Paolo Ardoino, the CEO of the world’s largest stablecoin issuer, laid out a detailed critique of Big Tech’s AI infrastructure spending on July 4, identifying what he called structural risks baked into the current capital expenditure frenzy. His core argument is straightforward. Companies are pouring unprecedented sums into data centers, GPUs, and power capacity while profits remain stubbornly distant and open-source competitors keep chipping away at any pricing power.
Four mismatches that should worry everyone
Ardoino’s warning zeroes in on four specific economic mismatches in the AI sector. First, compute token prices don’t accurately reflect their true underlying costs. Second, there’s a gap between the massive upfront investments required and the timeline to profitability. Third, the capital maturity timelines are misaligned with hardware lifespans. AI chips depreciate in roughly 3 to 5 years, but the debt and equity structures financing them often assume much longer payback periods. Fourth, open-source AI models are eroding the commercial revenues that were supposed to justify all this spending.
The numbers are staggering
JPMorgan projects that global AI-related spending could hit $5.5 trillion by 2030, a figure the bank revised upward in June 2026. Goldman Sachs estimates that just four companies, Microsoft, Meta, Amazon, and Alphabet, will account for roughly $5.3 trillion in spending between 2025 and 2030.
