硅谷教父马克·安德森2026开年访谈：AI革命才刚开始，智能的价格正在崩塌

Group 1 - The AI revolution is considered the largest technological revolution ever witnessed, surpassing the internet and comparable to electricity and the steam engine, with the industry still in its early stages [4][25][26] - The emergence of AI technologies has been accelerated by the democratization of access, with powerful AI models available to the public through platforms like ChatGPT, Grok, and Gemini [5][6] - The revenue growth of leading AI companies is unprecedented, indicating that the industry is still in its infancy and will continue to evolve significantly over the next five to ten years [6][25] Group 2 - The cost of AI is experiencing "hyper deflation," with unit costs declining faster than Moore's Law, leading to increased demand as prices drop [10][11] - The AI industry operates on two core business models: consumer and enterprise infrastructure, with consumer AI products rapidly deploying to a global audience [7][8] - The pricing model for AI services is shifting towards a "tokens by the drink" approach, which will further drive demand as costs decrease [9][10] Group 3 - The GPU market is expected to transition from a shortage to an oversupply, driven by significant investments in AI infrastructure and the emergence of smaller, efficient models that can replicate the capabilities of larger models [12][14][15] - Companies like DeepSeek and Kimi are emerging as significant players in the AI space, demonstrating that breakthroughs can come from unexpected sources, including smaller firms and open-source models [16][18] - The competitive landscape is evolving, with a mix of large models and smaller models, leading to a diverse range of applications and capabilities in the AI sector [19] Group 4 - Startups in the AI space are not merely "wrappers" around large models; they are developing their own models tailored to specific industry needs, enhancing productivity for professionals [20][21][24] - The pricing strategies of AI startups are more innovative compared to traditional SaaS companies, allowing for better research and development funding [24] - The current state of AI products is still early, with significant improvements expected in the coming years as costs drop and capabilities expand [25][26]