Tesla builds physical cars, energy grids, and robotics. SpaceX builds physical rockets and satellites. That tangible hardware scale is real.
What isn't real is pegging the upside to the myth that closed-source compute margins stay astronomical forever.
Valuing an AI arm on the premise that compute cartels will keep buying overpriced inference ignores basic technology cycles. Compute commoditizes; open-weight models and lean silicon will inevitably crush foundation-model moats. When inference runs cheaply at the edge, artificial scarcity evaporates.
In fact, you can totally interchange any of Claude, Chatgpt, Gemini, Minimax, Deepseek, GLM, Qwen, Kimi, even Mistral and others, for consumers they can easily just switch to any other models to resume their work. AI benchmarks are often bullshit, with smart "Goal oriented" AI prompts all these models can reach pretty much the same results with their "thinking".
A decade ago, the 11 year younger Elon was right: AI had to be open-source and non-profit to prevent bloated, rent-seeking monopolies like Google or whatever scammers comes out of OpenAI/Anthropic. Anchoring Tesla and SpaceX’s future to a hyper-capitalized, proprietary US compute hegemony, when Chinese AI runs on 100x cheaper hardware, contradicts that original vision.
Tesla holders shouldn't surrender governance over a future of millions of re self-driving Robotaxis for a speculative AI token bubble. Give them a ratio grounded in real hardware.