Ubiquitous local AI agents that can script any service and reverse-engineer APIs fundamentally threaten the SaaS recurring revenue model. If software lock-in becomes impossible, business models may shift back to selling expensive, open hardware as a one-time asset, a return to the "shrink wrap" era.

The open vs. closed debate overlooks a key strategic threat: frontier model companies could offer their smaller, older, cheaper models as fine-tunable products. This would directly compete with the primary use cases for open-source models today, potentially reshaping the entire ecosystem.

Creating frontier AI models is incredibly expensive, yet their value depreciates rapidly as they are quickly copied or replicated by lower-cost open-source alternatives. This forces model providers to evolve into more defensible application companies to survive.

As enterprises replace expensive proprietary models with cheaper open-source alternatives, frontier labs like OpenAI and Anthropic face an existential threat. Their strategic response could be to lobby for regulations that effectively make open-source models illegal, creating a protective moat.

The assumption that startups can build on frontier model APIs is temporary. Emad Mostaque predicts that once models are sufficiently capable, labs like OpenAI will cease API access and use their superior internal models to outcompete businesses in every sector, fulfilling their AGI mission.

Unprofitable frontier AI companies are expanding into application-layer verticals like drug development as a defensive strategy. They aim to build defensible, high-margin SaaS revenue streams to prove their business model to investors before their core inference and training services are fully commoditized by cheaper open-source alternatives.

While some vendors push self-hosting an open-source model as a safer alternative, Anthropic argues the real business risk is falling off the intelligence frontier. As model capabilities improve exponentially, the competitive advantage gained from using the most advanced models will far exceed the perceived benefits of a static, self-hosted system.

The business model for AI agents fundamentally shifts the value proposition from selling a tool (license) to selling an outcome (automated work). This allows vendors to tap into operational or labor budgets, not just IT budgets, unlocking a new price-for-value equation and exponentially larger contract sizes.

Enterprises distrust AI vendors policing themselves, creating a need for independent security firms. Crucially, these firms gain access to sensitive historical agent data that companies refuse to give to 'data hungry' labs like OpenAI, creating a powerful, non-technical moat.