Benedict Evans Says AI is in the 1997 Internet Era, Suggests Embracing Radical Uncertainty
Former a16z analyst Benedict Evans stated on the Lenny Rachitsky podcast that the scale of AI is comparable to that of the internet or mobile internet, being neither large nor small, and is currently in 1997 rather than 2007, with most applications yet to be built, necessitating an assumption of radical uncertainty.
He believes that the doomsday scenario for jobs is largely fictional, that foundational model companies will struggle to maintain pricing power, and that value will shift to upstream application layers; anti-AI voices mix real and false discontent, and the only correct response is to actively engage with AI.
In terms of market mechanisms, developers and businesses buy practical AI tools and adaptability, while selling a wait-and-see or resistant mindset; event-driven dissemination of Evans' podcast views leads to capital flowing towards application layers and distribution platforms built on foundational models, benefiting from distribution endpoints and actual products from Google, Meta, Apple, etc., while being pressured by pure foundational model companies and those resistant to transformation.
Source: Public Information
ABAB AI Insight
Benedict Evans is known for his annual trend reports over the past six years, and this podcast continues his consistent framework, directly comparing AI to the early internet, emphasizing characteristics of 1997: most things are still unfeasible, predictions are largely promotional, and value ultimately flows to the application layer rather than the underlying infrastructure, similar to how telecoms generated huge revenues but had flat stock performance, while companies built on top created trillion-dollar value.
On the capital path, Evans suggests that businesses and individuals accumulate knowledge of AI's capabilities through practical use (such as proofreading, image generation, voice transcription), motivated by the desire to avoid being replaced rather than to resist, strategically positioning "immersing in technology" as the only career insurance, promoting a shift from a narrative of fear to one of productivity leverage.
Currently, AI is in a phase of rapid capability iteration but still has unclear business models; Evans' views remind the market that foundational models are becoming commoditized, and success will be determined by distribution, application scenarios, and real user value, rather than purely model performance.
Essentially, this is a technological replacement: AI automates repetitive labor and assists creativity, replacing traditional workflows, with the mechanism being the reshaping of actual job content rather than simply eliminating positions, pushing the entire industry from a defensive posture to actively embracing and reconstructing upper-layer value capture.
ABAB News · Law of Cognition
The mindset of 1997 will determine the winners and losers of 2026; early entrants gain the leverage of capability.
When foundational models are commoditized, application distribution becomes the new moat.
Before the essence of work is redefined, active use surpasses passive prediction, and adaptability equates to long-term pricing power.