Many AI companies are financing rapid data‑center expansion through large borrowings, creating rising costs for equipment, metals and energy inputs.

Debt and rising build costs

Two years ago, constructing a given AI data center might have cost $1 billion; today a similar build can reach $3 billion, and may rise to $6 billion.

That dynamic means the current AI investment wave relies heavily on debt and long‑term expectations that future revenue will recoup today's capital expenditures.

Sustainability and profitability concerns

The thesis often presented is simple: invest $50 billion now and later capture market share to recover costs, yet historically 8 out of 10 such firms never reach profitability.

When investors reassess those odds, funding can evaporate quickly and growth plans become unserviceable under heavy leverage.

Competitive pressure from cheaper chips

Dimmao argues a tipping point may arrive when Chinese firms produce chips domestically on local lithography, delivering similar performance at a fraction of current costs.

If equivalent hardware appears for $5 billion where Western builds cost $50 billion, existing investments in capacity could be rendered economically obsolete.

Market fallout scenarios

In such a scenario, valuations tied mainly to infrastructure could collapse: a company once valued at $50 billion might fall to $5 billion or less, while leading chip stocks could slump by around -85%.

Bankruptcies would likely follow, with banks liquidating data centers at steep discounts; assets written down from tens of billions to single‑digit prices could trigger a multi‑year recession.

Timing and broader implications

The collapse could begin when investors conclude they will not recover capital and withdraw further financing; this may even precede broader technological shifts in chip costs.

Over roughly a decade, however, the sector could still mature and deliver substantive improvements to science, business and everyday life after the adjustment period.