacity a year, a factory build-out that has shrunk every month of 2026, and a blue-collar skilled-labor pool aging out of employment. Every week Washington spends litigating whether Americans and US firms should be able to download a free model is a week conceding the necessary hard work to build our energy and manufacturing capabilities. America must win both the weights and means of production to secure the next century of prosperity.

_{Disclaimer: The views and opinions expressed in this article are strictly those of the author and do not reflect the official policy or position of any current or former employer. This content is provided for informational purposes only and does not constitute investment, financial, legal, or professional advice}