Vercel Founder Rauch: The Future of AI is Multi-Model
Guillermo Rauch, founder and CEO of Vercel, commented that "the future of AI is multi-model," criticizing current products that attempt to hide underlying model choices from users.
His first argument is that product designs that conceal model choices from users/developers prevent consumers from participating in what he calls "the most exciting market competition of this era." In other words, the performance and price competition among model vendors should translate into perceptible choices and cost advantages for end users, but this is interrupted by "hidden choices" in product design.
The second argument is that hiding model choices also prevents users from "matching the best tool for the job," meaning they cannot select the optimal model based on task characteristics (such as code generation, long text reasoning, multi-modal processing, etc.), but instead passively accept a single model solution predetermined by the product provider.
Based on publicly available information, this stance aligns with Rauch's own product strategy: he is the creator of the Next.js framework and the founder of the front-end cloud platform Vercel. Vercel's AI SDK allows developers to uniformly call and freely switch between multiple model providers such as OpenAI, Anthropic, and Google, and its generative interface product v0 also adopts a multi-model architecture rather than being tied to a single foundational model.
From a market mechanism perspective, the beneficiaries of the "multi-model open" approach are developer tools/infrastructure companies that can interface with multiple model APIs simultaneously. The lower the switching costs between models, the stronger the stickiness and bargaining power of these intermediary tools. In contrast, products that tend to hide model choices and present a single brand to end users typically aim to lock in purchasing relationships and control cost exposure by reducing the user's cognitive burden. Financially, the more the price wars and performance competitions among model providers are amplified and transmitted to end users by developer tools, the more it benefits the reduction of model procurement costs for downstream applications, while further concentrating bargaining power and traffic distribution rights in the hands of intermediary tools.
Source: Public Information
ABAB AI Insight
Guillermo Rauch's historical trajectory began with infrastructure tools: he founded the real-time communication library Socket.io in 2011, then established ZEIT (later renamed Vercel), and led the development of one of the most widely used React application frameworks, Next.js. In the 2020s, Vercel gradually expanded from a front-end deployment hosting platform to the AI application infrastructure field, launching an AI SDK that allows unified access to multiple model providers and a generative interface product v0. This series of actions has transformed Vercel from a "front-end hosting provider" to an "AI application tool layer" company.
In recent years, Vercel has continued to expand its AI-related product line through multiple rounds of financing, with the core logic of resource mobilization being: rather than participating in the model layer "arms race" by developing or binding to a single foundational model, it is better to position itself as an intermediary tool connecting developers with multiple model providers, absorbing the benefits brought by competition among all model providers through an abstracted SDK, rather than directly competing with model vendors like OpenAI, Anthropic, and Google. This alignment naturally keeps Vercel's commercial interests consistent with the "multi-model open" stance.
A similar "neutral tool layer, betting on ecological openness" path can be compared to Kubernetes and Terraform in the early cloud computing era: both maintained compatibility and abstraction across multiple underlying cloud vendors (AWS, Azure, GCP), becoming the default tools for developers rather than binding to a single cloud vendor. Currently, Vercel is attempting to replicate this path at the AI model invocation layer, at a stage where the competitive landscape of the model layer has not yet solidified, and the developer tool layer is still vying for the "de facto standard" position.
This essentially represents a "transfer of pricing power": if the multi-model, freely switchable product route becomes the industry mainstream, the price comparison and substitution elasticity among model providers will significantly increase, and pricing power will shift from single model vendors to intermediary tools and end developers. Mechanically, this is because once developers can switch between multiple models at low cost, model vendors will find it difficult to maintain premium pricing based on lock-in effects and must continuously compete on performance and price; the tool layer, by controlling this "switching entry point," gains new bargaining power and traffic distribution rights.
ABAB News · Cognitive Laws
- Whoever controls the switching entry point controls the bargaining power.
- Hiding choices does not protect users; it locks them in.
- A neutral tool layer benefits from the premiums of the giants.