The real story isn't Anthropic's headline revenue growth—it's that they're burning serious cash on inference while chasing an IPO, and their acquisition of Decart (likely for video/multimodal capabilities) signals they're scrambling to compete with OpenAI's product velocity rather than leading on safety. Nvidia's $500B financing theater is exactly that: theater masking slowing enterprise AI adoption and capex discipline that investors are finally demanding. Skip the papers on "omni-modal scientists" and motion tracking benchmarks; watch instead how llama.cpp's relentless optimization keeps closing the performance gap between $10K GPUs and $100K clusters—that's the actual hardware story, because efficiency beats raw compute in deployment.