$NVIDIA(NVDA.US) is really hilarious. A bunch of people who don't use AI are constantly shouting about excessive capital expenditure. Just try using it yourself: the biggest ones like Open AI and the so-called most powerful Anthropic, none of them can keep up with response speeds; you hit rate limits quickly. Whether open source or closed source, everyone says Kimi K3 challenges the closed-source business model, but where is that challenge? Look at the current parameter counts. Forget about the cost of local deployment for a moment; just see if you can even subscribe to their plans. They can't even guarantee service for existing users—over 30 tokens per second, barely able to hold a coherent conversation. Running a task takes at least three to five hours. Even if subscription numbers weren't limited, delivery quality would inevitably drop. It's speechless. AI applications are completely a wasteland; true demand hasn't emerged yet. We're not there until full App AI-ification happens at the stage of flat-rate token pricing. I mean genuine AI integration, not just stuffing a chatbot into Alipay. Treat tokens like phone bills or data plans, and then take another look.