20VC: Is SaaS Dead in a World of AI | Do Margins Matter Anymore | Is Triple, Triple, Double, Double Dead Today? | Who Wins the Dev Market: Cursor or Claude Code | Why We Are Not in an AI Bubble with Anish Acharya @ a16z

What It Covers

Anish Acharya, GP at Andreessen Horowitz, challenges conventional wisdom on AI disruption, arguing SaaS is oversold and enterprise software remains defensible. He covers foundation model competition, application layer opportunities, defensibility in AI-native companies, the future of agents, pricing dynamics, and why Series A remains the optimal investment stage despite competitive intensity and high valuations.

Key Questions Answered

• •SaaS Disruption Overstated: Software represents only 8-12% of enterprise spend. Even if companies vibe-coded their entire ERP and payroll systems, they would save just 8-12%. The innovation opportunity lies in optimizing the remaining 88-92% of spend, not rebuilding existing software. 75% of public SaaS companies have raised prices 8-12% since ChatGPT launched, with many raising 25% or more, indicating strong product-market fit rather than competitive pressure threatening their existence.
• •Multi-Model Aggregation Creates Value: Foundation models are innovating in lockstep with 80% substitutability but 20% specialization. Gemini excels at front-end coding while Codex handles back-end. Midjourney offers aesthetic opinions while Ideogram provides neutral graphic design. Application layer companies like Cursor aggregate multiple models, allowing developers to orchestrate the best tool for each task. This aggregation layer captures significant value as cost optimization alone does not drive model selection today.
• •Switching Costs Declining via Agents: Coding agents dramatically reduce the complexity, speed, and risk of transitioning between enterprise systems. Companies have hostages, not customers. Moving from SAP to Oracle previously required multi-year, high-risk projects that typically failed. AI-powered migration tools transform this dynamic, creating more customers and fewer hostages. This increased competition incentivizes the entire ecosystem to improve rather than enabling wholesale SaaS replacement.
• •Margins Require Nuanced Analysis: AI companies show worse blended margins due to subsidized user acquisition through free inference credits, but this represents healthy calories compared to 2021's empty calories from Google and Facebook ad spend. Separate month-one organic traffic from true customer acquisition cost. Evaluate margin profiles of converted power users separately from free trial costs. Power users now pay 10x higher subscription rates than pre-AI, with ChatGPT at $200 monthly and Grok at $300 versus Spotify's $20-25 ceiling.
• •Series A Optimal for Risk-Adjusted Returns: Competitive risk and pricing risk represent the correct risks for investors to take, not team risk, geographic risk, or fundraising risk. Seed investing requires seeing potential in nothing, while Series A provides dramatic signal through shipped product and actual sales. Companies that achieve zero-to-one typically become greater versions of themselves. Inertia is the most powerful force, so formidable founders making nonlinear progress should be underwritten to continue succeeding indefinitely.
• •Consumer Discretionary Spend Shifts to Software: Consumer discretionary spend will asymptote to 80-90% on software from current levels of a few hundred dollars monthly. This expansion covers companionship, entertainment, therapy, healthcare, professional development, and education. The frontier opportunity lies in pushing capabilities forward into new categories rather than cost optimization of existing use cases. AI-native categories emerging in 2026 will create entirely new markets beyond obvious 2023-2024 ideas like customer support and coding tools.