The Economics of AI Usage and What's Next For SaaS | Benedict Evans on a16z
a16z · 1:00:32 · 1 months ago
AI is currently in a transformative "hype" phase similar to the early internet or mobile eras. While it is generating significant excitement, the industry is still in a period of extreme, temporary disequilibrium. Long-term, foundation models will likely become cheap, commoditized infrastructure, while the true economic value will shift up the stack to specialized software applications that eventually become as invisible and "magical" as modern internet connectivity.
-
Coding agents — these have achieved clear product-market fit because software developers immediately found ways to use them for building more software .
-
Commoditized models — foundation models will likely become generic infrastructure, shifting the true economic value to the application layer where the actual work happens .
-
Supply imbalance — the market is currently experiencing a temporary, extreme shortage of capacity that is driving high prices and massive capital expenditure .
-
Chatbot limitations — current chat interfaces are early, basic tools, not the final form that most commercial software will eventually take .
-
Historical patterns — like the early PC or internet eras, we are seeing massive excitement before the industry settles on clear winners or standardized business models .
-
Enterprise adoption — businesses are still trying to bridge the gap between interesting technology and practical, daily-use workflows .
-
Capital limits — current massive infrastructure spending is unsustainable in the long run, and financial gravity will eventually force a correction .
-
How do organizations transition from testing AI tools to integrating them into daily operations?