“短缺终将导致过剩”，a16z安德森2026年展望：AI芯片将迎来产能爆发与价格崩塌

AI Technology Scale - AI represents a technological revolution larger than the internet, comparable to electricity and microprocessors, and is still in a "very early" stage [2] - The unit cost of AI is decreasing at a rate that far exceeds Moore's Law, leading to explosive demand growth [2] Market Dynamics - Following the historical pattern of "shortage leads to surplus," the large-scale construction of GPUs and data centers will eventually result in oversupply, further driving down AI costs [3] - The future AI market structure will resemble the computer industry, with a few "god-level models" at the top and a vast number of low-cost "small models" proliferating at the edges [3] US-China Competition - The competition between the US and China is characterized as a dual hegemony, with Chinese companies like DeepSeek and Kimi making remarkable progress in speed, open-source strategies, and chip self-research [3][10] - The emergence of DeepSeek has surprised both Washington and Silicon Valley, indicating a shift in global price competition that may influence US regulatory approaches [10] Business Model Evolution - AI applications are transitioning from "pay-per-token" to "value-based pricing," with startups moving beyond being mere wrappers to integrating their own models [4][12] - High pricing can benefit customers by supporting better research and development, as AI startups demonstrate more creativity in pricing compared to SaaS companies [12] European AI Landscape - The EU's inability to lead in innovation has led to a focus on "regulatory leadership," which has stifled local AI development and caused major companies like Apple and Meta to withhold new features in Europe [5] AI Democratization - Advanced AI technologies are now accessible to anyone, breaking down barriers and allowing immediate use and validation of previously expensive technologies [6] - Public sentiment shows fear of AI replacement, yet actual behavior indicates a rapid adoption of AI technologies [6] Cost Deflation and Investment Outlook - The extreme deflation of AI input costs is expected to drive demand growth beyond expectations, with significant investments anticipated in the coming years [8] - Historical cycles suggest that shortages will lead to oversupply, resulting in a dramatic decrease in AI companies' unit costs over the next decade [8] Model Competition - The future of AI will not be a zero-sum game between closed large models and open-source small models, but rather a clearly defined "intellectual pyramid" [13] - The industry structure will feature a few supercomputer-like "god models" at the top, with numerous smaller models extending to embedded systems [13]