Analysis
AI's infrastructure buildout is projected to hit $11 trillion in cumulative capex by 2030, according to a SemiAnalysis estimate cited in a sharp new column by The Register -- and we think the more interesting number is how much of that bet is actively eating the revenue streams meant to repay it.
The column's strongest point isn't speculative: Microsoft and OpenAI's own internal emails reportedly acknowledge that AI is accelerating an "ad-revenue doom loop," siphoning the advertising dollars that fund the content AI models are trained on in the first place. Separately, AI-generated Rust reimplementations of Microsoft Office and Adobe Creative Suite -- billed by their creator as "clean-room" clones -- show how easily AI can now reproduce the lock-in that funds enterprise software margins.
“Room for disagreement: Gartner's view, that there isn't yet enough ROI to justify AI's growth, assumes today's monetization is the ceiling rather than the floor.”
Nvidia's practice of renting back unsold capacity from neocloud partners at fixed prices is the same backstop mechanism Pulse has tracked in its $500 billion financing push -- a way to keep demand looking real even as datacenter chips that normally depreciate over five years keep earning the same or more revenue well past their expected shelf life.
Room for disagreement: Gartner's view, that there isn't yet enough ROI to justify AI's growth, assumes today's monetization is the ceiling rather than the floor. Cheap, efficient Chinese models undercutting Western pricing could just as easily be read as proof the cost curve is bending faster than the capex curve, not that the whole bet is broken -- and $11 trillion projected out to 2030 is a SemiAnalysis estimate, not a committed number.
Our read: the funding gap the column can't close -- "nobody knows" where the money comes from -- is the same gap every AI infrastructure LP should be pricing into neocloud debt right now, not a rhetorical flourish. The column also notes Beijing is treating cheap, widely distributed AI models as a deliberate instrument of global influence, which means the price pressure on Western labs' revenue isn't a side effect of competition -- it's a policy choice by a rival state, and it will keep showing up in every pricing deck AI infrastructure financiers build from here.
