Venture’s biggest questions bubbling up

Five-trillion dollars: the estimated funding required for digital infrastructure and power underpinning the future of AI. Billion-dollar deals for computing power feature in the news almost daily and datacenter campuses may soon cover the area of Manhattan. The industry is building AI infrastructure at a relentless pace. Everyone is asking the same question, “is this a bubble?”.

This is arguably the most important question in the market today. If this furious build-out turned out to be speculative, "building too much, too quickly for use cases that have not arrived yet", a subsequent pop could trigger a crisis of confidence, leading to significant equity market losses and a tightness in private markets that could rival the 2022-2023 downturn. Further, without a handle on the true size of the infrastructure underpinning AI, its impossible to properly size the AI application market.

Our analysis, however, suggests the build-out is both rational and supported by highly favorable early fundamentals. What we’re experiencing is a logical response to the core insight of modern AI: the more computing power you apply, the better the model (or system) performs, and that next step up in performance might unlock not just automating route work but a cure to cancer or nuclear fusion.

Further, demand for AI computing is not just theoretical. The hyperscale compute providers persistently describe demand as "significantly ahead of supply" and their own businesses as "compute starved". Companies like Google and Meta are deploying AI to billions of users and finding incremental revenue opportunities even at their $100bn+ revenue scale, almost immediately validating their headline-grabbing capex figures. Simultaneously, a new wave of AI-native startups is generating real revenue: going from $5bn to $47bn in ARR in the last year.

The (general) rationality of this infrastructure buildout lays the foundation for the success of the entire AI infrastructure stack, from AI cloud compute all the way up to model observability. We believe that the AI infrastructure market is at the early stages of a structural expansion that will create an AI Cloud Computing market worth ~$450bn and support many significant businesses all the way up the stack.

Read the rest on our website 🔗