Large models are becoming "free infrastructure"; are the real alpha opportunities in the application layer?

A senior investor has spoken out through a hedge fund CIO, suggesting that investment focus in the AI sector should **shift from infrastructure to the application layer.** According to an article published by Eric Peters, Chief Investment Officer of One River Asset Management, on November 24, 2025, entrepreneur and seasoned investor Sparks pointed out that there is a misconception in the market about large language models (LLMs). The real long-term investment value does not lie in building these models themselves, but rather in the ecosystem of applications built on top of them. Sparks compares the current situation of LLMs to the early days of broadband network construction. He believes that **LLMs are rapidly becoming commoditized; these models are "basically being given away for free," just like tap water that everyone loves.** > “When broadband was first built, it was not the operators who made the big money, but the companies that built businesses on top of broadband.” Sparks believes that **the current market landscape is that multiple large language model developers (such as OpenAI) are fiercely competing, attempting to outdo each other on nearly identical features, which is no different from "creating 10 Googles at the same time."** In his view, LLM developers play the role of the operators—they provide the fundamental tools, but the richest profits in the value chain will flow to those who best understand how to use these tools to create real business value as application developers. This judgment directly influences his investment strategy. Sparks stated that he would rather "be the person who earns real money using search engines" than the developer of the search engine itself. He is currently dedicating all of his "forward-looking energy" to searching for opportunities in the next 2-3 years (which he refers to as “T+3”). He believes that **the biggest opportunities lie in startups that can deeply integrate AI capabilities with specific industries, thereby creating significant efficiency improvements and transformations in business models.** Regarding the AI infrastructure that is currently hyped by the market, Sparks takes a cautious stance. He compares Nvidia’s valuation of up to $5 trillion to Cisco’s valuation back in 1999, believing that this is more of a "retrospective" valuation, reflecting achievements that have already occurred rather than future potential. He predicts that although the United States will invest $500 billion in the coming years to build data centers to meet "crazy" demand, this wave of construction is more akin to a "boomlet;" the capital and attention focused on it have already become "offsides." Risk Warning and Disclaimer The market involves risks—investment should be made cautiously. This article does not constitute personal investment advice, nor does it take into account the special investment goals, financial circumstances, or needs of individual users. Users should consider whether any opinions, perspectives, or conclusions in this article fit their specific situations. Investing based on this article is at your own risk.