# a16z, Anish Acharya: Is SaaS Dead in a World of AI? | Who Wins the Dev Market: Cursor or Claude Code Source: https://www.youtube.com/watch?v=Aq0JSbuIppQ Recap page: https://rapidrecap.app/video/Aq0JSbuIppQ Generated: 2026-02-09T15:34:43.419+00:00 --- ## Quick Overview Anish Acharya asserts that the narrative of completely rebuilding all software with AI is incorrect, arguing that traditional enterprise revenue remains sticky because pointing the innovation bazooka at rebuilding payroll or ERP only saves 8 to 12% of spend, while application layer companies will capture significant value by aggregating specialized foundation models and decreasing switching costs for customers who are currently 'hostages' of legacy SaaS providers. **Key Points:** - Anish Acharya strongly disagrees with the idea that the whole market will 'vibe code everything,' stating that rebuilding core systems like payroll or ERP only impacts 8 to 12% of enterprise spend. - The cost of transitioning between SaaS providers is dramatically lowered by coding agents, which decreases switching costs and turns customers from 'hostages' into less captive users. - Incumbent SaaS companies are capable, noting that 75% of public SaaS companies have raised prices since ChatGPT was released, with the mean increase being 8 to 12%. - Startups will win in native categories that did not exist before the current product cycle, whereas incumbents like Microsoft will improve existing product lines like word processors. - Foundation model providers are largely substitutes for each other, creating value for application layer companies that aggregate these specialized models, such as Cursor orchestrating both Gemini and Claude for coding. - The 'weird wins' in AI involve products that touch on core aspects of humanity, like companionship (e.g., Replica, Grok), which large corporations are uncomfortable building due to internal committee structures. - For power users, AI products command significantly higher subscription rates than previous consumer ceilings, with Grok at $300/month and ChatGPT at $200/month, plus consumption revenue. **Context:** The conversation features Anish Acharya, a General Partner at Andreessen Horowitz (a16z) leading consumer and fintech investing, discussing the future of Software as a Service (SaaS) and the impact of generative AI with the host. They explore topics ranging from optimal locations for building technology companies (SF vs. London/Tel Aviv) to the evolving nature of venture outcomes, defensibility, and the role of AI in enterprise software, contrasting the potential of foundation models versus specialized application layers. ## Detailed Analysis Anish Acharya dismisses the notion that AI will entirely replace existing software, arguing that the marginal utility of pointing the 'innovation bazooka' at rebuilding core systems like ERP or CRM is low, saving only 8 to 12% of IT spend. He suggests that this spend will instead be used to extend core business advantages or optimize the other 90% of non-software spend. Acharya highlights that incumbents are not weak, as 75% of public SaaS companies have raised prices since the release of ChatGPT. A key disruptive force is the reduction in switching costs due to coding agents, which lowers the risk associated with migrating between providers like SAP and Oracle, thereby turning 'hostages' into customers. In terms of market winners, history suggests incumbents will improve existing categories, but startups will own native categories created by the new product cycle where no incumbent exists. In the AI stack, application layer companies gain value by acting as aggregation layers over specialized foundation models (e.g., Cursor combining Gemini and Claude for coding), as models are becoming substitutes in general tasks but specialists in niche areas. Acharya also champions 'weird wins' in areas touching on core human experiences, such as companionship products (like Replika or Grok), which large, risk-averse corporations avoid building. He dismisses the idea that this is a bubble because supply (inference capacity) is being built directly in lockstep with demand (topline growth), and customers are paying higher prices, unlike previous bubble periods characterized by oversupply and price compression. Finally, he notes that influence is the new sales and marketing, and power users are now paying 10x historical consumer price ceilings, allowing companies to justify high acquisition costs for converts. ### Geographic Strategy for Startups - SF offers an enormous network effect for builders and a selection bias toward singular focus, while Tel Aviv forces immediate global thinking due to its small domestic market size, unlike London which might satisfy sufficient $3-5 billion outcomes domestically. ### Venture Outcomes and Scale - Building a trillion-dollar company requires specific initial assumptions; while $3-5 billion is an extraordinary enterprise outcome, it is insufficient if the intention is trillion-dollar scale. ### SaaS Market Health and Pricing Power - The 'SAS apocalypse' narrative is flawed; 75% of public SaaS companies have raised prices since ChatGPT, indicating product-market fit allows for price increases, contradicting the idea that revenue is not sticky. ### Disruption and Switching Costs - Coding agents dramatically lower the complexity and risk of systems integration, which is the primary mechanism for decreasing customer 'hostage' situations in legacy enterprise software. ### Foundation Models vs. Application Layer Value - Because foundation models are innovating roughly in lockstep and specializing, value accrues to app companies that aggregate them for users needing multimodal orchestration (like coding) or specialized feature surface that model companies prioritize less. ### Defensibility and Moats in the AI Era - Traditional moats like networks remain the gold standard, but proprietary, live data sets (e.g., health data) provide a powerful moat by allowing commodity models to outperform cutting-edge models lacking that data. ### Margins and Power Users - The distortion in AI markets is healthier than the 2021 subsidy (empty calories from ad spend); current subsidization via free trials converts to high-paying power users who pay 10x historical consumer ceilings ($200-$300/month), meaning customer acquisition cost (CAC) for converts is well-invested. ### UI Paradigm Shift - Intent-based UIs like chat are optimal for high-agency people, but browse-based interfaces remain necessary because many consumers want to 'spend time,' not just 'save time,' suggesting chat is overstated for general consumer use.