The real story this week isn't Zuckerberg's manifesto or passive component shortages—it's the infrastructure squeeze. Nvidia's half-trillion dollar ask signals that the actual bottleneck isn't model capability or software (see: the steady drip of open-source updates like llama.cpp and Ollama) but physical AI compute and the geopolitical capital racing to secure it. Meanwhile, governments are quietly scrambling with evaluation frameworks (Dutch LLM benchmarks, NUS mandating AI literacy) because they've realized they can't regulate what they don't understand—and the vendors know it. Watch the passive components angle hard: if YAGEO and Walsin are posting record sales on AI demand, margin compression is coming for everyone building inference hardware, and that's where real differentiation dies.**