Key Moments
Stanford MS&E435 Economics of the AI Supercycle | Spring 2026 | Applications, Coding AI
Stanford OnlineWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
AI coding agents now deploy software 86% of the time on Vercel; this shift to agentic infrastructure is the biggest expansion of software creation.
Key Insights
Vercel's platform has seen significant growth since October last year, particularly with the advent of models like GPT-4.5, where coding agents now account for a substantial portion of deployments.
The concept of 'elastic compute,' previously designed for human-written code, is being re-evaluated as AI agents can write, repair, and secure software, vastly increasing compute demand.
Amazon's historical data shows a 1% conversion drop for every 100ms of webpage slowdown, highlighting the importance of speed, a principle that now extends to agent response times.
Vercel's AI Gateway acts as a 'CDN for tokens,' optimizing and securing AI model outputs, similar to how CDNs accelerate and secure web page delivery.
Vercel's support agent now handles 93% of user inquiries, demonstrating how AI can drive massive efficiency gains and improve customer experience.
The 'block economy,' as described by Mitchell Hashimoto, emphasizes the importance of building reusable components and infrastructure that AI agents can easily integrate with.
From Human Code to Agentic Deployment: Vercel's Evolving Role
Guillermo Rauch, founder of Vercel, discusses how the internet's infrastructure has fundamentally shifted with the rise of AI coding agents. Vercel, initially focused on optimizing the developer experience for deploying JavaScript applications, has become a critical platform for deploying software generated by AI. Rauch highlights that even everyday interactions, like ordering a Big Mac or using systems like Grok, can involve Vercel's infrastructure. The key shift is from a human-centric approach to software creation to one where AI agents are primary creators and deployers. This expansion of who can create software is unprecedented, transforming the total addressable market for software development. Vercel's platform, often described as the 'peanut butter and jelly' for coding agents, provides the essential infrastructure needed to deploy the software these agents write, addressing a critical bottleneck in the AI development lifecycle.
The democratizing power of open source and developer experience
Rauch's journey and the foundation of Vercel are deeply rooted in a commitment to open source and prioritizing developer experience. Growing up in Argentina, he taught himself to code using limited resources, which instilled a desire to make tools accessible to everyone. This philosophy led to the creation of popular open-source projects like Next.js and the subsequent building of Vercel into a multi-billion dollar company. The initial frustration with the complexity of deploying web applications using existing cloud providers like AWS, Google Cloud, and Azure fueled Rauch's ambition to simplify the process. He aimed to empower JavaScript developers, even those focused primarily on front-end user experiences, with the ability to deploy scalable, high-performance applications without needing deep infrastructure knowledge.
AI's massive expansion of software creation capabilities
The arrival of advanced AI models, particularly since last October with developments like GPT-4.5, has dramatically altered the landscape of software creation. Rauch emphasizes that this marks the 'biggest expansion and revolution in the total addressable market of software and specifically the creation of software in the history of our times.' Where previously Vercel catered to an estimated 20 million developers, AI agents are now enabling a much broader populace to generate code. He draws a parallel to the historical expansion of computing access, from mainframes to personal computers. Vercel's role has evolved to become the 'jelly layer' of deployment for software produced by these agents, which are less prone to the human bias of 'it works on my machine.' This massive influx of AI-generated code necessitates robust infrastructure capable of handling unprecedented deployment volumes.
The inversion from pages to agents and the changing cloud
Rauch posits that the very essence of cloud computing is transforming, moving from a focus on 'web services' to 'agent services.' Amazon Web Services (AWS), for instance, might be rebranded as Amazon Agent Services (AAS) if launched today, reflecting the new primary entity developers want to deploy: AI agents. This shift involves new infrastructure primitives and frameworks designed specifically for AI applications and agents. The traditional internet was built on instantaneous request-response cycles, epitomized by Amazon's observation that a 100-millisecond delay can reduce conversion rates by 1%. However, agentic workflows often involve 'long-thinking streams' where agents perform tasks over extended periods, from minutes to hours or even days. This necessitates different infrastructure capable of handling these asynchronous and extended computations, including agents that can manage and grow entire businesses behind the scenes.
Compute for agents and the commoditization of tokens
The economic model of the cloud is also evolving. Elastic compute, popularized by services like AWS EC2, was designed for human-written code and its scalability was bound by the number of available programmers. With AI agents now writing code, repairing code, and securing code, the demand for compute is exploding. Rauch notes that agents are not only writing code but also potentially 'writing software, are writing software, are repairing software, are securing software,' leading to both beneficial applications and concerns like hacking agents. Concurrently, tokens are emerging as the new 'hot commodity,' shifting pricing models from seat-based SaaS to token-based consumption, which measures intelligence usage. This signifies a move from streaming pixels to streaming intelligence, fundamentally altering business models and the perceived value of software.
The 'block economy' and composable infrastructure for agents
Rauch explains Vercel's strategy through the lens of the emerging 'block economy,' where composable building blocks are essential for AI agents. He draws an analogy to LEGO blocks, emphasizing the need for technologies that agents can easily build upon. Vercel's own frontend UI engine, 'shadcn,' is cited as having a 'near monopoly status' (90.1%) among React deployments, and Vercel itself experiences a similar dominance in deployment options for agents. This success is attributed to Vercel's focus on technologies like Tailwind CSS and frameworks like Next.js, which offer 'local reasoning' properties. This means components are designed to be easily understood, portable, and future-proof, which is crucial for LLMs with limited context windows. By providing these well-defined, composable blocks, Vercel makes its technologies attractive targets for AI agents seeking to build software efficiently.
Agentic infrastructure: The three pillars of Vercel's vision
Vercel's framework for the future is centered around 'agentic infrastructure,' which comprises three key components. First, it provides the necessary infrastructure for coding agents, acting as the deployed environment for tools like Claude Code or Codex. Second, it enables the creation of custom agents, allowing businesses to build their own AI agents tailored to specific needs, such as an AI-native school or a support agent. Vercel's own support agent, handling 93% of inquiries, exemplifies this. Third, agentic infrastructure facilitates automation by these agents. This vision extends to new paradigms like the 'AI Gateway,' a CDN for tokens, and 'Sandbox' environments, which function as the fundamental compute units for agents, analogous to providing a computer to a new human employee. This holistic approach aims to provide a full-stack cloud for AI and agents.
The evolving nature of software value and customer engagement
The concept of 'throwaway software' or 'free software' is emerging, where custom solutions can be rapidly generated and deployed for specific tasks, such as a single sales call or a quick business process optimization. Vercel's platform enables customers to 'vibe code' tools quickly, sometimes replacing existing commercial software and saving significant costs. Rauch notes that while some applications might be ephemeral, this rapid iteration drives engagement, fueled by the reflexivity of AI and the difficulty of foregoing efficiency gains. However, he clarifies that not all software is fleeting; complex infrastructure and core systems still require significant engineering effort. The most significant shift is in the audience, with agent-to-agent communication becoming more prevalent, where customer agents file feature requests or bug reports with service providers' agents, redefining customer support and development feedback loops.
From EC2 to agentic cloud and the future of infrastructure
Rauch contrasts the traditional elastic compute model (EC2) with the emerging needs of agentic infrastructure. EC2 was built for human-coders, its capacity limited by programmer availability. Now, agents are writing code at an exponential rate, demanding massive compute. Vercel is evolving to meet this demand, offering services like 'Sandbox,' which provides agents with compute environments similar to how companies provide laptops to human employees. Security is a paramount concern, with Vercel developing cybersecurity solutions for these agent sandboxes to prevent data exfiltration. Looking ahead, Rauch envisions the cloud itself becoming a 'self-driving car,' automating maintenance, optimization, and scaling. This aims to eliminate the painful 'pager duty' experiences associated with traditional cloud operations, ushering in an era of self-optimizing, highly efficient cloud infrastructure managed by AI.
Vercel's integrated approach and value accrual in the AI economy
Vercel's decision to 'do it all' – from developer tools like sandboxes and AI gateways to deployment and UI components – stems from a philosophy of comprehensive product development, often built through 'dogfooding' their own platform. Rauch believes that agents represent the 'last class of software,' and Vercel is building the supporting infrastructure. He argues that while value might currently accrue heavily in hardware (chips, data centers), it will increasingly shift to the software and infrastructure layers that make AI useful. For instance, domain names have become a key entry point for new users interacting with AI, a bet Vercel made early on by simplifying DNS configuration. The company aims to be the 'front door to any emerging idea on the planet,' positioning itself to capture value by facilitating the creation and deployment of these ideas, with a strong emphasis on security and governance for AI agents.
The future of SaaS and the shift to consumption-based models
Rauch is short on businesses that assume code scarcity or rely on restrictive, drag-and-drop builders, viewing them as patronizing and outdated in the age of AI-generated code. He predicts that many businesses built on these foundations will need serious pivots. Instead, he is bullish on companies that embrace open access, consumption-based pricing, and rapid deployment – essentially, any business 'moving at the speed of tokens.' He notes that traditional enterprise sales cycles and restrictive pricing models are ill-suited for the agility demanded by AI. Vercel, by removing rate limits (with caveats for security and abuse prevention) and favoring pay-as-you-go models, is aligning with this shift, catering to the immense demand from AI agent platforms that can generate extraordinary volumes of deployments.
Beyond Vercel: Space Tech and Energy Innovations
When asked what he would build if not Vercel, Rauch expressed a strong interest in space technology, citing a desire to become a multi-planetary species and his inherent inclination as an 'infrastructure guy' who values high availability and failover. He is also deeply excited about breakthroughs in energy, including fission, fusion, and geothermal power. He views the emergence of intelligence as a bidirectional flow of energy – energy in, intelligence out – making advancements in energy production a critical area for future innovation and human progress.
Mentioned in This Episode
●Software & Apps
●Companies
●Organizations
●Concepts
●People Referenced
AI Agent Preferences for Frameworks and Deployment
Data extracted from this episode
| Component Type | Preferred Technology | Market Share | Agent Example |
|---|---|---|---|
| UI Components | shadcn | 90.1% | Claude |
| Deployment | Vercel | 100% | Claude |
Common Questions
Vercel is a developer infrastructure company that provides tools and frameworks, especially open-source ones, to help developers deploy and scale web applications and AI agents efficiently. It aims to simplify the complexities of infrastructure, allowing developers to focus on building user experiences.
Topics
Mentioned in this video
The publisher of Guillermo Rauch's book, 'Smashing Node.js'.
A self-driving car company, mentioned alongside Tesla as an example of autonomous vehicle technology.
A financial services and technology company, its founder is cited with an anecdote about pager duty alerts.
A Q&A website for programmers, cited as an example of a static content business facing challenges in the current AI landscape.
A CRM platform that Vercel has effectively reinvented for its internal sales processes, demonstrating the power of customized agent applications.
Mentioned as an example where Vercel's technology might be used, such as in ordering a Big Mac.
Not mentioned in the transcript.
Mentioned in relation to its self-driving car technology as an analogy for the future of the cloud.
Mentioned as an example of a company building a business on open source.
Mentioned regarding their influence on the cloud with AWS and their data insights on website conversion rates.
Mentioned as the accelerator for a startup experiencing massive deployment demand on Vercel.
A technology company that has used Vercel's infrastructure to accelerate development, including meta.ai.
Mentioned for its CEO Toby S. who coined the term 'reflexivity of AI'.
The company behind Claude, mentioned in the context of its AI models generating tokens that require infrastructure management.
A company whose founder, Mitchell Hashimoto, is now on the Vercel board and coined the term 'block economy'.
A platform for version control and collaboration, mentioned as an example of software storage that often lacks practical deployment.
Mentioned in the context of a potential IPO, and as a forward-looking venture.
A CDN provider mentioned as an example of infrastructure built to scale, accelerate, and secure the delivery of web pages.
A CDN provider mentioned alongside Akamai in the context of scaling and securing web content delivery.
Not mentioned in the transcript.
A developer infrastructure company specializing in building and deploying web applications and agents, founded by Guillermo Rauch.
A popular JavaScript library for building user interfaces, central to Vercel's focus on front-end developers.
An open-source system for automating deployment, scaling, and management of containerized applications.
An interactive encyclopedia that Guillermo Rauch recalls from his childhood, cited as an example of a rich, engaging digital experience.
An early open-source software repository, mentioned in the context of historical platforms for code storage.
A productivity and note-taking application that is incorporating AI and agentic capabilities, using Vercel for infrastructure.
Microsoft's cloud computing service, mentioned as a competitor in the cloud infrastructure space.
A Vercel product or feature used for live coding AI-generated interfaces, mentioned in the context of replacing existing software.
An AI model which, like other coding agents, preferentially selects Vercel for deployment and uses shadcn for UI components.
An example of a system that interacts with agents hosted on Vercel, potentially referring to a model or CLI tool.
Elastic Compute Cloud, a service from AWS that popularized the concept of elastic compute for human-written code.
A CSS framework that Vercel bet on, known for its utility-first approach and property of 'local reasoning', which benefits agents.
A hypothetical advanced AI model mentioned in the context of token usage and the potential for optimization with smaller models.
A large language model developed by Google, mentioned alongside Claude and GPT in the context of token generation and infrastructure needs.
A coding agent mentioned as a user of Vercel's infrastructure and as an example of a team discovering user needs like domain names.
Vercel's compute platform that powers services like AI Gateway.
A major cloud provider mentioned alongside AWS and Azure as existing options before Vercel's focus on developer experience.
Amazon Web Services, cited as a traditional cloud provider that Vercel aimed to improve upon in terms of developer experience.
A platform for developing, shipping, and running applications in containers, used in training AI models.
A coding agent that requires deployment infrastructure on platforms like Vercel.
The programming language used by Vercel and its popularity, especially with frameworks becoming more accessible.
These models are increasingly writing, repairing, and securing software, leading to massive compute demand.
A core component of Vercel's infrastructure that reuses virtualization primitives to provide ephemeral computers for agents.
The new commodity in the AI economy, replacing traditional software models and influencing pricing and business strategies.
Software as a Service, a traditional business model discussed in the context of its potential demise and replacement by token-based pricing.
Undergoing a fundamental change with the rise of AI agents, moving from 'Amazon Web Services' to 'Amazon Agent Services'.
The future of how systems and agents will interact, moving away from direct human-to-human technical support.
Constantly evolving, with Vercel's tools shaping its modern iteration and new agentic infrastructure set to redefine it.
Guillermo Rauch's early career involved JavaScript contracting at a young age.
A trend in Silicon Valley mentioned as indicative of the low learning curve for modern JavaScript tooling at the time.
A critical consideration for the future of AI and agentic systems, with Vercel developing products to protect sandboxes and computers.
A type of energy generation that Guillermo Rauch is deeply interested in, seeing it as a key to future breakthroughs.
A core philosophy behind Vercel's creation and business model, emphasizing accessibility and community contribution.
A characteristic of the audience and the general spirit of innovation discussed in the context of starting new ventures.
The new paradigm in software development and cloud computing, driving Vercel's evolution towards agentic infrastructure.
The core focus of Vercel's current development, providing the necessary tools and frameworks for building and deploying AI agents.
A property of code and APIs, particularly emphasized in technologies like Tailwind CSS and Next.js, that makes them scalable and understandable for agents.
The ability of components and services to be easily combined and reused, a key prerequisite for agentic scalability.
The primary obsession when Vercel was founded, aiming to simplify the deployment process for developers.
A form of energy generation that Guillermo Rauch finds deeply exciting.
The driving force behind the expansion of software creation and the shift towards agentic infrastructure.
A term coined by Mitchell Hashimoto describing a future economy based on reusable building blocks, relevant to agentic development.
A form of energy generation that Guillermo Rauch finds exciting, alongside fusion and geothermal.
The significant increase in Vercel's popularity and deployment volume, particularly since October of the previous year.
Guillermo Rauch's aspiration for humanity, reflecting his infrastructure-minded approach and desire for high availability.
A cloud computing concept popularized by EC2, designed for human-written code and its limitations in the age of AI-generated code.
The destination Guillermo Rauch would focus on if not building Vercel, reflecting his interest in space exploration and multi-planetary aspirations.
The suburb where Guillermo Rauch grew up, highlighting his humble beginnings and self-taught coding journey.
The city Guillermo Rauch moved to pursue the startup world, and a location chosen for its structural safety.
More from Stanford Online
View all 80 summaries
50 minStanford MS&E435 Economics of the AI Supercycle | Spring 2026 | Building AI Factories
42 minStanford CS153 Frontier Systems | Scale, AGI, and the Future of Everything
39 minAI in Healthcare Series: Inside the Rise of AI in Healthcare, Open Evidence and Cyber Risks
72 minStanford CS336 Language Modeling from Scratch | Spring 2026 | Guest Lecture: Dan Fu
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free