Tether CEO Warns AI Spending Faces $5.5T Repricing Risk

Published by James Harris on July 4, 2026July 4, 2026

Paolo Ardoino published a warning on X this week that the AI infrastructure buildout is structurally misaligned across four dimensions: token pricing versus actual compute costs, profitability timelines that don’t match investment horizons, capital maturity mismatched against asset lifespan, and open-source AI eroding the revenue base that justifies the spending. The Tether CEO is not a disinterested observer, but the numbers he’s pointing at are real.

The scale of the bet is difficult to overstate. JPMorgan raised its estimate for global AI-related capital spending through 2030 to $5.5 trillion in its midyear outlook, with hyperscaler capex projected to exceed $1.1 trillion in 2027 alone. Microsoft plans to spend roughly $190 billion in 2026, a 61% increase year-over-year. Goldman Sachs estimates Meta, Microsoft, Amazon, and Alphabet will collectively spend $5.3 trillion on capital expenses between 2025 and 2030. The problem is that the average company deploying AI is spending $11.5 million annually and cannot demonstrate a clear return, while Information sector growth in the US slowed to 1.5% in Q1 2026, down from 3.2% the prior quarter. For crypto markets, this matters because the same institutional desks that built exposure to Nasdaq-correlated risk assets are the ones now sitting on AI infrastructure positions that haven’t paid off.

The open-source pressure is already showing up in corporate behavior before any formal repricing: Amazon quietly dropped its internal AI usage leaderboard, Uber burned through its entire 2026 AI coding budget in four months, and Meta cautioned roughly 6,000 staff about rising costs.

The AI chip supply chain sits underneath all of this. Hardware that depreciates in three to five years, financed by $4.1 trillion in projected debt, against revenues that haven’t materialized, is the structural mismatch Ardoino is describing. IDC projects that 70% of leading AI adopters will use multiple models by 2028, which would compress margins across the entire stack. The Bank for International Settlements named AI investment as one of three primary economic risks, with its Asia-Pacific chief warning that any correction could move faster than past banking crises. Wedbush’s Dan Ives pushes back, calling it an arms race no major player can exit, and JPMorgan forecasts operating cash flow above $900 billion by 2027, but those projections assume the revenue side eventually catches the cost side.

Great Hill Capital’s Thomas Hayes offered the most testable near-term prediction: one or more major hyperscalers may announce reduced capital spending in the upcoming earnings season. If that happens, the mismatch Ardoino outlined stops being a theoretical concern and becomes a market event with consequences well beyond the AI sector.