$SoundHound AI (SOUN.US)$$Microsoft (MSFT.US)$$NVIDIA (NVDA....

$SoundHound AI (SOUN.US)$ $Microsoft (MSFT.US)$ $NVIDIA (NVDA.US)$ $Apple (AAPL.US)$ $Palantir (PLTR.US)$ The enthusiasm in the Capital Markets for foundational models may significantly cool down. In economic terms, capital may gradually recognize that there are clear externality issues in developing foundational models. Developing large language models (LLMs) requires enormous capital investment and computing resources, and the results are often difficult for a single enterprise to fully internalize. Once the model is released, the benefits of technological innovation will quickly diffuse throughout the entire ecosystem, benefiting many competitors and various sectors of society. This positive externality means that even if a company incurs high R&D costs, the ultimate benefits cannot be completely locked in by the company itself. In the future, capital may flow more towards innovations at the application layer of models, rather than continuing to endlessly invest in the stacking of foundational models and the race for parameter scale. For matters with externalities, government leadership becomes particularly necessary.