a16z Partner straightforwardly states "Vibe Code Myth Debunked": Not all software is suitable for AI programming

Venture capital firm Andreessen Horowitz (a16z) general partner Anish Acharya stated in the latest episode of the “20VC” podcast that the idea that all software can be built using “Vibe Code” (AI-assisted coding) is “completely wrong.” He pointed out that companies should not indiscriminately apply AI programming tools to every business module, as this not only offers limited savings but could also amplify systemic risks.

Acharya explained that software costs typically account for only 8% to 12% of a company’s total expenses. Even with full adoption of AI-assisted coding, the maximum savings would be about 10% of software expenditure, having a very limited impact on overall cost structure. Using such tools in critical business modules like ERP, payroll systems, or CRM could instead lead to higher hidden costs due to security, stability, and compliance issues.

He bluntly stated that current generative models possess powerful innovation capabilities and are better suited for driving core business, creating new products, or improving operational efficiency, rather than “rebuilding highly mature enterprise software systems.” He also named vendors like Salesforce, Microsoft, Oracle, and SAP, arguing that these systems should not be simply replaced.

Acharya also warned that the market is overhyping the narrative that “AI will rewrite all software,” a perception that significantly diverges from reality. “Some companies will inevitably be eliminated, but claiming the entire software industry will be replaced by AI is an inflated illusion.”

Notably, his comments come after a significant correction in tech stocks. Previously, Anthropic launched AI tools aimed at the legal industry, sparking investor concerns about the impact on traditional software models, which dragged down related sectors.

Regarding the long-term value of AI, Acharya shares a similar view with well-known investor Vinod Khosla, emphasizing that one should not focus solely on stock price fluctuations. Khosla has stated that as long as API call volumes remain high, genuine demand persists, and the so-called “AI bubble” is a misconception.

This discussion about Vibe Code and the boundaries of AI programming is also seen as an important indicator for enterprise technology strategy choices in 2026.