AI Is Eating Software: Reshaping the Infrastructure Landscape

“AI is eating software.” This stark declaration from a16z’s Jennifer Li, echoing Martin Casado, encapsulates the profound disruption currently reshaping the technology landscape. It signals a paradigm shift where the very tools and foundations of software development are being re-architected, presenting both unprecedented challenges and immense opportunities for founders, VCs, and AI professionals.

In a recent a16z podcast, Erik Torenberg spoke with General Partners Martin Casado, Jennifer Li, and Matt Bornstein about the evolving definition of infrastructure in the age of artificial intelligence. They posited that AI models are emerging as a critical fourth pillar of infrastructure, joining the established triumvirate of compute, storage, and networking. This isn't merely an additive layer; it fundamentally changes the way computers are programmed and reshapes the entire software stack.

The panelists highlighted that infrastructure, at its core, is "the stuff you use to build the stuff." In this context, AI models represent a new foundational layer, demanding rethinking of traditional memory, latency, and programming paradigms. Martin Casado emphatically stated, "From an application standpoint, we've abdicated logic." This radical shift means that applications are increasingly relying on AI to generate core logic, moving beyond merely abstracting resources.

This transformation is also driving a significant evolution in who holds purchasing power. The rise of the "technical buyer", developers, data scientists, analysts, and cybersecurity professionals, means that marketing and sales motions increasingly resemble consumer-driven adoption. Jennifer Li noted that low-code solutions, long a promise, are finally realizing their potential through natural language interfaces, empowering individuals with good ideas, regardless of their deep computer science background, to build sophisticated applications.

Comparing the current AI wave to past supercycles like the internet and mobile, the discussion emphasized that new infrastructure always creates massive market expansion and new user behaviors. Existing companies, often tied to old behaviors and sales motions, struggle to adapt, creating white space for new challengers. Matt Bornstein characterized this AI surge as "by far the biggest thing that I've seen happen sort of in my life."

Defensibility in this new era isn't about traditional moats. It’s about navigating an industry that is simultaneously expanding and consolidating. While building robust AI products is inherently challenging, the true long-term value lies in deep technical expertise, particularly in "context engineering." This involves understanding how to feed the right data and context into models to achieve optimal performance and reliable outcomes. As Martin Casado articulated, the fundamental process of articulating and specifying a problem for a machine will not disappear, nor will the need for skilled professionals to do it. It requires a pragmatic, non-blinkered approach, rather than succumbing to the anthropomorphic fallacy that AI will solve all problems autonomously.