About
BubbleTracker shows what’s being spent building AI infrastructure in 2026, and what has to come back to justify it. All figures are sourced from institutional research and updated periodically. The site doesn’t have an opinion about whether this is good or bad. It just shows the numbers.
New here? How to read this explains how the numbers are chosen and how each chart works, and the sources page lists every reference.
Common questions
- Is AI spending really this large?
- Yes. Goldman Sachs projects 2026 AI infrastructure capex at about $765 billion across the major hyperscalers, and other institutions such as Bain & Company and Morgan Stanley publish figures in a similar range. These are projections and scenario models, not audited actuals.
- Are you saying AI is bad?
- No. The site doesn't take a position. Whether this spending is justified is a question economists, analysts, and investors disagree about. The site just shows the numbers.
- What is capex?
- Capital expenditure: money spent on physical infrastructure, like data centers, chips, and power systems, rather than on salaries or day-to-day operations. When a tech company builds a data center, that's capex.
- What is GDP?
- Gross Domestic Product: the total value of everything an economy produces. When analysts say AI capex is driving GDP growth, they mean the building of AI infrastructure is showing up as a significant portion of measured economic output.
- How often is the data updated?
- It depends on the panel. The spending counter and stock charts move on their own; equipment figures update each quarter; longer-range projections update only when new institutional research is published. Each figure shows its own as-of date.
- Who made this?
- An independent side project by Aline Aronsky. I am not an economist and I do not calculate any of these figures; I collect them from named institutions and lay them out clearly. No funding or ads. I built it because the scale of the numbers surprised me and I wanted to see them in one place.