20VC with Harry Stebbings• Apr 25, 2026• 48:49Interview

Replit CEO: Why the SaaS Apocalypse is Justified & Why Coding Models are Plateauing | Amjad Masad

Amjad Masad•Co-founder and CEO, Replit

Executive Summary

Replit's vision has evolved from 'learn to code' to 'learn to create,' enabled by the rise of agentic AI capable of long-horizon tasks, aiming to expand the software creator base from millions to a billion.
The company employs a 'society of models' strategy, using a mix of frontier models (Anthropic, OpenAI), cost-effective models (Google Gemini), and custom-trained models to optimize performance and cost for different tasks.
AI-native platforms like Replit are disrupting the enterprise software market by enabling non-developers to build custom applications that replace vertical SaaS tools, often building directly on data warehouses like Databricks.
The AI model market faces economic challenges, with a duopoly at the frontier (OpenAI/Anthropic) and low profitability for model providers due to the high cost of NVIDIA chips, which keeps token prices from falling rapidly.

9 quotes

Concerns Raised

The duopolistic nature of the frontier AI model market limits price competition.
High costs of NVIDIA chips are preventing significant decreases in token prices and squeezing model provider margins.
Platform risk from gatekeepers like Apple, who can arbitrarily block applications and stifle distribution.
The cyclical nature of the 'build vs. buy' decision for custom models requires constant strategic adaptation.

Opportunities Identified

Empowering a new generation of non-technical creators to build and deploy software.
Replacing expensive and inflexible vertical SaaS tools with custom, AI-generated applications.
Enabling enterprises to build solutions directly on their data warehouses, increasing efficiency and data control.
Automating the entire software lifecycle, including maintenance, testing, and security, with AI agents.

Key Themes

The Rise of Agentic AI

The key technological unlock for Replit's mission was not just AI, but agentic AI capable of performing complex, multi-step tasks autonomously. This shifts the paradigm from simple code completion to full application generation, maintenance, and security monitoring, making software creation accessible to non-coders.

This marks a fundamental change in human-computer interaction, where users can specify outcomes and have AI agents execute the steps, dramatically lowering the barrier to creating sophisticated software.

AI Model Market Dynamics

The frontier model market is effectively a duopoly between OpenAI and Anthropic, with Google's Gemini competing on price-performance. This lack of competition, coupled with high NVIDIA chip costs, means model providers have surprisingly low margins (~40%) and token prices are not decreasing as quickly as expected.

Understanding these economics is crucial for any company building on top of AI, as it impacts cost structures, profitability, and the strategic decision of whether to use third-party models or invest in custom, fine-tuned open-source alternatives.

The 'SaaSpocalypse' and Custom Application Building

Platforms like Replit are enabling enterprises to replace niche, vertical SaaS products (e.g., survey software) with custom-built applications. While foundational 'systems of record' like Salesforce remain, this trend threatens the growth of many point-solution SaaS companies and shifts value towards platforms that facilitate creation.

This represents a significant threat to the traditional SaaS business model and an opportunity for enterprises to build more tailored, cost-effective solutions directly on their own data infrastructure.

The Evolution of the Development Environment

The traditional Integrated Development Environment (IDE) is considered obsolete in the age of AI. The future belongs to comprehensive, AI-native platforms that manage the entire software lifecycle, including automated code review, testing, and security agents that monitor applications in production.

This signals a major platform shift for software developers and creators, moving from passive tools to active, agent-driven partners that handle the complexities of maintenance and security, not just initial code generation.

Platform Risk and Gatekeepers

Replit's experience with its iOS app being blocked by Apple for months highlights the significant risk faced by companies building on top of major, closed ecosystems. Despite a long history on the App Store, a company's distribution can be halted by the platform owner's unilateral decisions.

This is a critical strategic consideration for any tech company, emphasizing the need to diversify distribution channels and be aware of the power wielded by platform gatekeepers like Apple and Google.

Get started free

Topics

Processed Apr 25, 2026 yt-dlp + mlx-whisper + Gemini